P. 1
Design distillation

Design distillation

|Views: 16|Likes:
Published by jcruz_255081
destiling process
destiling process

More info:

Published by: jcruz_255081 on Jan 20, 2013
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less










Edward Arnold

Design of Distillation Column Control Systems
8 Instrument Society of America 1985 All rights reserved
Printed in the United States of America In preparing this work, the author and publisher have not investigated or considered patents which may apply to the subject matter hereof. It is the responsibility of the readers and users of the subject matter to protect themselves against liability for infringement of patents. The information contained herein is of a general educational nature. Accordingly, the author and publisher assume no responsibility and disclaim all liability of any kind, however arising, as a result of using the subject matter of this work. The equipment referenced in this work has been selected by the author as examples of the technology. No endorsement of any product is intended by the author or publisher. In all instances, the manufacturer's procedures should prevad regarding the use of specific equipment. No representation, expressed or implied, is made with regard to the availability of any equipment, process, formula, or other procedures contained herein.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written permission of the publisher:
Instrument Society of America 67 Alexander Drive P. 0. Box 12277 Research Triangle Park North Carolina 27709 United States of America

ISBN 0-7131-3551-4
Library of Congress Cataloging in Publication Data Buckley, Page S. Design of distillation column control systems. Includes indexes. 1. Distillation apparatus. 2. Chemical process control. 1. Luyben, William L. 11. Shunta, Joseph P. 111. Title. TP159D5B83 1985 660.2'8425 84-27813

ISBN 0-7131-3551-4

Book design by Raymond Solomon
Production by Publishers Creative Services Inc., New York



his is a book about the design of disullation column control systems. It is written primarily fiom the standbint of an engineering design organization, and is based on years of experience with large design projects as well as on personal plant experience. Most new investment dollars go into new or modemized facilities, and it is in the design phase of projects for these facilities that the most opportunities occur and flexibility exists to influence process control. Consequently this book is aimed primarily at design personnel. It is our hope, however, that it will also be usell to those who have to operate or troubleshoot existing plants. P r I is an introduction, including a perspective on control and a brief at review of fundamentals of &stillation, with emphasis on topics that will be of interest to the control engneer rather than to the column design engineer. The distillation review, it is hoped, will be particularly usell to nonchemical enpeers. P r I1 of the book, on concepts and configurations, discusses some practical at aspects of distillation control. Once the requirements for a particular column in a particular process are understood, design engineers must make at least a preliminary choice of equipment arrangements and control system configuration. In this section we have mostly avoided the use of mathematics and control theory. It is our hope that our discussions of equipment and control system arrangements will be usell to process engineers, production supervisors, maintenance engineers, and instrument engineers seeking guidelines, alternatives, and perspectives. P r I11 focuses on the quantitative design of distillation control systems. at It is aimed at professional control engineers and any others concerned with the numerical definition and specification of control system performance. Probably the most important development in process control system design since about 1950 was the evolution of a substantial body of theory and mathematics, plus a large catalog of control system studies. Together, these permit quantitative design of most process control systems with a considerable degree of multivariable control. It is the purpose of this book to indicate the range of this technology, which has been developed for distillation control, to the point where it can be economically and reliably used for design. The ultimate economic advantages include lower plant investment (particularly in tankage), lower operating costs, and closer control of product quality. For the most part, we have stayed with

the modest theory of single-input, single-output (SISO) systems presented in f previous books: Techniques o Process Control by P. S. Buckley (Wiley, 1964) and Process Mohling, Simulation, and Control fm Chemical Engineers by W. L. Luyben (McGraw-Hill, 1973). This kind of theory and mathematics not only is adequate for noninteracting systems and for simple interacting systems, but it has the advantages of requiring minimum formal training and of permitting low design costs. “Modernyyor “optimal” control techniques are mentioned only briefly here because their use on real, industrial-scale distillation columns has been quite limited to date. These techniques are still being actively researched by a number of workers, and it is hoped that they eventually will be developed into practical design methods. As of the date of the writing of this book, however, these mathematically elegant methods are little used in industry because of their complexity, high engineering cost, and limitation to relatively loworder systems. Simulation techniques also are not covered since there are several texts that treat this topic extensively. In the past five years, we have witnessed the introduction and proliferation of microprocessor-based digital controls of various sorts that are intended to replace analog controls. In fact, most of the newly installed control systems are of this type. In addition, we are seeing more control being implemented in process control computers. Sampled-data control theory has taken on new importance because of these developments and so we have included a chapter on previous work we have done in this area as it relates specifically to distillation columns. The concepts we present are quite basic as opposed to the recent advances in adaptative, multivariable, and predictive control, but we hope they will benefit those interested in synthesizing single-loop sampled-data controllers. Many thanks are due our associates in the Du Pont Company, particularly R. K. Cox, and throughout the industrial and academic communities for helpful comments and suggestions. Many of the concepts presented in this book have been vigorously debated (over untold cans of beer) during the Distillation Control Short Courses held at Lehgh University every other spring since 1968. We also wish to thank Leigh Kelleher for major assistance in formatting and editing, Arlene Little and Elaine Camper for typing, and Ned Beard and his Art Group for preparing the illustrations.

Pade S . Buckley W l i m L. Luyben ila Joseph P. Shunta


n this work an effort has been made: (1) to use symbols and u i s commonly nt employed by chemical engineers, ( 2 ) to define each symbol in a chapter when the need for that symbol arises, and (3) to keep symbols and units as consistent as possible from chapter to chapter. A few symbols, however, have different meanings in different parts of the text. The list that follows contains the major symbols and their usual meanings: transportation lag or dead time, usually seconds or minutes area, ft2 bottom-product flow, mols/min C specific heat, pcu/lbm "C acoustic capacitance, fi5/lbf C control-valve flow coefficient, gallons per minute of water flow Cll when valve pressure drop is 1 psi D diameter, feet, or top-product flow rate fi-om condenser or condensate receiver, mols/min E Murphree tray efficiency cycles/minute or cps f F feed rate to column, mols/min ft Ibmass mass-force conversion factor, 32.2 -B c sec' Ib force local acceleration due to gravity, ft/sec2 BL heat-transfer f m c&cient, pcu/sec h i l "Cfi? head of liquid or liquid level, feet H fl(has different meaning when used as subscript) i static gain K 1 distance, feet external reflux, mols/min L o liquid downflow in column, mols/min LR M liquid holdup, mols M w molecular weight pressure, psi P P pressure, lbf/fi?, or atmosphere, or mm Hg
a A B




pound centigrade u i s (heat required to heat one pound of water nt 1°C) 5 vapor pressure of pure component, speciesj heat flow, pcu/sec, or 4 fraction of feed that is liquid (molar basis) flow rate, ft3/sec or ft3/min Q R reflux ratio, LJD 5 Laplace transform variable t time, seconds or minutes T temperature, degrees Celsius or Kelvin, or sampling time interval in sampled-data control systems U overall heat-transfer coefficient, pcu/sec ? "C f vapor flow, mols/min, or V volume, ft3 VI- volume in tank corresponding to level transmitter span, AHT 29 weight rate of flow, usually Ibm/sec weight, lbm W mol fraction more volatile component in a liquid X mol fiaction more volatile component in a vapor Y z z-transform variable, or mol fraction more volatile component in feed ZF Z acoustic or hydraulic impedance, Ibf sec/fi5 a relative volatility specific heat ratio, or Y activity coefficient M -liquid-level transmitter span, feet, corresponding to full-scale output l E difkrence between set-point signal and signal from measurement device damping ratio in a quadratic expression 5 arbitrary input signal 6, arbitrary output signal 60 h latent heat of vaporization, pcu/lbm molar latent heat of vaporization, pcu/mol Am viscosity, lbmlft sec CL = centipoise/ 1488 density, lbm/ft3 P 7 time constant, usually seconds or minutes enthalpy, pcu/lbm fi-equency, radians/unit time 0







quadratic bottom of tower reset, or reflux


light component or key heavy component or key f feed f feedforward i inlet i arbitrary tray location or component outlet 0 S stripping section set point SP steam st C controller distillate (top product) D OL open loop (used outside of brackets)


Symbols on Illustrations


CC or

composition control flow control liquid level control pressure control temperature control high signal selector low signal selector high signal limiter low signal limiter cooling water
- -


Individual barred terms (e.g., V, P) indicate average values. Combined barred terms [e.g., HG(z)] have special meaning in sampled-data control systems (see Chapter 21).

K,G,(s) K,G,(s) K,,G,(s) KpGp(s)

measurement transfer function controller transfer function control valve transfer function process transfer function

Preface Part I INTRODUCTION Chapter 1 Strategy for Distillation-Column Control 1.1 Distillation Control Objectives 1.2 Arrangements for Maw-id-Balance Control 1.3 Fundamentals of Composition Control 1.4 Compensation for Various Disturbances 1.5 Startup and Shutdown 1.6 Control System Design Philosophy 1.7 Procedure for Overall Control System Design 1.8 Column Design Philosophy and Control System Design 1.9 Existing Columns-Typical Practices and Troubleshooting 1.10 Conventions Followed in This Book 1.11 Literature

pwe mtt
1 3 3 6 11 12 13 14 19 19 20 21 22

Chapter 2 Fundamentals of Distillation 2.1 Introduction
2.2 2.3 2.4 2.5
Tray Hydraulics Vapor-Liquid Equilibrium Fundamentals Graphical Solution Techniques Effects of Variables

25 25 28 30 49 65 67 69 69 70 72 80 84

Part II CONCEPTS AND CONFIGURATIONS Chapter 3 Overhead System Arrangements 3.1 Introduction 3.2 Types of Condensers 3.3 Atmospheric Columns 3.4 Vacuum and Pressure Columns-Liquid Product 3.5 Pressure Columns-Vapor Product 3.6 Miscellaneous Pressure-Control Techniques 3.7 Gravity-Return Reflux Versus Pumped-Back Reflux 3.8 Control Techniques with Air-Cooled Condensers

90 99


c ne t o t ns

3.9 ‘Tempered” Versus Once-Through Coolant 3.10 Level Control of Condensate Receiver and Required Holdup

100 100 109 109 110 114 116 117 119 122 126 130 133 137 137 137 140 141 143 144 145 149 151 153 153 154 157 166 166 169 169 169

Chapter 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7

Column-Base and Reboiler Arrangements

Introduction Vertical Thermosyphon Reboilers Flooded Thermosyphon (Steam-Side) Reboilers Forced-Circulation Reboilers Flooded-Bundle Kettle Reboilers Internal Reboilers Steam Supply and Condensate Removal 4.8 Required Holdup for Level Control 4.9 Miscellaneous Column-Base Designs 4.10 Miscellaneous Reboiler Designs

Chapter 5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9

Feed System Arrangements
General Comments Feed Flow Control Feed Temperature Control Feed Enthalpy Control Feed Tray Location Feed Tank Sizing Feed System for Double-Column Systems Feeds with Makeup/Purge to Tankage Feed Systems in Sequences of Columns With and Without Recycles

Chapter 6 6.1 6.2 6.3 6.4 6.5

Level Control and Feedforward Options
Introduction Material-Balance Control in Direction Opposite to Flow Material-Balance Control in Direction of Flow Unfavorable Control Schemes Unreasonable Control Schemes

Chapter 7 Control of Sidestream Drawoff Columns 7.1 Introduction 7.2 Side-Draw Columns with Large Sidestreams



7.3 7.4 7.5 7.6 7.7

Side-Draw Columns with Small Sidestreams Composition Control of Side-Draw Columns An Improved Approach to Composition Control of SideDraw Columns Prefiactionator Plus Sidestream Drawoff Column Other Schemes

170 170 174 176 180 181 181 181 182 183 186 189 189 189

Chapter 8 Minimizing Energy Requirements
8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8
Introduction Conservation Design Considerations in Heat-Recovery Schemes Multiple Loads Supplied by a Single Source Single Source, Single Load Split Feed Columns Combined Sensible and Latent Heat Recovery Energy Recovery by Vapor Recompression

Chapter 9 Application of Protective Controls to Distillation Columns
9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10 9.11 9.12 9.13
Introduction Overrides and Interlocks Implementation of Overrides Controllers in Override Circuits Anti Reset-Windup Feedforward Compensation with Overrides Overrides for Column Overhead System Overrides for Column-Base System Automatic Stamp and Shutdown “Idle” or Total Reflux Miscellaneous Overrides Design Considerations Overrides for Side-Draw Columns

193 193 194 195 199 200 202 205 208 211 213 214 21 7 220 229 229 229 230 231

Chapter I O Indirect Composition Measurements
10.2 10.3


Introduction Single-Tray Temperature Differential Temperature Differential Vapor Pressure

5 Column-Base Level Measurements 11.2 Tray Hydraulics 13.7 10.6 10.5 10.1 11.6 Control Valves 11.5 Ways of Designing Control Systems Kinds of Information Available Functional Layout of Control Loops Adjustment of Controller Parameters (Controller Tuning) Enhanced Control of Distillation Columns via On-Line Models 293 295 295 297 299 303 305 313 313 314 Chapter 13 Tray Dynamics-Material Balance 13.10 Control-Valve Split Ranging Part 111 QUANTITATIVE DESIGN OF DISTILLATION CONTROL SYSTEMS Chapter 12 Approaches to Quantitative Design 12.8 Temperature-Measurement Dynamics 11.1 12.4 Heat-Flow Computations 11.viii GmtenB 10.4 12.2 11.3 Pressure-Compensated Temperature Multicomponent Compositions Computed from Temperature and Pressure Measurements Double-Differential Temperature Average Temperature Composition Estimators 234 239 240 241 241 243 243 243 249 255 256 273 279 279 288 289 Miscellaneous Measurements and Controls Introduction Calculation of Distillation-Column Internal Reflux Temperature and Pressure Compensation of Gas Flow Meters 11.4 Mathematical Model for Combined Trays Chapter 14 Distillation-Column Material-Balance Control 14.7 Column AP Measurement 11.9 Flow and Flow-Ratio Conventions 11.3 12.1 Mathematical Model-Open Loop 14.3 Derivation of Overall Tray Equation 13.8 10.2 12.2 Control in the Direction of Flow 320 323 327 327 333 .1 Introduction 13.9 Chapter 11 1 1.

1 Introduction 17.2 Flooded Condensers-Open-Loop Dynamics 15.4 Partially Flooded Reboilers 15.2 Level Control of Simple Vessels 16.3 Pressure Control Via Vent and Inert Gas Valves 17.1 Introduction 18.4 Pressure Control Via Flooded Condenser 17.7 Column-Base Level Control Cascaded to Steam FlowControl 16.6 Column AI' Control Via Heat to Reboiler Chapter 1 8 Composition Dynamics-Binary 18.8 Column-Base Level Control Via Condensate Throttling from a Flooded Reboiler (Cascade Level-Flow Control) Chapter 17 Pressure and A P Control 17.5 Top and Bottom Level Control Combinations Chapter 15 Condenser and Reboiler Dynamics 15.3 Control in the Direction Opposite to Flow 14.c ne t o t ns ix 14.6 Column-Base Level Control Via Feed Flow Manipulation 16.2 Basic Tray Dynamics 18.4 Level Control of Overhead Condenser Receiver Via Reflux Manipulation 16.5 Pressure Control Via Condenser Cooling Water 17.4 Material-Balance Control in Sidestream Drawoff Columns 14.5 Column-Base Level Control Via Bottom-Product Manipulation 16.3 Feed Tray Dynamics Distillation 337 342 343 347 347 349 357 366 371 375 375 375 386 386 389 389 390 399 405 405 405 408 415 420 420 427 427 427 432 .5 Partially Flooded Reboilers for Low-Boiling Materials Chapter 16 Liquid Level Control 16.1 Liquid-Cooled Condensers with No Condensate Holdup 15.3 Level Control of Overhead Condenser Receiver Via TopProduct Withdrawal 16.1 Introduction 16.3 Reboilers-Open-Loop Dynamics 15.2 Heat-Storage Effect on Column Pressure 17.

5 Relative-Gain Matrix 20.X contents 18.7 Top-Tray and Overhead System Composition Dynamics Reboiler and Column-Base Composition Dynamics Inverse Response Overall Composition Dynamics 433 439 439 441 445 445 446 448 449 462 465 465 466 468 475 478 489 493 493 494 496 502 508 510 523 527 530 Chapter 19 Calculation of Steady-State Gains 19.4 18.6 Composition Measurement Location Chapter 21 Sampled-Data Control of Disti1lat.5 18.1 Introduction 21.5 Interaction Compensation 21.3 Servo and Regulator Control 21.2 Feedback Control of Composition 20.2 Control Algorithms 21.3 Interaction Compensation 20.2 Design Procedure 19.4 Column Operation Procedure 19.3 Exact R Procedure 19.5 Examples Chapter 20 Composition Control-Binary Distillation 20.6 18.6 Sampled-Data Control for Loops with Overrides Nomenclature Subject Index Author Index .m Columns 2 1.1 Introduction 20.4 Feedforward Control 2 1.1 Introduction 19.4 Feedforward Compensation 20.

2 Material balance control in direction opposite to flow Material balance control in dlrection of flow Overall material balance control in direction opposite to flow Overall material balance control with intermediate material balance control in direction of flow Distillation column with material balance control in direction of flow Nomenclature and conventions for typical dlstillation column Control variables for distillation column Schematic of typical sieve tray Vapor pressure and temperature measurement Temperature vs.19 29 31 32 33 34 35 36 37 38 41 42 44 46 47 48 49 50 .9 2.17 2. temperature Temperature vs.13 2.Figures 1. composition of binary mixture at constant temperature x vs y for binary mixture Bubble point and dew point at constant temperature Bubble point and dew point at constant pressure Isothermal flash Graphical representation of equation (2. composition of binary mixture at constant pressure Pressure vs.14 2.8 2.1 1.10 2.18 2.16 2.13) Relative volatility on x-y diagram Typical activity coefficients as functions of light component composition Typical homogeneous “maximum boiling” azeotrope Homogeneous “minimum boiling” azeotrope Heterogeneous azeotropes Simple distillation column xi 7 8 1.6 2.1 2.11 2.2 2.4 9 10 26 27 1.5 2.5 2. pressure for pure component Typical method of plotting vapor pressure vs.12 2.7 2.3 2.15 2.3 1.4 2.

18 Column pressure control with flooded condenser Gravity flow reflux (flow controlled) and distillate (level controlled) 92 .25 2.21 2.4 3.13 3.23 2.7 3. reflux ratio Minimum number of trays required at total r d w F&ures 52 53 54 55 56 59 61 62 63 64 3.27 2.28 2.15 Horizontal condenser.24 2.10 3.11 3. vapor in shell Vertical condenser.17 3.5 3.2 3.8 3.14 3.29 Material balance on stripping section Operating line of stripping section Material balance on rectifjring section x-y diagram showing both stripping and rerufying operating lines McCabe-Thiele dagram-stepping between VLE curve and operating lines to estimate number of trays required q-line on x-y dagram McCabe-Thele diagram for rating problem Column operation at minimum reflux ratio Costs vs. vapor in tubes Alternative overhead system for pressure column Air-cooled condenser Spray condenser Preferred overhead system for atmospheric column Alternative overhead system for atmospheric column Thermowell installation under vertical condenser Tempered coolant system Overhead system for vacuum or pressure column-large of inerts Overhead system for vacuum column-small amount 71 72 73 74 75 76 77 78 79 81 82 83 84 amount of inerts Alternative overhead system for pressure or vacuum columnsmall amount of inerts Alternative pressure control system Overhead system for pressure column-vapor product Alternative overhead system for pressure column-vapor product 85 86 88 89 90 3.16B Column pressure control by hot gas bypass 3.9 3.22 2.20 2.xii 2.3 3.26 2.6 3.12 3.1 3.16A Column pressure control by hot gas bypass 3.

17 Distillation column base with thermosyphon reboiler Relationship between vapor volume in tubes of thermosyphon reboiler.7 4. and base liquid level Flooded reboiler Flooded reboiler for boiling point materials Column base with forced-circulation reboiler Kettle-type reboiler with internal weir Protective circuits for tube bundle chamber in kettle-type reboiler Column base with internal reboiler-no Steam header configuration Improved steam supply and flow control system Level control of column base via bottom product throttling.3 4.23 3.4 4.19 3. surge tank with Sutro weir.11 4.6 4.25 3. cascade arrangement baffles or weirs Column base with isolated internal reboiler Column base design and arrangement for minimum holdup Reboiler piping arrangement for preferential boiling of reflux from lowest downcomer Arrangement for column base overflow into intermediate surge vessel .22 3.1A 4.10 4.1B 4.26 3.5 4.21 3. heat load.27 Liquid-vapor disengagement space built into condenser Gravity flow reflux system with ground-located surge tank for distillate Control of gravity reflux by throttling top product flow Control of gravity reflux flow rate by overflowing through Sutro weir and by throttling distillate flow Gravity-flow reflux. minutes T~ xiii 93 94 95 96 98 99 99 101 103 111 111 113 115 116 117 118 119 121 123 124 125 127 128 131 132 133 134 > 3-5 Undesirable piping arrangements for returning reflux to column Preferred piping arrangement for returning reflux to column Condensate receiver level control via distillate Proportional-only condenser seal pot level control via reflux flow Vertical thermosyphon-heat flux vs.24 3.16 4.13 4.15 4.F4qures 3. cascade arrangement Proportional-only level control system for column base Column base level control by steam flow manipulation. supply side liquid level 4.2 4.20 3.8 4.12 4.14 4.9 4.

overhead level control via top product.6 6.7 6.7 5.1 6.2 6. the toluene impurity content in the dlstillate producer is controlled by the reflux ratio. reflux drum level control via distillate.8 6.8 but with reflux ratioed to dstdlate Basic control scheme for column with sidestream drawoff Controls for liquid sidestream drawoff column Alternate control scheme for column with sidestream drawoff Scheme for control of sidestream composition Control of terminal composition (A) In the control system finally chosen. reflux drum level control via base level control via feed Material balance control in direction of flow.5 6.9 5. reflux drum level control via reflux. base level control via feed Distillate demand. overhead level control via reflux.1 7.XiV Fgures Feed system for distillation column 138 139 140 141 142 5.2 7. base level via feed Bottom product demand.6 5.9 7.5 5. base level control via bottom product Material balance control in direction of flow.3 6.6 Column feed systems with positive displacement pumps Column feed preheat via exchange with bottom product Column feed temperature control with economizer and preheater Column feed enthalpy control with economizer and preheater Column with multiple feed trays Feed system for a split column Feed system for split vacuum columns Feed systems for column in parallel Makeuptpurge feed systems Bottom product demand.10 6.4 7. base level control via bottom product Material balance control in direction of flow.3 5. reflux drum level control via reflux. base level control via feed 144 146 147 148 150 155 156 158 159 160 162 163 164 165 171 172 173 ol Bottom product demand.3 7. overhead level control via b i up. base level control via boilup Like Figure 6.8 5.4 5.1 5.2 5. (B) The five alternative sidestream tray positions and 175 176 .5 7. base level control via feed Distillate demand.4 6. reflux drum level control via dstillate.

10 9. single load source.11 9.18 9.19 9.1 8. single load-scheme feed.17 9. are shown in this blowup. xv 177 178 179 7. multiple loads 184 185 187 Scheme for establishing heat load priorities source.9 9.4 8.5 8.2 9.6 9.7 9.4 9.2 8.3 9.8 9. which regulate the benzene and xylene impurities in the sidestream drawoff.1 9.6 9. single load 2 188 190 191 196 197 197 198 198 203 204 205 206 208 209 214 216 216 218 219 219 Heat recovery via vapor recompression Median selector (J.14 9.20 9.7 7.5 9.16 9.13 9.Fipres their controls.21 D-scheme L-scheme Heat recovery scheme-single Heat recovery-single Heat recovery-single Heat recovery-split source.P.15 9.12 9.8 8.3 8. Shunta design) High h t e r Low limiter High limiter schematic Low limiter schematic Column base temperature control with AP override Anti reset-windup for cascade loops Impulse feedfonvard with PI controller and overrides Overrides for column overhead system High cooling water exit temperature override on condensate temperature control Overrides for column base system Scheme for protecting centrifugal pump against dead heading Effect of entrainment on overhead composition Entrainment override Limited utility override on feed Steam header pressure protective override Control scheme for balancing condenser and reboiler heat loads Hard and soft constraints Flow rate controls for composition control Feed flow system Low temperature overrides for drawoff valves 220 221 222 223 .

1 10.1 repeater .12 11.14 Typical gas flow purge system 11.6 11.19 Level measurement with two flush diaphragm transmitters and a summing relay 11.3 Steam valve overrides Override for minimum vapor flow up column Override for minimm liquid flow down column 224 225 226 232 232 235 237 245 245 251 255 257 259 260 261 262 244 265 267 268 269 270 270 271 272 273 274 DVP cell schematic Pressure compensation of temperature measurement Composition vs.22 9.11 Spechc gravity compensation of head measurement of liquid level 11. t and p Measurements needed for internal reflux computation Pneumatic hardware configuration for internal reflux computation Typical compensated gas flow metering scheme Heat flow computer for heat transfer Head-level relationship in a vessel Column base-reboiler manometer Schematic diagram of Isplacement-type level transmitter Internal damping chamber External damping for AP level measurement 11.4 11.9 11.2 10.23 9.3 10.17 Angled nozzle with dip tube 11.10 Velocity error in head measurement 11.8 11.5 11.24 10.16 Best gas flow purge system 11.13 Insufficient purge: transmitter lag in response to rapid rise in level InsufFcient purge: level transmitter erroneous response to rapid rise in pressure Improved gas flow purge system 11.2 11.1 11.20 Level measurement with flush diaphragm AP transmitter and 1.3 11.18 Level measurement with A' transmitter with double remote I seals 11.9. t and p Composition vs.7 11.15 11.

7 14.5 14. and predictor Single-loop with feedforward compensation Primary controller with optional enhanced control faturescascade control system Distillation tray schematic for flows and liquid elevations Preliminary signal flow diagram for tray material balance dynamics Material balance coupling with vapor and liquid flow Distillation column material balance Signal flow diagram for column material balance Signal flow diagram-condensate receiver Signal flow diagram--column base Signal flow diagram-material flow balance control in direction of 300 302 304 315 322 324 328 334 335 337 338 340 341 344 13. of f l d e d condenser 350 353 354 355 .1 12.4 14.1 14.2 14.27 Split ranging of large and small valve 11.3 14.3 15.2 Final signal flow diagram for P.6 Material balance signal flow-vapor sidestream drawoff sidestream drawoff Material balance signal flow diagram-liquid 345 Horizontal condenser with coolant in tubes and partially f l d e d on shell side First signal flow diagram for P of flooded condenser .F&%WJ xvii 281 282 284 285 286 287 290 291 11.9 15. PI controller.24 Effect of fluid velocity on step response 11. First reduction of signal flow diagram of Figure 15.22 Effect on step response of various annular fills 11.2 13.4 Signal flow diagram-material balance control in direction opposite to flow Rearranged version of Figure 14.1 13.1 15.25 Effect of annular fill on step response.3 Single loop with feedforward compensation.2 15.28 Split ranging reflux and distillate valves 12. derivative.23 Effect on step response of annular clearance 11.3 14. overrides.6 14.2 12.8 14. v = lO/sec 11.21 Effect of velacity on step response 11.26 Effect of annular clearance on step response 11.

5 15.10 Signal flow diagram for base level control cascaded to steam flow control 15.7 15.1 16.xviii 15.4 16.2 Reduced version of Figure 16. of flooded condenser Reduced signal flow diagram for w.10 Simplified treatment of heat storage effect on column pressure dynamics Preliminary signal flow diagram for column heat storage dynamics 407 408 .14 Reduced signal flow diagram for flooded reboiler 15.10 Preliminary signal flow diagram for column base level control via condensate throttling from a flooded reboiler 16.11 Final signal flow diagram for base level control cascaded to steam flow control 15.7 16.12 Signal flow diagram for base level control by direct manipulation of steam valve 15.6 15.6 16.9 First signal flow diagram for w.5 16.16 Reduced signal flow diagram for flooded reboiler for low boiling point materials 16.8 FiguVes 358 359 360 362 363 364 366 367 368 369 372 372 15.13 Preliminary signal flow diagram for flooded reboiler 15.8 16.15 Signal flow dagram for flooded reboiler for low bohng point materials 15.2 16.8 15.1 17.11 17. of flooded condenser Schematic representation of column base and reboiler holdup Preliminary signal flow diagram for heat transfer dynamics Partial reduction of Figure 15.3 16.7 Inverse response predictor for base level control via steam flow 16.9 Level control of simple vessel Signal flow diagram for simple level control system 376 378 384 388 391 392 396 397 398 402 403 PI level control cascaded to flow control Signal flow dagram for proportional-only condensate seal pot level control via reflux flow manipulation Signal flow diagram for proportional-only column base level control via feed flow manipulation Responses of PI averaging level control system with dead time to step change in outflow Base level control via steam flow control with inverse response and reboiler swell Partial reduction of Figure 16.

3 and 17.9 Reduced signal flow diagram of preliminary l a g r a m Partial signal flow diagram for reboiler dynamics Reduced form of signal diagram of Figure 17.1 19.9 17.5 18.2 18.00000 Reflux via reflux drum level control: bottom product via base level control 416 417 418 419 421 423 424 425 428 431 434 436 437 440 443 450 452 .3 17. significant inerts 17.4 Combined signal flow diagram for Figures 17.5 Partial signal flow diagram for column pressure control via manipulation of inert gas and vent valves Reduction of signal flow diagram of Figure 17.7 Signal flow diagram for column pressure control via manipulation of inert gas and vent valves when reboiler steam is flow or flow ratio controlled xix 409 410 410 410 412 412 413 414 17.7 19.5 17.6 17.14 Partially reduced version of Figure 17.3 18.4 18.Rippin-Lamb model for binary distillation column dynamics Effect on calculation of rectifying section when: A.2 Flows to and from basic tray Signal flow diagram for basic tray Signal flow diagram for feed tray Signal flow diagram for top tray and overhead system Partly reduced signal flow diagram for top tray and overhead system Signal flow diagram for reboiler composition dynamics Signal flow l a g r a m for . reboiler steam not flow or flow-ratio controlled.1I 17.18 Column P (base pressure) control via direct manipulation of steam valve 18.12 Partially reduced version of Figure 17. Guess for R is too small or B.10 Reduced form of Figure 17.8 17.16 Equivalent network for vapor flow and pressures in column 17.4 17. Guess for xD is too close to 1.17 Column AP (base pressure) control cascaded to steam flow control 17.7 17.15 Signal flow diagram for column pressure control via manipulation of condenser cooling water 17.1 18.13 Column pressure control via flooded condenser.6 18.11 Column pressure control via flooded condenser drainnegligible inerts and reboiler steam flow or flow-ratio controlled 17.13 17.Fkures 17.

5 Composition control of distillation column with feedforward compensation and decouplers 454 456 20.1 Tracking sampled-data control of X2 with set-point disturbance 1 .5 21.2 with decouplers Partly reduced signal flow diagram of Figure 20. v.1 Partial signal flow diagram for system with reflux manipulated by reflux drum level Signal flow diagram for system of Figure 20.7 20.4 Distillate via reflux drum level control: bottom product via base level control Distillate via reflux drum level control: boil up via base level control Distillate via reflux drum level control. vs x.7b Set-point change with compensators 2 1 .4b “Dual” and PID control for feed composition disturbance 2 1 .7a Sampled-data feedforward/feedback control loop Interaction compensation Set-point change without compensators 509 511 511 511 511 512 513 516 517 2 1.2 20.1 2 1.6 20. 7 ~ Disturbance in feed cornposition 21.4 20.8 21. R qvs.10 Tracking “dual” control in loop with overrides Conventional control of X2 with set-point disturbance 2 1.10 20.4 Partly reduced signaVflow diagram of Figure 20.xx 19.4a 466 467 469 470 472 473 476 485 486 487 488 496 498 499 503 504 505 507 yTvs. 4 ~ “Dual” and PID control for feed rate disturbance 2 1.3 21. R YT vs. bottom product via base level control Partial signal flow diagram for Figure 20.8 20.11 21.1 20.2 21.9 21.9 20.3 20.5 20. Sampled-data control Discrete PID sampled-data control ccDual” sampled-data control “Dual” and set-point control 21. vs.7d Disturbance in feed composition Convential “dual” control in loop with overrides 21.3 19.6 21.

12 Conventional control of X2 with feed composition disturbance 2 1.14 Comparison of conventional and tracking PL control 518 519 520 .21.13 Tracking sampled-data control of X2 with feed composition disturbance 21.

Proper original design is by far the best way to guarantee satisfactory operation and control. but the one chosen here is one the authors have found broadly useful in virtually all kinds of processes. and protective controls wdl be coordinated with the sizing and proper location of process holdups to achieve both automatic startup and shutdown and smooth. in many cases. For distillation there are many possible approaches. (2) product quality control.i DISTILLATION CONTROL OBJECTIVES The starring point of any design project is a definition of objectives. and (3) satisfaction of constraints. today. 3 .Strategy for Distillation-Column Control n chemical plants and petroleum refineries. Troubleshooting of columns that are already in operation is frequently necessary. Failure to obtain performance specified by the column design engineer is due. As applied specifically to distillation columns. The authors of this book have been unable to find any special merit for this scheme except for some high reflux ratio columns. The application of feedforward. many distillation columns that are working well. and distillatelfeed ratio is set manually or by a composition (temperature) controller. this philosophy suggests the following: 1.’ It has three main facets: (1)material-balance control. i 1 . in this book we will approach the design of integrated distillationcolumn control systems as a systems problem in process design. noninteracting control of column product compositions. and at least a few that function very poorly. feedback. There are also many others that are not working well. or not at all. to faulty or inadequate control system design. but practical considerations usually limit corrective measures to relatively minor items. Therefore. Material-balance control” * This term is sometimes used by others” to mean a control system in which reflux is set by reflux drum level control. there are.

4 StrMeD J%. It is important to note that the material-balance controls on any given column must be consistent with the material-balance controls on adjacent process equipment. --Column holdup and overhead and bottoms inventories should be maintained between maximum and minimum limits. such as that caused by rainstorms 3. that is. This is so because the relationship between thermodynamic work of separation and purity is nonlinear. . -Maintain the composition at the other end of the column as close as possible to a desired composition.t It is usually true that minimum operating cost is achieved when the products are controlled at minimum acceptable purities. With only two drawoffs. the composition at one end n must change a little. For example: -The column shall not flood. --Column pressure drop should be high enough to maintain effective column operation. In most cases material balance will be controlled by so-called “averaging” liquid-level or pressure controls. If feed variations are i the nonkey components.Dthdktim-Column Control -The column control system must cause the average sum of the product streams to be exactly equal to the average feed rate. and sometimes both ends. Satisfaction of constraints For safe. to prevent serious weeping or dumping. but not the nonkey components. satisfactory operation of the column.” For some columns compositions are allowed to vary at one end. Harbed4 has called this requirement that of keeping the column in “balance. certain constraints must be observed.” -The resulting adjustments in process flows must be smooth and gradual to avoid upsetting either the column or downstream process equipment fed by the column. to satisfjr certain economic constraints. Product quality control The control system for a binary distillation in most cases must: -Maintain the concentration of one component in either the overhead or bottoms at a specified value. Both material-balance and composition controls must function satisfactorily in the face of possible disturbances in: -Feed flow rate -Feed composition -Feed thermal condition -Steam supply pressure --Coohg-water supply temperature --Cooling-water header pressure -Ambient temperature. t For multicomponent columns subjected to feed composition changes. composition may vary somewhat at both ends of the column. 2. it is not possible to hold exactly constant the compositions at both ends of the column. we can conml two keys.

several other facets of column control. --Column pressure should not exceed a maximum permissible value. Heat Recovery Increasingly there is an interest in recovering as much heat as possible. and so on. as. not only to hold design costs down. heat and material balances. Startup and Shutdown Column controls should facilitate startup and shutdown and. flooding. Recently more ambitious schemes have been employed in which the reboiler for one column is used as tha condenser for another. whether a cathode-ray-tubekeyboard console or a panelboard with gages. This will facilitate. Transitions When it is desired to change product compositions. and so on. but also so that the work can be readily discussed with other design and plant personnel.I Distillutk Control O&jem’ves 5 -The temperature difference in the reboiler should not exceed the critical temperature difference. -The control system design engineer should use the simplest possible design procedures. switches. The petroleum industry has frequently used the sensible heat in a column bottom product to preheat column feed. This is particularly important when feed stock composition varies widely and it is desired to optimize column or train operation.1. should be carefully designed according to human engineering principles for easy use. Such schemes magnify control problems and sometimes limit process turndown. For example: -The operator‘s work station. for example. with a computer. --Column feed rate should not be so high as to overload reboiler or condenser heat-transfer capacity. The hardware should be designed and arranged for convenient access and quick repair or replacement. There are. any necessary minor redesign at the plant site. Failure . recorders. Testing There should be enough instrumentation so that testing may be carried out for tray efficiency. the column controls should facilitate doing so. The need for frequent or critical “tuning” should be avoided. in addition. should make it easy to achieve total reflux operation when desired. by implication. column control should be designed with human engineering in mind. -The controls should be so designed as to require minimum maintenance. -Boilup should not be so high that an increase will cause a decrease in product purity at the top of the column. at a later date. Miscellaneous In addition to the above.

while the level in the base of the column sets the bottom product flow.’ the size and location of tanks and the concept of overall material-balancecontrol used can have a great influence on plant investment and process control. is the ratio of maximum required flow rare to minimum required flow rate. Generally speaking the direction of material-balance control is determined by the demand stream. concepts.4. or distillate flow. For the last tank in a series (final product storage). In the first case. Control in the direction of flow is the most commonly used concept (although the least desirable).5.2 ARRANGEMENTS FOR MATERIAL-BALANCE CONTROL As shown elsewhere. Here level in the condensate receiver (also commonly called reflux drum or accumulator) sets the top product. for simple tanks with level controllers. tanks may be smaller and plant fned investment and is used. in response to a given change in demand flow. it is more common to have an operator make the adjustment. the choice in control strategy is between adjusting the flow into the last tank (control in direction opposite to flow) or adjusting flow into the first process step (control in direction of flow).3 and 1. in Figures 1. and methods is one of the frequent causes of unsatisfactory plant operation.6 Stratemf Didhttim-Column Control w of design and plant personnel to achieve a mutual understanding of design objectives. Once the basic concept of material-balance control has been selected for a process. one must apply the same concept to all process steps. . we can easily use simple automatic controls. “Turndown. the required change in the manipulated flows will be smaller in one case than in the other. If the design engineer uses the concept of control in the directwn opposite t o w . As shown by Figures 1. other advantages of control in the direction opposite to flow are (1) less difficulty with stability problems. and a frequently encountered arrangement is shown on Figure 1.” as used here. the demand flow is always the shipments to a customer. in the direction opposite to flow. The working capital can be lower than if umtroZ in the directwn meaning of these expressions is illustrated.2. and (2) reduced internal turndown requirements. In this instance the meaning is that.1 and 1. in other columns base level sets steam or other heat-transfer media to the reboiler. in which case the condensate receiver level sets top product flow. These are discussed in derail in Chapter 6 . 1. can lead to many interesting level-control and flow-ratio options. It is for this reason that the first step in designing column controls is to determine the material-balance control arrangement. In recycle systems we may find some material-balance controls in the direction of flow while others are in the direction opposite to @ e Material-balance control. When more than one tank is involved. In the second case.

2 A w a q m fm M a t d .1 Material balance control in direction opposite to flow FIGURE 1.2 Material balance control in direction of flow .B a l a n c e Control ~ 7 FIGURE 1.1.

C 0 0 ki @?I4 3 s m - n B -E W a- E8 32 .DistiuatiOn-column Control er -g k z = u z Ly 0 G 3 s Q c .8 VI Strazcm fb. a a 0 c E - 0 L a 3 U 2 C c . c 0 .

2 Arrangements fm Material-Balance Control 9 3 3 C L Q) C 0 E 0 0 8 s 2 m 8 2 d CI 3 z E 8 8 c m d 3 3 c .1.o U Q iz&% . e 3 - E d m -?$ ' zE= 3 e.

5 Distillation column with material balance control in direction of f o lw .10 Strategy for Dljtillatwn-Column Control FIGURE 1.

although the results will be applicable in a general way to multicomponent systems. basis for column composition control.3 Funahmentnls of Composition Control 11 1. particularly those that may be treated as quasib&ary or pseudobinary systems. Visualization of column operation. for example. Similarly. if feed composition and feed thermal condition are constant. we see that if the rectification section vapor-to-feed ratio is fixed (and it will be if the boilup-to-feed ratio is fixed) and the reflux-to-feed ratio is fixed. binary distillation. one should change the boilup and reflux according to the following equations: AV = K ~ A z F+ KfzAF where V ALR = Kf3AzF + Kf4A. Rippin and Lamb’ have shown that. reflux-to feed. Methods of handling other disturbances will be discussed later. Throughout the rest of this book. To simpl@ the analysis.3 FUNDAMENTALS OF COMPOSITION CONTROL Let us consider briefly what must be done to a column to keep terminal compositions constant on a steady-state basis when the column is subjected to sustained changes in feed flow rate or feed composition. This is somewhat restrictive. and bottom-product-to-feed ratios are held constant. it will be used as the primar). for small perturbations. If feed rate and feed composition are not constant. Luyben6 has shown that it is necessary to be quite careful in designing feedforward compensation for feed composition changes. The operating lines (defined in the next chapter) will not change as long as the distillate-to-feed. and (3) the boilup-to-feed ratio and the distillate-to-feed ratio. (2) the reflux-to-feed ratio and the bottom-product-to-feed ratio. As will be shown in Chapter 2. boilup-to-feed. the bottom-product-to-feed ratio will be fixed. particularly when the column is not making a sharp separation. since the bottom product flow is the difference between the stripping section liquid and the boilup. in terms of reflux-to-feed and boilupto-feed ratios. Practically speaking one may hold all four ratios constant by fixing anv one of the three pairs: (1) the reflux-to-feed ratio and the boilup-to-feed ratio.1. then the distillate-to-feed ratio will be fixed. was suggested by Uitti7 and has since been proposed in varying degrees bv many others.F = vapor flow from the reboiler LR = internal reflux flow at the top of the column The A’s represent departures from average operating conditions. then we want the “operating lines” on a McCabeThiele diagram to remain constant when feed flow changes. The constants KO Kf4 may be calculated approximately by the column designer. since the distillate flow is the difference between the reflux flow and the vapor flow. In considering case (1). . let us limit our attention to an ideal.

If the pressure drop is high enough. The design should avoid having sonic flow corresponding to low feed rates and nonsonic flow corresponding to high feed rates. To determine the control accuracy obtainable by this approach. When significant changes are anticipated. too. a heat exchanger and feed-enthalpy control system should be provided.4 COMPENSATION FOR VARIOUS DISTURBANCES Feed Thermal Condition Feed should enter the column with a constant enthalpy. A high pressure drop is also undesirable for energy conservation. sonic flow through the valve results. It should be noted. For columns that are not operating too close --. The best preventive is a steam-to-feed ratio control system combined with temperature and pressure compensation of steam flow. a properly designed column feed system can play a very important role in filtering out hsturbances in feed rate. that this "feedforward" approach to column control has a particular limitation: In general one cannot calculate the constants Kfl Kf4 with great accuracy. and Kf4 therefore will be constants. As will be shown later (Chapters 5 and 20). is discussed in Chapter 6. . and consequent changes in boilup and reflux. will not change tray efficiency appreciably. and reboiler steam-side pressure has no effect on flow rate. A high pressure drop across the steam valve favors smooth control but velocity-limiting trim may be required to minimize noise and plug and seat wear. that feedforward from feed composition may not be needed if the feed comes from a process step with discharge composition control. small changes in feed rate. however. Where really close control is required. accurate control of composition. such as bottom product or distillate demand flow. feed composition. 1. one must supplement feedforward control with measurement of the column terminal compositions and subsequent feedback control. Feedfonvard compensation for other process variables.12 StrateBy fm DirtillatMn-Column Control It should be noted. This is discussed in more detail in Chapters 5 and 11. to either upper dr lower h i t s of capacity. Steam Supply Pressure Changes here can cause changes in boiling rate. The usual philosophy will be to use feedforward for fast. thereby making composition control much easier. then Kfl Kf3 and may also be treated as constants. If the feed composition The terms Kf2 changes are not too large (as will usually be the case). since required controller gain changes make tuning very difficult. at least at one end of the column. and feed enthalpy. one should make the necessary calculations or tests for each individual column. approximate control and feedback for long-term.

By measuring the temperature rise of the cooling water across the condenser.5 STARTUP AND SHUTDOWN Startup and shutdown are often dismissed as relatively unimportant. To put it another way.5 Startup and Shutdown 13 Cooling-Water Supply Temperature Cmling-water temperature changes are usually seasonal. startups and shutdowns are far more common. the vapor piping from the top of the column to the condenser is both uninsulated and long. Another way. transferred. In the chemical industry. Columns are commonly started up in total reflux-no product is taken off at top or bottom. and will require no specific correction. This calculated value of qc can serve as the measured variable in an enthalpy control system. In this event it may be desirable to use an internal reflux computer. and multiplying it by the cooling-water flow rate. 1. they are large and rapid. is costlv. auxiliaries. If. suggests that faster and smoother . have limitations. Further. then it may be desirable to provide an enthalpy control system for the condenser. one has a measure of the heat .perhaps monthly. which will be discussed later. however. however. and piping are properly insulated. for some reason.1. Ambient Temperature If the column. This may be true in a petroleum refinery where shutdowns may occur at intervals of two or three years. Cooling-Water Header Pressure One of the best ways to make a flow system immune to pressure changes is to provide a high system pressure drop. ambient temperature changes mav cause fluctuations in pressure and the rate of condensation. is to use a cooling-water flow control system. however. If. however. and if the column is properly controlled. since they happen so seldom that it is not economical to spend much time and monev on improvements. or even daily. elaborate heat and material recycle schemes may require inmcate startup/shutdown procedures as part of the original design. This. 4. columns and their control systems may have to be designed specifically to accommodate a particular startup/shutdown procedure. discussed in Chapter 11. weekly. Limited experience. as is often the case. For column pressure control. which should be satisfactory for some distillation columns. ambient changes should cause little difficulty unless the condenser is of the air-cooled type. where process equipment is often plagued by severe corrosion or by plugging process materials. It does. the enthalpy control system can serve as the secondary loop in a cascade system.

process control was hardware limited. it is not surprising that some loops never work in “automatic. let US look at typical controls in existing plants. and speed of response. operation closer to hard constraints. progress in single-loop design has been essentially at a standstill. microprocessor controls.” and must implement process logic by switching in and out of ccautomatic.* In addition.1 4 Stratgy for Dtddhtkfi-Cul~mn Cuntrui startups may be achieved by recycling top and bottom products back to feed during part of the startup sequence. whether one uses analog pneumatic or electronic gear. one controller. sensitivity. and one final control element. although stable in “automatic. As recently as 1950.6 CONTROL SYSTEM DESIGN PHILOSOPHY Current Design Practices In the preface we noted that we would try. of these loops are represented in the central control room (CCR) by control stations. For newer plants with higher throughputs. The controllers usually have neither antireset windup nor automatic tracking. Further.” are so sluggish that they are ineffective in dealing with typical ‘disturbances. or a digital computer. energy-recovery systems. to present a multivariable control approach to distillation column control. Many. and sometimes compositions. . The term “single loop” means there is one measurement. for a given algorithm. There is a traditional pattern of what is called “instrumentation” in the chemical and petroleum industries based on single-loop control (sometimes called SISO-single input. however. * Digital computers and microprocessor controls. Startup/shutdown will be discussed in more detail in Chapter 9. usually offer a wider range of controller parameter adjustments and facilitate the design of control systems more sophisticated than most of those discussed in this book. usually a valve. pressures. Before discussing this. Consequently these characteristics less and less frequently pose limitations to the single-loop designer. Since then primary measuring elements. flows. the quality of control achieved by single-loop systems is not greatly affected by type of hardware. it makes little difference. however.” Since the original design procedures are usually qualitative and intuitive. controllers. For some years now. This statement is true for both analog and some digital hardware. As a consequence the operators must perform startup operations with the control stations switched to “manual. Each process operation has a number of independent or single loops for feedback control of temperatures. control stations. with heavy emphasis on conformity to past practices. computing devices. and elaborate materialrecycle systems. in this book. and there is little or no logic circuitry to tie the many loops together. 1.” Others. smaller and fewer holdups. single output). the traditional “instrumentation” approach to control is often seriously inadequate. but not all. liquid levels. and control valves have improved greatly in reliability.

in a given situation. Multivariable Control To avoid the limitations of single-loop design and to provide a more flexible and sophisticated process operating logic than can be implemented by human operators. however. they yield results that are far inferior to those obtainable with well-damped feedback controls with simple feedforward and override control. this is a variable configuration. the best of several preprogrammed strategies (algorithms) for manipulating one or more control valves (or other final control elements). precision. with their extensive lateral and diagonal crossovers. the steam valve for a distillation column reboiler.” Multivariable control may involve both variable configuration and variable structure controls. we use an approach we call multivariable control. base composition is normally controlled by steam flow that can be taken over or overriden by high column AP. Hardware permits us to automate this kind of control with a speed. If base level control is normally achieved bv a PI controller that can be overridden by high or low base level proportion&-only controls.6 Control System Des@ Philosop@ 15 optimized tuning procedures for unaided feedback controllers have limited practical value for continuous processes.1. If. we call that “variable structure. For our purposes we define a multivariable control svstem as one that has the built-in intelligence to look simultaneously at two or more process variables and to choose. we make extensive use of “variable configuration” controls that are usually implemented by overrides. To provide automatic control of this sort. are functionally more like the modern “matrix” concept of management. and reliabilinr that are completely beyond the capabilities of human operators. may respond to controllers for: Steam flow rate Column AP Column pressure Base temperature Column feed rate Column base level Column bottom-product rate The seven variables listed may also exert control on five or six other valves.* For plants being designed today (late 1983). It is common to think of process control fimctions stacked one above another in a pyramid or hierarchic arrangement as with traditional business or military organization structures. but most of them are expressed in mathematical terms rather than in terms of process hctions. depending on circumstances. Multivariable control structures. From .’ Many definitions of this term may be found in the literature. For example. it is increasingly common to use microprocessor controls instead of analog (see discussion under “Hardware Conventions and Considerations”). for instance. particularly when computers are involved.

Further. for example. Our design procedures are accurate enough that only a modest amount of empirical controller tuning is required at startup. A typical chemical plant or refinery has hundreds of single loops with many interactions among them. 3. Stability. It is usually far more important to design for a dynamic balance among these loops with a minimum of interaction than to strive for maximum speed of response. The traditional objectives of feedback control system design are: 1. For microprocessor computer controls. distillation column reflux flow and boilup should not be changed too rapidly since these might cause transient flooding or weeping in part of the column. Design material-balance control loops to be at least a factor of 10 slower than related composition control loops. Our preferred philosophy of controller tuning is discussed in a book’ and a paper. By now it is probably apparent that we are striving for control system designs whose performance and design parameters are specified in advance of plant startup.” There are five simple methods by which to make a system noninteracting: 1. To compensate for or to attenuate disturbances as much and as quickly as possible. Similarly. 2.” In process control the objectives are often quite different. To get the fastest possible response to set point changes. we calculate scaling parameters for computation blocks (either in software or hardware).16 Strategy fm Distillation-Column Control a control engineering standpoint. and Interactions Most existing literature on automatic control is concerned with the stability and speed of response of single loops. Avoid designs that are intrinsically interacting. One of the two controllers must be detuned. which means that controls must be added and must be kept in “automatic” for stable operation. . troublefree operation. it is usually undesirable to make rapid changes in manipulated variables since these may upset the process. For example. The objectives of averaging level control. In practice we furnish calibration data for controller parameters and computational devices for the majority of control loops prior to startup. clearly are different from those just mentioned. 2. make the secondary or slave loop at least a factor of 10 faster than the master loop. in cascade systems. and more reliable. For example. These must be accomplished with a reasonable degree of closedloop stability. the shorter lines of communication and decentralized control functions permit more rapid and stable control. such as pressure control and flow control at the same point in a pipeline. Select process designs that eliminate or minimize interaction. Speed of Response. We calculate these from simulations or simple linear models. * Some processes are open-loop unstable.

There is also a “manualautomatic” switch. This may or may not be subject to restrictions imposed bv feedforward compensation. and the valve loading signal (controller output. with process design engineers. like an analog station.” These may be analog pneumatic. they do not permit ready implementation of feedforward. really). flow-rate interactions may be reduced to an arbitrary level by providing a very high pressure drop across each feed valve in comparison with the reactor pressure drop. it is normal practice (as mentioned earlier) to furnish “control stations. the feedback controller is disconnected and there is a knob that enables the operator remotely to set the valve position. interact. the station may be physically distinct. the secondary controller set point comes from the output of the primary controller. eliminate interactions. temperature. valve position may be set manually or the controller set point may be set by the operator (c‘local-auto”). and so forth. or microprocessor based. and s for& is that this discussion may be very helpfid. level.^^. analog electronic. but also another fUnction-“remote-locaI. and so on. liquid level control cascaded to flow control. which some vendors label “hand-automatic. such as top and bottom temperature controls on a distillation column.Although cascade functions are sometimes combined into one station (or CRT “faceplate”) for space and money-saving reasons. or may be represented on a CRT display as a “faceplate. override circuits. etc. One compensates for the action of the top temperature controller on the bottom loop while the other compensates for the action of the bottom temperature control loop on the top one. In addition. Although not specifically intended for this purpose. If two control loops.1. This is discussed in Chapter 20. for example.” In the “remote” position. Use interaction compensators (decouplers). there are some very sophisticated mathematical methods for dealing with interaction^. overrides.” In the “manual” mode. overrides.6 Control System Des@ Philosqphy 17 if a tubular reactor is fed at several points along its length from a common header. we can eliminate the interactions by installing two compensators. depending on the design philosophy for a particular project.^^'^^ Some are intended for noninteracting design while others seek a design that provides an optimum amount of interaction. 4. by permitting only one controller at a time to control a given valve. If cascade control is involved.” Each provides an indication of the process variable (flows. the desired value (set point).). as. Use override circuits. column designers. * Much of this section may seem superfluous to instrument and plant personnel. Our experience. In the last case. Hardware Conventions and Considerations* For control loops represented in the CCR (central control room). we recommend dual stations. 5. . o however. Most single-station designs with which we are familiar are very inflexible and complicated. the secondary station not only has manual-automatic switching. In the c‘loca”’ position.

As far as CCR hardware is concerned. . Better measurements. are limited to 0. Sometimes a consolidated console is also provided for production supervision. For maximum advantage a supervisory computer should be programmed to have the control algorithms discussed in Chapter 12. however. and contain more control algorithms. two-stage positioners are recommended. perhaps at another location. many are featuring satisfactory antireset windup and override capabilities. and adaptive tuning. They are technically more versatile and are less expensive (some versions) than analog. coercive stem forces when the valves are in flowing streams. With regard to these. Most analog readout devices. especially of compositions. For both there are substantial.18 Strategy fm Dirtillation-Column Control As of this writing (late 1983). Other advantages include freedom from drift. can be reconfigured or restructured without wiring changes.000.5-2 percent. or microprocessor). * Most control valves today are either single-seated global types or rotary types. it is possible to find worthwhile applications for a supervisory digital computer with a good data historian. we have a decided preference for microprocessor controls. It should be noted. their performance can generdy be approximated by analog control algorithms. Microprocessor controls usually have a sampling time of a fraction of a second. Positioners compensate for this and maintain the valve’s inherent flow characteristic expressed as a function of controller output signal. Computer consoles were originally provided in the CCR for the convenience of the operators. Engineers’ consoles.* Piston actuators with double-acting. It is our opinion that using such a computer to imitate unenhanced two. Separate consoles for maintenance personnel (the computer can be a powerful maintenance tool) are highly desirable. Some worthwhile applications for computers will be discussed later.and three-mode analog controls is poor practice. that some users prefer not to use positioners. Digital readouts for important variables are worthwhile because they permit seeing their magnitudes with sensitivity approaching that of the original analog measurements. regardless of the type of basic controls selected (pneumatic. Others have gain scheduling. electronic. In adchion. Although slightly slower than analog controls. Valve positioners should always be used (it is assumed in this book that they will be). and the fact that they can be calibrated more precisely. 2. As of late 1983. facilitate technical studies. more progress has been made in the design of valve bodies and trim than in the design of actuators. however. the two biggest hardware needs appear to be: 1. These are position rather than velocity algorithms. Better control valves. For most projects today. dead-time simulation. have a larger range of tuning parameters. where reset time and proportional gain are functions of some process variable or the controller error signal. Some of the last named achieve self-tuning via stochastic techniques or by pattern recognition. Most typical analog measurements have a sensitivity ranging from one part per 1000 to one part per 10. they provide more advanced logic capability.

It may also save startup time. 1. With holdups determined calculate column-composition transfer functions. d. b. Use simulation for some columns. Prepare material-balance and composition-control signal flow diagrams. Calculate al other overrides. This is due to the trend toward increasinglv tight column design. l 8. The number of trays is held down because smaller allowan& are made for uncertainty in tray efficiency. Determine holdup volumes required for smooth material-balance control and for liquid-level override controls at each end of the column. 5. with the . prepare a simplified flow sheet that defines the control concepts: a. Calculate feedforward and interaction compensators. and after careful review of the overall process flow sheet. Check feed-tank material balance and mixing time constants for adequacy.8 Column Des@ Philosophy and Control System Des@ 19 1. Quantitative information for the other steps is presented in Part 111. preferably l proceeding from final product inventory to raw material inventory. e. 3. Simulation of the column and its control system will be usefd in confirming control concepts and controller tuning parameters. 9 . Select composition control schemes. 2. 6 . Al individual equipment-piece material-balance controls must be consistent with this scheme. Add miscellaneous temperature and pressure controls. Column diameter and tray spacing are now kept to smaller values. After careful discussion with the process engineer and column designer. Calculate feedback-controller gain and reset settings and control-loop natural frequencies. 7. 1. Hardware vendors may now be selected and the measurement and control equipment may be ordered.7 PROCEDURE FOR OVERALL CONTROL SYSTEM DESIGN We now have enough information to suggest a sequence of design steps to follow that will lead to a quantitative definition of column and hardware performance. simple columns benefit from a modest application of feedforward compensation and overrides. Select the overall material-balance control scheme first.1. Select measurement spans and calculate control valve sizes. Add feedforward and interaction compensators as required. Add protective controls and automatic startup/shutdown circuits. Part I1 of this book deals with the qualitative and heuristic aspects of steps 1 and 3. c. 4. particularly those in critical service or with a new untested control system or process configuration.8 COLUMN DESIGN PHILOSOPHY AND CONTROL SYSTEM DESIGN Experience on many projects shows that even small.

8000 hours per year). . 8. If water cooled. and f l d g or dumping. It is. and columns with heat recovery schemes. 3. Avoid sidestream drawoff designs. This incremental investment not only provides better normal control. Provide larger reboilers (lower heat flux). Provide larger condensers.* 2. our opinion that these controls should be used to some extent on almost every column. Experience indicates that typical incremental instrument investment over that required for unenhanced feedback controls is 5-10 percent (large projects tend toward the lower figure). To save this amount of steam would probably be only a modest accomplishment for most columns. This not only wastes energy. 6.00/1000 pounds. we suggest the following: 1. 9. existing columns. it also reduces column capacity. Provide 100 percent reserve capacity in heat-recovery schemes or avoid them altogether. or flow control steam. Provide surge tanks between columns with at least 30-60 minutes of holdup each. particularly sidestream drawoff columns. Frequently encountered problems include unstable or ineffective controls. 7. it is fairly common practice to use excessive boilup and reflux to make sure of meeting or exceeding product specifications. the opportunities for their improvement. hs it is pertinent to consider briefly the controls of typical. off-specification product or products. 1. use tempered water. Provide five extra trays or increase the number of trays by 10 percent. but it also helps to avoid inadvertent shutdowns. and how to troubleshoot them when necessary. What column design philosophy should be followed?Having had considerable adverse experience with columns with primitive controls. that the customer insists on minimum application of feedback controls with no feedfonvard compensation or overrides. one may note that 100 Ibm/hr of steam is worth $3200 per year (basis $4. therefore. 5. Control column Al’ by boilup. Design for normal operation at 60 percent of the rate for flooding.9 EXISTING COLUMNS-TYPICAL PRACTICES AND TROUBLESHOOTING Although ti book deals primarily with &stillation control in design projects. however. Provide increased tray spacing.20 Strategy fm Dljtillatwn-Column Control result that columns typically operate closer to flooding. Incorporated. The combined effect of these design policies is to make columns much touchier and harder to control. whichever is larger. To provide a perspective on energy savings. In addition. If composition control of each product stream is desired (and this is usually * In some companies columns are designed by the probabilistic methods recommended by Fractionation Research. Suppose. 4.

ratio controls should be replaced by impulse feedforward compensation . Provide a linear flow measurement for steam flow control. 4 If condenser controls are a problem. pressure. and that hardware. For further improvement provide steam-to-feed ratio control. 8. With rare exceptions flow-control set points must be changed to accomplish either composition control or material balance control. Make sure that column material-balance controls are properly designed and tuned. internal reflux-to-feed or distillate-to-feed ratio control. and sometimes at neither end. 3. . make sure an adequate scheme is provided for maintaining an inert balance. 7. and. smoothed. is in good working condition. 1. Provide averaging level control of column feed. and a minimum steam flow ltmiter to protect against dumping. in which case flow measurement should be linear. At this point with flows established. or both. Column feed rate should also be held between maximum and minimum limits. in some cases. Any such unaided flow control should be regarded with suspicion. 6. If composition is measured and is cascade controlled via reflux or boilup. If PI level controllers are’used. Another shortcoming frequently observed is the use of fixed flow controls for steam. Most commonly temperature in the upper or lower section of the column (or both) is used in lieu of true composition measurements. we will frequently illustrate control schemes with pneumatic components. follow the tuning procedures of Chapter 16. the most obvious deficiency of most exisring columns is the lack of appropriate composition measurements. This is done for convenience.I 0 CONVENTIONS FOLLOWED IN THIS BOOK Throughout the remainder of this book. high A P override on steam to protect against flooding. Frequently composition control is attempted at only one end of the column. especially level and flow transmitters and control valves. If temperature by itself is not an adequate measure of composition. reflux. For pressurized or vacuum columns. In view of the preceding comments about problem areas and likely opportunities for improvement of composition control and reduction in energy consumption. 2. it will probably be possible to see some improvement in composition control. Pneumatics. Chapter 3. review the control schemes in . flow measurements to deduce compositions. auto overrides or nonlinear controllers should be used. Also provide temperature and pressure compensation. and in some cases limited.IO Cunventwns Followed in thlj Book 21 the case). It is usually desirable to cascade level control to flow control. or product drawoff. (see Chapter 12) if feed flow turndown is greater than 2: 1. . consider one of the schemes in Chapter 10 for using temperature. 5. the following guidelines are suggested: 1. at least for part of the time.1 .

1 . some of these work fairly well.” which stands for proportional-integral. in programs for computers or programmable calculators. multipliers.12It relies heavily on conventional. It also contains an extensive bibliography. feet. Units used in this book are those commonly employed in chemical engineering: pounds. G.22 Strategy fm Dirtillation-ColumnControl with few exceptions. In pneumatics it is c o m o n to refer to most signal-conditioning devices other than controllers as “relays. and so on.g. It is also becoming common today to speak of iontroller gain rather than proportional band (PB = 100/Kc). and has been adopted by one manufacturer of analog electronic instruments and by several vendors of microprocessor-based distributed controls.. and explores painstalungly a large number of possible control systems. For projects that make i] partial use of metric or SI units. The second book is by Rademaker. etc. while some are quite inflexible.” usually in minutes. and so on. The most common type of controller used in the chemical and petroleum industries was once called “proportional plus automatic reset.” later shortened to “proportional reset.” The treatment of energy conservation alone is worth the price of the book. feedfonvard. can achieve antireset windup by the same external reset feedback method. ALL pneumatic controllers. In drawings where we are trying to illustrate concepts of structure (overrides. Ibm/min. To facilitate the calculation of control engineers’ “time constants. mols. we have found no advantage in writing equations with SI units.15 psig.I I LITERATURE For anyone seriously interested in distillation control.” Today it is more common to use “PI. Ibm/sec. For electronic analog and digital controls. A table of nomenclature and symbols will be found at the end of this book.” we have mostly used time units of minutes or seconds [e. single-loop control theory. Other vendors of electronic analog and digital controls feature a wide variety of techniques. The first is an easy-to-read. rather than its older reciprocal. it is more common to use terms such as signal scalers and mult@lien. we mostly write the equations in the older units and add subroutines for going back and forth to metric or SI units.We will also use “reset time. . degrees Celsius. Instead. subtractors. two books are highly recommended. we found it convenient to convert them to the above units. features a standard signal span of 3. (pcu/sec>/”C f’. we use a more detailed symbolism that we have found useful in our design work. and Maarleveld.” Included are adders. we use very simplified symbols. Generally speaking. “repeats per minute.). This appears to be the most universally usell method (we use it extensively). as far as we know.” For drawings in which we are trying to present a perspective of control concepts (configurations). nontheoretic (as far as control is concerned) work by F. Shinskey. A brief discussion is presented in Chapter 12. Rijnsdorp.

D. ProcessMod.. M c A v o ~ ~ addressed the specific subject of has interaction analysis. 1970. Modelling o Separation Processes. Techniques o Process f o Continuous Distillation Units. McGraw6. “A Theoretical Study of the DyN. Jan.J. Prentice-Hd. 10. Recent books bv Rav”3 and by S t e p h a n o p o l ~ u discuss applications of “modern control s~~ theor). T b e o q . Chem. D. McGraw-Hdl. Van Winkle. N. 1960. “MultivariableControl 18. New York. sevier. Hill.. 7. 1968. C. we recommend two written by the one by Harriott. E. Introduction to Control 1973. J. J. New York. G.” and one bv Murri11. Enginem. D. J. S. R$27(4): 10613.. New York. McGraw-Hill. namics and Control of Binan Distillation Columns. f D. New York. R~ppin. 5. . Buckle?. K. Douglas. 61(8): 74-78 (1965).J. DG-tilhim. we have made much use of those Treybal. a Chemical Reactor. Holland. S.. Harbert. I.Refevences 23 For basic reference books on distillation. 9. 15. Computer Methodr fm Soltin8 Dythe Universitv of Delaware. P. 49-9-2. T. Separation Processes. 16.l6 and Hengstebeck” by Van Winkle13 and King. 4 Harbert. Pet. Ref 29(10): 11714. W. 1967. Pet. F. Ref 35(11): 151. Process Control.21 and Gould... Uitti. Harriott. wood Cliffs. J. New York. Holland. New York. Fundamentals and 159 (1956). Hill.. 1983.. 3. Pet.. D. DynamicJ and Control 1. 1964.. D.. 8. E. Mms-Tran$er Operatwns. End.. we suggest texts by Koppel. Rijnsdorp. King. W. Apr.. ISA Paper no.. 1977. Luyben. m u Separmim P r o b h . Distillation. 12. 1961. and 21. E. 1964. D. R.20 Douglas..” Bulletin from 15A. New York. Englewood Cliffs. McGraw-Hdl.ellin8. REFERENCES A. P. Koppel.. Harbert.” presented at Purdue University. 2. Jan. Lamb. M.. 1973. New York. “A Modem Perspective on Controller Tuning. P..24It is clearly written and easv to read. 1975. Reinhold. Symposium. 1975. P. and D. C. Liapis. New York. P q . B. Simulation. Trevbal. Rademaker.28 For more advanced treatments. DljtiLLatwn Control. Shinskey. McGraw109 (1948). L. Hengstebeck.22 The last contains some perceptive comments about the difficulties of applving advanced control theory developed by electrical and mechanical engineers to chemical processes. R. S. at Texas A&M University. W. 1971. Prentice-Hall. L. For basic books on control. Buckley. 20. Maarleveld. 17. 1975. L.. W. Process Dynamics and . Buckley. Wiley.. . . Luyben. New York. 122 (1950).14Others by also have been useful.” in Proceedings of Texas A&M Instrumentation 17. W. Nonchemical engineers with no background in distillation may find an introductonr text bi7 Nisenfeld and Seemann useful. P. C.. Buckley.. and A.” presented 19.. W. “Override Controls on 1968.” ’in chemical processes. Engle11. Elf Control. O. M. and Control f i Chemical in the Process Industries. S. McGraw-Hill.. McGrawHill.

INTECH.C. Chemical Process Control. H. Prentice-Hall. Interactwn Analysir. P. 1972. Murrill. Control. W. N. Triangle Park. 1981. Englewood Ray. Fundumntals $Process Control Theory. T.. Seemann. Buckley. ISA.. Advanced Process Control.En1974. and R.24 Stratgy fm Dirtillation-Column Control in Recycle Systems. C. 28.. May Control ( 2vols. McGraw-Hd. 25. Instrument Society of America. Prentice-Hall. pp 29-34. L.. 24. . 27. Mass.J. Nisenfeld. Chemical Process 1969. 1983.. N. McAvoy... G. 23. 1981. W.J. Addson-Wesley. 26.. Cliffs. Research of America. New York. Instrument Society Dictdlation Columns. N. I. 1983...). glewood Cliffs.. 1981. A. E. Reading. S. P. Stephanopolous. Gould. Materral Balance Control 22.

Most readers with chemical engineering backgrounds will be familiar with this material and can skip some sections of this chapter. For example. The control engineer must have a basic understanding of any process before an effective control system can be developed. components are removed from the top of the column. This separation is based on differences in “volatilities” (tendencies to vaporize) among various chemical components. or heavier. Feed rate is F mols per minute. with flow 25 . a mixture of methanol and water can be separated by distillation because methanol is more volatile or boils at a lower temperature than water. components are removed from the lower part of the column.” and “bottoms” or “bottom product. the more volatile.1 summarizes the nomenclature conventions that will be used throughout this book. The column trays are numbered fi-om the base upward. or lighter.’’ respectively. and the less volatile. Feed composition is zF mol fraction of the more volatile component. Fundamental Objectives Distillation columns are very widely used in the chemical and petroleum industries to separate chemical components into more or less pure product streams. At this point we will consider only a simple single-feed.1 INTRODUCTION he purpose of this chapter is to review briefly the essential features of the distillation process. Products removed fi-om the top and bottom of the column are called “distillate” or “top product. The total number of trays in the column is NT. with feed introduced on the NE tray. In a distillation column. two-product column separating a binary (two-component) mixture.2 t Fundamentals of Distillation 2. Nomenclature Figure 2.

26 Fundamentals o Distillation f FlGURE 2.1 Nomenclature and conventions for typical distillation column .

This liquid reflux and the vapor boilup in the base of the column are necessanr to achieve the separation or “fractionation” of chemical components. and x . Heat is transferred out of the condenser at a rate g. Heat is transferred into the process in the “reboiler” (n7pically a tube-andshell heat exchanger) to vaporize some of the liquid from the base of the column.2 Control variables for distillation column . mol fraction light component.. and reflux are inputs into this box (see Figure 2. some liquid. etc.. pcu/hr.. Liquid from the condenser drops into the reflux drum.) . Feed. Outputs from the box are the two product streams D and B with compositions x. The usual situation . In addition. is fed back to the top tray of the column. with a distillation column is that the feed rate and feed composition must be FIGURE 2. mols/min). heat. and reflux drum as a “black box” process. The heat-transfer rate is qR energy units/time (e. joules/minute.1 Introduction 27 rates D and B mols/min and compositions xD and X. pcu/hr.2. required to make the separation is approximately the heat added to the reboiler. The energ\.2). Btu/hr. condenser. called ccreflux” ( L o . Overall View f o a Control Perspective rm One can stand back and look at a distillation column with its associated reboiler.g. The vapor coming from the top of the column is liquified in another tubeand-shell heat exchanger called a condenser. Distillate product is removed from this drum.

heat input to the reboiler is the only variable that can be used. The vast majority of industrial distillation columns are equipped with trays or plates (sometimes called ccdecks”in the petroleum industry) located every 1-3 feet up the column. alternatively reflux can be adjusted. and the intrinsic difficulty of separating the components. and B. They are not independent. D is controlled. if this is more convenient. however.” there are basically two fundamental manipulated variables that affect compositions. D. preventing the liquid from falling through. We will discuss the pros and cons of various choices of control schemes in more detail in later chapters. Sieve trays are simple flat plates with a large number of small holes. particularly when product purities are high. Fundamental Manipulated Variables Two of the four manipulated variables listed above must be used to maintain liquid inventories in the reflux drum and in the column base. then B is dependent. for example. the energy input to the reboiler. Fractionation also affects product composition. Therefore. Distillate and bottom product rates can also be manipulated. q R . The D/F and B/F ratios can be manipulated either lrectly (as proposed by Shinskey in his “material balance control” scheme) or indirectly. A slight change in feed split can change product compositions very drastically. It varies with the number of trays in the column. 2 2 TRAY HYDRAULICS .28 Fundamentah of Dictillatimt considered as disturbances. Vapor-liquid contacting is achieved by a variety of devices. since reflux and heat input are tied together through overall energy and mass balances. No matter what “manipulated variables” are chosen to control what “controlled variables. In either case feed split has a very strong effect on product composition. If. The most widely used trays in recent years have been sieve trays and valve trays because of their simplicity and low cost. These are “feed split” and “fractionation.” Feed split means the fraction of the feed removed as either distillate or bottom product. we are left with just two manipulated variables that can be used to control compositions in the column. The steady-state effectiveness of both the Qrect and indirect schemes is identical. These trays promote mass transfer of light components into the vapor flowing up the column and of heavy components into the liquid flowing down the column. These are called manipulated variables. Vapor flows up through the holes. Fractionation means the degree of separation. so there are four variables that can be adjusted: La. Heat input q R and external reflux Lacan be adjusted to achieve the desired control objectives. For a fixed column operating at a fixed pressure with given chemical components. Heat input can be used directly. Liquid .

This fairly complex process of flow of vapor up the column and of liquid across each tray and down the column is called tray “hydraulics.3 Schematic of typical sieve tray . FlGURE 2. Vapor flows from one tray up through the tray above it because the pressure is lower on the upper tray. 29 flows across each tray. providing more or less effective hole areas as vapor flow rate changes. Valve trays are built with a cap that fits over the hole in the tray and that can move up and down.3. Liquid must flow against this positive pressure gradient. and drops into a “downcomer. or if vapor-liquid contacting is poor.” It is important in control system design because it imposes very important constraints on the range of permissible liquid and vapor flow rates.2 2 Tray Hydrauliu . See Figure 2.” which provides liquid for the tray below through an opening at the base of the downcomer. passes over a weir. Thus there is an increme in pressure from the top of the column to its base. the separating ability of the column drops drastically. If liquid cannot flow dowh the column. A liquid level is built up in the downcomer to a height sufficient to overcome the difference in static pressure between the tray onto which the liquid is flowing and the trav from which it is coming. It is able to do so because the liquid phase is denser than the vapor phase.

see pages 424-430 of reference 8.). On the other end of the scale. These hydraulic constraints can be handled in control system design by using maximum and minimum flow limiters on heat input and reflux.4).. Therefore. Pressure is measured at various temperatures. speciesj . 2. Vapor Pressure The liquid phase of any pure chemical component. Tray “flooding”* occurs when the liquid height in the downcomer equals or exceeds the height between trays (tray spacing).3 VAPOR-LIQUID EQUILIBRIUM FUNDAMENTALS Distillation columns can be used to separate chemical components when there are differences in the concentrations of these components in the liquid and vapor phases. if vapor rates are reduced too much. The temperature at which the pure component exerts a pressure of one atmosphere is called its “normal boiling point. The same thing occurs if liquid rates are so low (as they often are in vacuum columns) that it becomes difficult to hold enough liquid on the tray to get good vapor-liquid contacting. outlet weir height. This is usually due to excessive boilup (vapor rate) but sometimes may be caused by excessive reflux. the vapor pressure drop through the openings in the tray will be too small to keep the liquid from weeping or dumping down through the ho1es. exerts a certain pressure at a given temperature. * For a further discussion of flooding. Vapor-liquid equilibrium (VLE) data and analysis are vital components of distillation design and operation. vapor-liquid contacting is poor and fractionation suffers.) and the average liquid height on the tray (which varies with liquid flow rate.30 Fundamentals o Dktillation f This pressure difference depends on the vapor pressure drop through the tray (which varies with vapor velocity. t This occurs at about 60 percent of design vapor rates for sieve trays and about 25 percent of design vapor rate for valve trays. etc. etc. These concentration differences are analyzed and quantified using basic thermodynamic principles covering phase equilibrium. A measurement of column pressure drop can also be used to prevent flooding. .t If this occurs. Vapor-pressure data are obtained by laboratory experiments where both liquid and vapor phases of a pure component are held in a container (see Figure 2. vapor density. number and size of holes. The control system must keep the column from flooding.” Light components have low normal boiling points and heavy components have high normal boiling points. It is a physical property of each component. there are maximum vapor and liquid rates. This pressure is called the pure component ‘‘vapor pressure” P.

6. . Note that the constantsA. atmospheres. H a ) absolute temperature (degrees Kelvin or Rankine) constants over reasonable range of temperatures Therefore. + B.5). Vapor-pressure data often can be described by the Antoine equation*: In P. psia. = = = vapor pressure ofjth component in any pressure units (commonly mm Hg.4 Vapor pressure and temperature measurement * A three-constant version of the Antoine equation is used in Chapter 10.3 Vapor-Liquid Equilibrrum Fundamentals 31 When the data are plotted on linear coordinates (see Figure 2.2. and BI must be determined for each pure component. vapor-pressure data are usually plotted using coordinates of log pressure versus reciprocal of absolute temperature as illustrated in Figure 2.1) P I T A. where = A. a nonlinear dependence of vapor pressure on temperature is obtained./T (2. ESSURE GAUGE VAPOR LIQUID TEMPERATURE MEASUREMENT FIGURE 2. and B.

Then temperature and pressure are measured. When working with a distillation column.32 Funhmentals o Dirtillation f They can be easily calculated by knowing two vapor-pressure points (P.T2 p1 (2-2) A=lnP2-- B T 2 (2-3) Experimental VLE Data for Binary Systems The VLE data for binary systems are obtained experimentally by mixing two components and allowing the vapor-liquid system to equilibrate. Mixture composition is then changed and the procedure repeated. Vapor composition is expressed as mol fraction light component. T1and T2 are usually selected to be near the temperatures at the top and at the bottom of the column. T1 at and P2 at T 2 ) . pressure for pure component . Samples of the vapor phase and liquid phase are taken and analyzed. Temperature vs. B = (Td(T2)In pz Tl . VAPOR PRESSURE P (mm Hg) 760 "NORMAL BOILING POINT" FIGURE 2 S . using the symboly. Liquid compositions are usually expressed as mol fiaaion of light component and the symbol x is used.

FIGURE 2. and at the pressure under which the data were obtained.2. T-xy Diagrams If the data are taken with pressure held constant (isobaric).” A. as illustrated in Figure 2. P-xy Diagrams Data taken at constant temperature (isothermal) are plotted as two curves: pressure versus x and pressure versus y. temperature s . Figure 2. To determine the composition of liquid and vapor phases in equdibrium with each other at a given temperature.8. one merely draws a horizontal line at the given temperature and reads off x and y values. Note that the ends of the curves must intersect at x = y = 0 and at x = y = 1.7 itlusrrates a typical T-xydiagram. either light or heavy. since these points correspond to pure components.3 Vapor-Liquid Equilibrium Fundamentals 33 These results are conveniently presented in graphic form using several types of “phase diagrams. B. it is convenient to plot two curves on the same paper: temperature versus x and temperature versus y.6 Typical method of plotting vapor pressure v .

If we impose a very high pressure on the mixture.9) is called the “equilibrium line. and subcooled liquid.10. superheated vapor.7 Temperature vs. composition of binary mixture at constant pressure . dew point. and we hold it at the same temperature T for which the diagram was drawn. This x versus y curve (see Figure 2.34 Fundamentaki o DirtiLjation f C. Consider the P-xy diagrams sketched in Figure 2. Suppose we have a mixture with composition z. we will be above the x versus P curve (saturated-liquid line) and no vapor will be present. Both T-xyand P-xy are u s e l l in illustrating the concepts of bubble point. x-y Diagrams Either isothermal or isobaric data can be represented by simply plotting liquid composition x versus vapor composition y. There will be only xandy (MOLE FRACTION LIGHT COMPONENT) FIGURE 2.” This type of diagram is the most widely used on distillation.

composition of binary mixture at constant temperature .3 Vapm-Liquid Equilihum Fundamentals 35 TEMPERATURE CONSTANT AT T x and y (MOLE'FRACTION LIGHT COMPONENT) FIGURE 2.2.8 Pressure vs.

The y versus P line is called the saturated-vapor line. At this high pressure.” FIGURE 2. As pressure is reduced further.36 Funhmntals of Dijtillation a liquid phase present with composition z.vapor will begin to appear. This is called the dew-point pressure of this mixture of composition z and at temperature T.9 x v .10. This therefore is c d e d the bubble-point pressure of this mixture of composition z and at temperature 7’. Finally. more and more vapor is formed. the liquid is called ccsubcooled. we will move down the vertical line drawn through composition 2. When pressure reaches the point labeled PBp. shown in Figure 2.’’ If we now begin to drop the pressure. y for binary mixture s . At pressures below Pop only a single phase exists. at a pressure Popd the liquid has vaporized. The composition of this first bubble can be read off the y versus P curve by moving across horizontally at PBp. “superheated vapor.

and flash calculations. The same is true for dew-point temperature and dew-point pressure. These calculations are called bubble-point. dew-point.10 Bubble point and dew point at constant temperature . depending on which variable is fixed (isothermal or isobaric situations). TEMPERATURE CONSTANT FIGURE 2. The mixture is superheated vapor at temperatures above the dew point TDp subcooled liquid at temperatures below the bubble point TBp. and Note that we can talk about either bubble-point temperature or bubblepoint pressure. we more commonly need to be able to calculate quantitatively various liquid and/or vapor compositions and temperatures or pressures given certain conditions in the column.11).2.3 Vapor-Liquid Equilibrzum Fundamentals 37 The same concepts can be visualized using constant-pressure T-xy diagrams (Figure 2. VLE Calculations Instead of using graphical techniques.

11 Bubble point and dew point at constant pressure .38 Fundamentals o Distillation f A. Thermodynamic Basis The second law of thermodynamics tells us that the chemical potential of each component must be equal in both liquid and vapor phases at phase equilibrium. y . pT = xj pj Y j (2-4) where x. A somewhat simplified equation representing this condtion is: 3. Pr = mol fraction ofjth component in liquid = mol fraction of jth component in vapor = total system pressure PRESSURE CONSTANT x and y FIGURE 2.

2.3 Vapm-Liquid Equilibrtum Fundamentals = = 39 P . The K value (K. The problem is to calculate the unknown temperature or pressure and the‘ composition of the vapor phase (3. distribution coefficients or “K values” are customarily used.).3. = 1). we know the composition of the liquid (XI’S are all given). When convergence has been achieved. The calculation procedure is iterative: 1. Calculate: x< P. Bubble-Point Temperuture Cakulutwn. We will assume ideal VLE behavior for the rest of this section for purposes of simplicity. If not. The system is “ideal” and obeys Raoult‘s law (7. Guess a temperature T . . decrease T. In the petroleum industry. calculate vapor compositions. If P T is greater than PT. reguess T and go back to step 2. is sufficiently close to PT.) of thejth component is defined as the ratio of vapor composition ( y t ) to liquid composition (x.P. In addition we must be given either the pressure or the temperature of the svstem. 2. Calculate vapor pressures of all components at T . ( T ) j=1 where N. 5. B. 3.+ ( T ) = CX. If the system is “ideal” ( y j = I). Bubble-Point Calculations In all bubble-point calculations. the K value is simply the vapor pressure divided by the total system pressure. y. = number of components 4 Check to see if .). Nonideality will be discussed in more detail in Section 2. C‘ 6. If the components are chemically quite s d a r . vapor pressure o f j t h component at the temperature of the system activity coefficient of thejth component in the liquid phase at the conditions of temperature and composition of the liquid This activity coefficient is a ‘‘fudge factor” that is used to account for nonideality. If PP is less than P. This is by far the most common type of calculation encountered in distillation work because column pressure is usually known. there is little attraction or repulsion of neighboring molecules of different types. increase C T.

25 120°C XIP. P. at XI B 0.9).35 X 0. C.25 = P. we know the vapor composition (y.35 Guess T 125°C 0-Xylene (X) x3 = 0.40 Toluene (T) x2 Guess T = 120°C = 0.’s are all given) and either temperature m pressure. 0-Xylene. Bubble-Point Pressure Calculation. This illustrates precisely why a distillation column can be used to separate chemical components. Solve iteratively €or the temperature that satisfies equation (2.258 0. Dew-Point Temperature (PGiven). The vapor rising in the column gets richer and richer in light components at each stage. the lightest component. 3 1 2300 1000 380 pcalc -j 9 5 1365 920 350 2600 1140 440 pcalc T - - 1040 399 130 1549 0. The liquid moving down the column gets richer and richer in heavy components.40 T 0. on the other hand.fDistillation Example. Dew-Point Calculations In dew-point calculations. at 125°C XIP. Given: P T = 1520 IIUII Hg Benzene (B) x1 = 0.071 Notice the enriching of lighter component that occurs in the vapor in the above example. Nc N‘ . In this case temperature T and liquidphase composition are known. the heaviest component. Total system pressure is easily calculated (with no iteration involved) from: Vapor pressures Pj are known since temperature is given. given P and 3.40 Fundamentals . has a higher concentration in the liquid than in the vapor. Benzene.’s.671 0. has a higher concentration in the vapor than in the liquid.

12 Isothermal flash . (2-9) Dew-Point Pvessu~e Given).) N. Calculate pressure directly fiom equation (2.9) (T D.12. The equations describing the system are: F=L+V z2F = x. See Figure 2. Both the temperature and the pressure in the drum are given. Isothermal Flash Calculations These calculations combine vapor-liquid equilibrium relationships with total mass and component balances.3 Vapor-Liquid Equilibrizcm Fundamentah 41 Rearranpg: 1 pT= J=1 c WP.L (2.12) + ~/IV 3/1 = xJpJ/pT FIGURE 2.10) (2.2.11) (2. Material of known composition zJ is fed into a flash drum at a known rate of F mols/min. Variables that are unknown are liquid and vapor compositions and liquid and vapor flow rates.

42 Fu-td 0fDirtiUat. =1 (2.13).13) The only variable that is unknown in this equation is the ( V / F )ratio since the P.V/PT 23 x3 = J=11+ 5 1 + (V/F)(<PT .13) .13) is a nonmonotonic function (see Figure 2.13) is satisfied. trialand-error solution is required. A value of zero for ( V / F ) always satisfies the 0 1-0 VIF FIGURE 2. / .13 Graphical representation of equation (2. One guesses a ( V / F )and sees whether equation (2.’s are functions of temperature only. Therefore.on Combining and rearranging: z3F = %(F - V ) + x~P.1) 23 ( ~ / q ( P .1) P. The lek-hand side of equation (2. another iterative.

The volatility of component j relative to component k is defined as: (2. For an ideal (Raoulfs . Before starting any flash calculation.18) where x and y are mol fiactions of light component in the liquid and vapor phases respectively. Relative Volatility Relative volatility is a very convenient measure of the ease or difficulty of separation in distillation. Rearrangement of equation (2. Values of ajkclose to 1 imply that the separation will be very difficult. (2. the system pressure must be between the bubble-point pressure and the dew-point pressure for a m i x t u r e with a composition equal to the feed composition and at the given temperature T.2.3 Vap-Liqud Equilitrrium Fundament& 43 equation.15) (2. but this is not the correct. the fatter is the equilibrium curve.18) leads to the very useful y-x relationship that can be employed when a is constant in a binary system. That is. For binary systems relative volatility of light to heavy component is simply called a: (2. requiring a large number of trays and high energy consumption.The larger the relative volatility a.16) Since T is known. it is vital that one checks to see that the pressure and temperature gven are such that the feed mixture is in the two-phase region.17) A large value of relative volatility a+ implies that components j and k can be easily separated in a distillation column.14 sketches y versus x lines for various values of a. Figure 2. both Pop and PBpcan be calculated explicitly with no iteration involved. real (V/F) ratio.

PL YIX . Relative volatilities usually decrease somewhat with increasing temperature in most systems. yP= = P L X (1 .44 FutzdmaentaLr o Didlatim f law) binary system.constant if the vapor pressure lines are parallei in a L P versus 1/T n plot. Distillation columns are frequently designed assuming constant relative volatility because it greatly simplifies the vapor-liquid equilibrium calculations. This is true for many components over a limited temperature range. a can be expressed very simply as the ratio of the vapor pressures of light and heavy components. particularly when the components are chemically similar.14 Relative volatility on x-y diagram . In other words. a will be independent of temperature. a= FIGURE 2.20) (1 .19) . relative volatility is._ (2.y ) / ( l .y)PT = Px(1 .x ) (2.4 px If the temperature dependence of the vapor pressure of both components is the same.

This is called a “heterogeneous” azeotrope (Figure 2.16.18 sketch typical phase diagrams for these. Positive deviations (repulsion) can give a lower temperature boiling mixture than the boiling point of the light component. where a is the volatility of thejth component relative to some arbitrary base component (usually chosen as the heaviest component.2.084 Nonideality In most distillation systems. If the molecules repel each other. the liquidphase activity coefficient of thejth component. When chemically dissimilar components are mixed together (for example. In this case the activity coefficients are greater than unity (called a “positive deviation” from Raoult‘s law).945 0.4) contains y.5 2. Figures 2.517 0.15 sketches typical activity coefficient data as a function of the light component composition x l . A modest amount of repulsion can lead to the formation of a minimum boiling azeotrope (Figure 2. there can be repulsion or attraction between dissimilar molecules.1 1 1. This can lead to formation of a “maximum-boiling“ moa-ope (Figure 2.225 0.17)] and rearranging lead to the following explicit relationship between any vapor composition y. Figure 2.17).35 0. (2. they exert a higher partial pressure than if they were ideal.3 Vap-LtquiCt Equdihum Fuhmentals 45 For multicomponent systems. If the molecules attract each other. Activity coefficients are usually calculated from experimental data. Wilson. 2. they exert a lower partial pressure than if they were ideal. the system may break into two-liquid phases with different compositions in each liquid phase. . x).16). Empirical equations (Van Laar.) Example. If the repulsion is very strong. Given the liquid compositions and relative volatilities. CYjXj Yj = N. Activity coefficients are less than unity (negative deviations).18). = There are several types of azeotropes. Equation (2.399 0. applying the basic definition [equation (2. An azeotrope exists when the liquid and vapor compositions are the same (x. and given liquid compositions (xis) and relative volatilities (ais). oil molecules and water molecules).200 0.45 0.21) . Margules. and 2.17.20 ffj xJ ffl YJ 3..) are used to correlate activity coefficient data. etc. Azeotropes occur in a number of nonideal systems. Negative deviations (attraction) can give a higher temperature boiling mixture than the boiling point of the heavier component. the predominant nonideality occurs in the liquid phase because of molecular interactions. calculate the vapor compositions: -5 0.

46 FunAamcntaLr o DirtiuatiOn f 4 3 7 1 2 1 I I 0 0.5 1.5 ' 1 0 05 1.o Xl 2.0 72 1.o Xl FIGURE 2-15 Typical activity coefficients as functions of light component composition .

3 Vapor-Liquid Equdibnum Fundammtak 47 FIGURE 2.2.16 Qpical homogeneous “maximum boiling” azeotrope .

17 Homogeneous “minimum boiling” azeotrope .48 FundamCnta.3 o~fDistilhm FIGURE 2.

4 GRAPHICAL SOLUTION TECHNIQUES The equations describing a binary distillation column can be solved graphically using the famous McCabe-Thiele diagrams. COMPOSITION OF ONE LIQUID PHASE COMPOSITION OF VAPOR 4 ABOVE TWO LIQUID PHASES 4 COMPOSlTlON OF OTHER LIQUID PHASE FIGURE 2. The control engineer should be aware that the existence of azeotropes imposes restrictions on the operation and performance of a distillation column.18 Heterogeneous azeotropes .2. These techniques are very useful in gaining an appreciation of the effects of various design and operating parameters. 2.4 Gqhical Solution Techniques 49 A detailed discussion of azeotropes is beyond the scope of this brief introduction.

19 Simple distillation column . in mols/minute.. Liquid and vapor rates in the section above the feed (“rectifjmg‘‘ section) are called LR and V.19 summarizes the system and nomenclature to be considered. Liquid and vapor rates below the feed in the stripping section are called Ls and V.50 Funriamentals $Dirtillation These effects are usually apparent on the diagrams. so one gets a picture of the process. These are assumed to be constant REFLUX RATIO = R FIGURE 2. System Figure 2.

The composition of this liquid pool is the same as the bottom product composition.25) (2.24) if feed conditions and product compositions are specified. F = D + B (2. Equations Overall Balances Mass and component balances can be written around the entire column system. The slope of the line is the ratio of liquid to vapor flow rates in the stripping section.4 G r a p W Solution Techniques 51 on all trays throughout each individual section.21).27) (2. The parameter q will be used to describe the thermal condition of the feed. This straight line can be plotted on an x y diagram (see Figure 2.” Primarily it assumes that the molar heats of vaporization of the components are about equal.28) This is called the operating-line equation.23) (2. In many systems this is a pretty good assumption.X B ) (2. A total condenser is used to produce liquid reflux and distillate product.26) B. This constancy of molar flow rates is what we call “equimolal overflow. + .22) Reflux ratio is widely used as an indication of the energy consumption.20): L x n +1 Rearranging: = vsyn + B XB (2. D = F(zF - xB)/(xD . R = LR/D (2. Thermosyphon. Note that distillate and bottom product rates can be calculated from equations (2. internal. Stripping Section B = F .24) F z ~ x ~ +I x ~ B = These relationships must be satisfied under steady-state conditions. A.2. kettle. The reboiler is a “partial reboiler” (vapor is boiled off a liquid pool).23) and (2. It has the form of a straight line: y = rmc b.D A light component balance around the nth tray in the stripping section yields (see Figure 2. and forced-circulation reboilers are all usually partial reboilers. The ratio of the internal reflux flow rate L R to the distillate flow rate D is called the reflux ratio R.

This is easily proved by letting the point of intersection be xi = yi.VS)xi = BxB LS %+I VS Yn B XB FIGURE 2. = Vsy..20 Material balance on stripping section .52 Fu&ntals of Dit&& The intersection of the operaring line with the 45" line (where x = y) occurs at x. -t B XB Ls xi = Vs xi + B XB (2.29) (Ls . Substituting into equation (2.27) gives: Ls X .

4 Graphad S o f d o n Techniques 53 SLOPE=- Ls vs STRIPPING OPERATING LINE Y 4!io LINE (y=x) XI3 X FIGURE 2.2.21 Operating line o stripping section f .

1 + D X D LR D (2.22) gives: VRyn = L R X n .= B x B x i = XB Therefore. Therefore. the operating line can be easily placed on an x-y diagram by simply starting at xB on the 45"line (see drawing a straight line with slope (Ls/Vs). Ls = V.V.21).29) gives: + Bx.20.30) + -XD VR VR This is the equation of another straight line called the rectifying operating line. = B. yn = -xn+l 7 LR X" +1 VR yn FIGURE 2. Figure 2. Substituting into equation (2.54 Fundumentah 0 f D i s t i l h k From a total mass balance around the system in Figure 2. Rectifying Section A similar component balance around the upper part of the column above the nth tray in the rectifjmg section (see Figure 2. Ls . C. B.22 Material balance on rectifying section .

23 shows both operating lines together with the y-x vapor-liquid equilibrium curve. Our VLE relationship gives us y. if we know x E . and has a slope equal to the ratio of the liquidto-vapor rates in the recufylng section.4 CraphiEal Solution Techniques 55 It intersects the 45" line at X.2.23 x-y diagram showing both stripping and rectiqring operating lines . V. Figure 2. a FIGURE 2.24. These tray-to-tray calculations can be solved graphically by stepping back and forth between the operating line (y = mx + b) and the VLE curve (y = f ( x ) ) . Stepping off Trays Tray-to-tray calculations involve the solution of vapor-liquid equilibrium relationships and component balances. let us start at the bottom of the column with known values of x.Graphically. For example. and Ls/ . as shown in Figure 2.

. (2.31) x B = -x1 L S VS + ($) (2.hti~t~ we simply move vertically up a line through x. We know y. We could plug into equation (2.24 McCabe-Thiele dlagram-stepping between VLE curve and operating lines to estimate number of trays requlred . until the VLE curve is intersected. Then this ni 1 Y YB X FIGURE 2. A component balance around the reboiler gives: Lsx. and then read off the yB value. Alternatively. YE = Vsy. we can solve graphically for x1 simply by moving horizontally on a straight line through y.32) This is the stripping operating line [equation (2.31) and solve analytically for xl. + Bx. utl the operating line is intersected.28)]applied to the reboiler stage.56 Fu-talr ofDktd.

having specified product compositions and operating line slopes.4 Gaphical Solution Techniques 57 same procedure is applied again to the first tray. Values of q greater than 1 indicate subcooled liquid feed.35) . is reached. Feed Thermal Condition The feed to a distillation column can be liquid or vapor.V.). q = 1.33) q is the fraction of feed that is vapor: (2. + . or both. pressure. x2 is determined from y1 by moving horizontally to the stripping operating line. q is a fraction. If the feed is a saturated vapor at its dew point. Use of equations (2. The number of trays in the rectifying section can be determined in this manner. Values of q less than 0 indicate superheated vapor feed. calculate the total number of trays NT and the feed tray N F ) using this graphical technique.30) gives: VRy = g LR x. yi) of the stripping and rectifying operating lines. Thus we can design a column (Le. q = 0. This stepping procedure is continued up through the stripping section u t l ni the intersection of the operating l n s is passed (see Figure 2.24). the operating line slopes are both known if the feed thermal condition and reflux ratio have been specified. This determines ie the number of trays required in the stripping section. As we will show in the next section.)yi = F(l (LR - L. To quanti@ the thermal condition of the feed. and composition of the feed.33) and (2.34) gives: - q)J. = + ZF F (2. (2. yl is determined from x1 (move vertically to the VLE curve). B % B) Using equations (2. depending on the temperature..28) and (2.LR F (2. q= It follows that 1 - L S . Then the rectifying operating line is used.34) If the feed is a saturated liquid at its bubble point.%D D Subtracting gives: (VR .%i + (x.2. D (-@)Xi + . the parameter q is defined as the fraction of the feed that is liquid. Now let us look at the intersection points ( x f . If the feed is a vapor-liquid mixture.24). and the stepping is continued until x.

AM temperature of Lo. intersecting the 45" line at xD. which can be easily drawn given zFand q. and T = R . the lower the reflux ratio specified. Thus the intersection of the operating .q / ( 1 . the more trays are required.F.4)). lines must lie on the q line. Desired product purities: xD. Feed: zF. distillation design involves a classical engineering trade-off between the two design variables: reflux ratio and number of trays. VLE curve in x-y coordinates 4. If the reflux is a saturated liquid at its on. while increasing trays increases investment costs. It is a straight line with a slope of R / ( R + l). As we will show later." Given: 1. q 2. To = vapor temperature. Then = molar latent heat. (2. It is a straight line joining xBon the 45"line with the intersection of the rectlfylng operating line and the q line. Then the smpping operating line can be drawn if xB and q are specified. Design Problem We are now ready to summarize the graphical design technique for determining the number of trays required to achieve desired product purities. Since increasing reflux ratio increases energy costs (V. Operating external reflux ratio: R = Lo/D * If external d u x is subcooled-mually the case-then where C. = Lo + D ) . Figure 2.36) Thus the rectiqing operating line can be drawn if xD and R are specified. This will be discussed fUrther under "Limiting Conditions. It is called the q line and intersects the 45"line at 2. The slope of the rectifjmg operating line (LJVR) can be expressed in terms of the internal reflux ratio LR/D = R. XB 3.58 Fu-tals ofDid&!&m This is the equation of a straight line with slope ( . = molar spec& heat. bubble p i t " LR = Lo. given a reflux ratio.25 shows q ha for several values of q.

Locate xE. 6. and ZF on 45" line.2. Draw 45" line.25) and (2. Calculate slope of 4 line ( . Total number of trays: NT 2.q / ( 1 . Draw VLE curve. 4. (SATURATED) LIQUID Y X FIGURE 2.xD. 3.25 q-line on x-y diagram . Feed tray location: NF Procedure: 1.q ) ) and draw 4 line from zF point on 45" line. 5. Calculate B and D fi-om overall balances [equations (2. Calculate liquid and vapor flow rates in reg and stripping sections.26)]. 2.4 Graphical Solution Techniques 59 Calculate: 1.

Switch to the rectifying operating line and continue stepping. The next step is tray 1. Rating Problems The graphical McCabe-Thiele methods studied in the previous sections for the design of distillation columns are also widely used to analyze the operation of an existing column. and so on. implying a noninteger number of trays. if there are multiple feed points available on the column. When this stepping procedure crosses the intersection of operating lines.(1 . Draw rectifjmg operating line from xD point on 45” line with slope LR/VR 9. The first step corresponds to the partial reboiler. 8.?any other feed tray would require a greater total number of trays). it may be varied.e. or. this is the total number of trays NT.37) (2. Calculate slopes of operating lines: rectifying = LR/VR. The feed tray may also be fixed. Lo. The two most commonly encountered are (with NT and N F fixed): . These fixed-column problems are called “rating problems. Check your calculations by seeing whether the slope of stripping operating line is Ls/Vs. Actual industrial columns seldom achieve this ideal situation. In this case the total number of trays in the column N T is fixed. Start from the XB point on the 45” line and step up the column from the stripping operating line to the VLE curve. Draw stripping operating line fiom XB point on 45”line to the intersection of the q line with the rectifying operating line. the next is tray 2. so an efficiency factor must be used to determine the number of actual trays installed in the column (which must be an integer number). that is. is at its bubble point: LR = Lo = (R)D (2. This last step will not go through the xD point exactly.39) (2. There are a variety of possible rating problems. Thus the feed tray NF has been determined.” in which NT is calculated. When the xD point is crossed.” as opposed to the “design problems. Typical efficiencies run from 40 to 90 percent. stripping = LS/VS.38) (2. Don’t let this worry you. trays on which the vapor and liquid streams leaving the trays are in perfect phase equilibrium. 12.40) VR = LR Ls +D + qF v = VR . In this procedure we have assumed “perfect” or “theoretical” or “100 percent efficient” trays.If the external reflux. depending on the system. this is the “optimum” feed tray (i. 13. 14. 10. 11.q ) F s = LR 7.

4 Graphical Solution Techniques 61 XB -To determine the reflux ratio required to achieve specified product purities and XD -To determine the product compositions that result from specified values of reflux ratio and distillate flow rate Both of these calculations involve iterative. VLE equilibrium relationships.26 McCabe-Thiele diagram for rating problem . and energy balances. component. Notice that in both of these problems. two variables must be specified to define the system completely. Y X FIGURE 2. This magic number of two occurs again and again in distillation (see Section 4).2. mal-and-error solution techniques. Mathematically the two degrees of freedom are the result of subtracting all the constraining equations describing the system (mass. is often called the “degrees of freedom” of the It system. and specified variables) from the total number of system variables.26). Basically one guesses a solution and sees if the stepping procedure produces exactly the same number of trays in each section as has been specified (see Figure 2. Limiting Conditions McCabe-Thiele diagrams are useful for getting a clear picnrre of some of the limiting conditions on the separation that can be achieved in a distillation column.

The actual reflux ratio used must be higher than the minimum.28).1 to 1.27 shows the normal minimum reflux ratio situation.fDistih&n A. Increasing reflux ratio requires fewer trays (less capital cost) but increases energy costs. An infinite number of trays are required to step past the feed plate because of the “pinch” condition (the converging operating and VLE lines). It occurs when the operating lines just intersect on the VLE curve.27 Column operation at minimum reflux ratio . Economic optimization studies have led to the commonly used heuristic (rule of thumb) that the optimum actual reflux ratio is 1. X FIGURE 2.62 Funahmat& . Figure 2. Minimum Reflux Ratio The minimum reflux ratio (for specified product purities and feed conditions) occurs when an infinite number of trays are required to make the separation.2 times the minimum reflux ratio (see Figure 2. Y TAKES AN INFINITE NUMBER OF TRAYS TO STEP PAST THIS “PINCH” POINT.

Ls _-.29).2. reflux ratio s .4 Graphiurl Solution Techniques 63 In some unusual VLE systems. COST It) REFLUX RATIO (R) FIGURE 2. even at very high reflux ratios. No feed is introduced and no products are withdrawn. RID = m. B.28 Costs v . Since D = 0. Also: LR . The L/V ratios in both sections of the column become unity and lie on the 45" line (Figure 2. the pinch between the VLE curve and an operating line can occur at some point other than the feed point. Minimum Number o Trays f The minimum number of trays to make a specified separation is found when an infinitely large reflux ratio is used. but heat is added in the reboiler and all the overhead vapor is condensed and returned to the column as liquid reflux. This situation actually takes place in a column when it is operated under "total reflux" conditions.=1 VR v s Thus a column with fewer than the minimum number of trays cannot achieve the desired separation.

64 Fundumtak OfDistdkkm For a system with constant relative volatility aLH.29 Minimum number o trays required at total reflux f . The L and H subscripts refer to light and heavy components. xDLIxDH ( ~ T ) m + 1= log XBLIXBH log ~ L H A partial reboiler is assumed in the above equation. Y X FIGURE 2. Fenske equation can the be used to solve analy~callyfor the minimum number of uays (AIT)-.

B. If two or more feed streams with different compositions are to be separated in the column. Increasing product purities increases reflux ratio. Going to lower pressures would require refiigeration. Increasingfeed q reduces reflux ratio (condenser load) but increases heat input. they should not be mixed and fed in at a single feed point. but this increase is only very gradual since the upper end of the rect@ing operating h e merely approaches closer and closer to the (1.5 Efem $ Variubles 65 2. Changing Feed Composition z . A. 2. since lower pressure means lower temperatures and higher relative volatilities in most systems. D. and/or Decreasing xB) The number of trays reqrrired in the column is increased. Therefore.2. Operating Case (NTand NFFixed) 1. C. The minimum reflux ratio is increased somewhat. . Increasing Feed q Reflux ratio is reduced as the feed is made colder (as q increases) but energy input to the reboiler increases. columns are usually designed to operate at the lowest economical pressure. Design Case Increasing Product Purities (Raising X. which is very expensive. Thermodynamics tells us that the maximum work of separation occurs when the feed is a 50/50 mixture. 4. 1) point. Redzuzngpressure usually reduces reflux ratio and energy consumption if product purities are kept constant. Instead they should be fed on separate feed trays at locations where tray compositions approximate feed compositions. Minimum reflux ratio is also reduced. Changingfeed mposition changes reflux ratio and energy input but not in the same way for all columns. higher or lower feed compositions should require less energy. The lowest economical pressure is usually the pressure that provides a temperature in the overhead condenser that is high enough to permit cooling water or air to be used for heat removal. Increasing Relative Volatility The number of trays is reduced. Relative volatility has a strong effect on the cost of separation.5 EFFECTS OF VARIABLES Now that the McCabe-Thiele method has been introduced. 3. we can visualize the effects of various operating and design parameters. which changes the operating line slope only slightly at high purities. Therefore. The effect depends on product purities and relative volatilities (see reference 9 ) .

R. B. 8. Robinson.. D. Treybal. 6. 321 (1975). 1969. McGraw-Hill. Elements . DistuatiOn and Rect$catiun. 3. W. McGraw-Hill. I O E C . Kirschbaum. J. E. McGraw-Hill. M. C.66 REFERENCES 1. 4.f Fractional Diadlutwn. New York. Holland. King. Luyben. 7... McGraw-Hill. 5. 1950. R. Desgn $Equilibrium Stwe Promses. New York. New York. New York. Dirtillation En&eerinJ. Fundtwentals ... Dirtillation MCG~W. Springer. Mass-Transfer Operatwns. Chemical Publishing Co. 9. C. . R. L. D. Separation Promes.. f M ~ l ticmnpmKnt Dist&ion. E. 1971. Gilliland. 1981. S. Billet. 1963. New York. Fundamentals 14. 1968. Smith. Hill. McGrawHill. 1967. 2.. C. New York. and E..1979. Van Winkle.

Material balance control on the condensate may be accomplished in several ways : 69 . If subcooling is required for a pressure or vacuum column.3 h Overhead System Arrangements 31 INTRODUCTION . The engineer must have some idea of what constitutes effective equipment configurations and arrangements. the preferred method is to have the condensate-temperature controller manipulate the vent flow in some way. and having reviewed the fundamentals of distillation. and satisfy part of the column material balance requirements. partly to control the amount of inerts in the system. aving considered a particular approach to an overall strategy for controlling distillation columns. and partly to control product losses through the vent. problems. The supporting mathematics and theory are covered in Part 1 1 1. The column overhead system is generally more complicated than either the feed system or the bottoms system. Condensate is generally subcooled at least slightly. control schemes. The design of a satisfactory distillation control system involves far more than theory or mathematics. compared with that of atmospheric or pressurized columns. at un least for simple columns. provide reflux flow back to the column. In the case of vacuum columns. there may be problems associated with the control of the vacuum jets. Typical equipment. It usually must condense most of the vapor flow fiom the top tray. and must be able to recogme when undesirable side effects are apt to interfere with an otherwise good control system. remove inerts. as well as an appreciation of equipment performance limitations. This arrangement avoids the instabilities and other control difKculties that often characterize condensate-temperaturecontrol systems based on manipulation of the condenser cooling water. and column turndown is usually limited. let us now t r our attention to some practical aspects. and solutions are discussed in this section. partly to minimize the likelihood of flashing and cavitation in valves and pumps. maintain column pressure in the right range.

The design itlustrated in Figure 3.2). This type is popular i the chemical industry because it m i n i m i z e s condenser cost when highly corrosive process materials must be handled. This is probably the most popular type in petroleum refineries. the cooling water valve is located at the exit to ensure that the shell is flooded. column inerts are usually vented more easily (ix. For the same reason. this type of condenser cannot be operated partially “flooded. we find at least five different kinds of condensers: 1. As an alternative. each with a valve if the exchanger is operated flooded (see further discussion in Chapter 15). 3. 2.Although it is less exact. In addition. takeoff should be by “averaging” pressure control. The cooling water valve is normally at the exchanger exit to make sure the tubes are fdled at all times. thereby minimizing corrosion. Since the exit water is hot. fiom column overhead composition. Because all vapors must pass through the tubes. Horizontal shell-and-tube condenser with liquid coolant in the tubes and vapor on the shell side (Figure 3. -Same as foregoing except that distillate flow is set by reflux drum level control. With a longer condensing path. Vertical shell-and-tube condenser with liquid cool an^ on the shell side n and vapor entering the tubes at the top (Figure 3. a reflux drum. r r a y m a A -Flow control of reflux is cascaded. the speed of venting inerts at startup time is limited. This condenser commonly has at its lower end a vapor-liquid disengaging pot. If the column top product is a vapor. If a smooth flow to the next step in the process is needed... the valve may need anticavitation trim.2 TYPES OF CONDENSERS In chemical and petroleum plants.1). with less pressure drop) through condensers of this design. . this term is widely used in the petroleum industry. at startup time. if possible.1 has two vents. should be employed. with averaging level control of distillate. Some designs bring the vapor in at one end and vent uncondensables at the other. it is much better suited to partially ‘‘flooded” operation. vapor may be taken off on flow control cascaded fiom top composition control while column pressure is controlled by heat input. which also serves as a condensate receiver.” Again. In this book we usually refer to the condensate receiver-the vessel that receives condensate &om the condenser-by another name: reflux drum. By comparison with the vertical design discussed below. it is also better suited to applications in which it is desired to absorb the maximum amount of low boilers in the condensate. and distillate overflows from the vapor-liquid disengagement space beneath the condenser. Sometimes condensate is taken out through two drawoffs instead of one.70 Overhead Syst. particularly if 316 SS tubes are used.

3. in-column head condenser (sometimes called “dephlegmator”). -Air-moZed umdensers (Figure 3. vapor in shell . The latter has a “hat” over it to prevent condensate fiom dropping back down the column. or externally through a gravity flow line with a control valve. and comes down the tubes with the condensate.4) -Spay ma’ensen (Figure 3. -Ve?-td bunde with wohnt on shell sia!e. Here condensate is recirculated through VAPOR CONDENSATE FIGURE 3. reverses direction.1 Horizontal condenser. -Horizontal tube bundle with wohnt i tubes (Figure 3. Reflux may return internally via an overflow weir. The vapor n comes up &om below and condensate drops into an annular space around the vapor nozzle.2 TrparofCondensen 71 3. Internal.5). This design comes in two variations: “reflux” design. Here we have a number of different designs. and a design with a chimney in the center such that the vapor rises in it. where the vapor goes up the tubes and is countercurrent to the condensate f a h g down.3).

Vertical condenser. The preferred overhead system for atmospheric columns is shown in Figure 3. This type of condenser is most commonly used in vacuum service because of its low pressure drop. and inerts are vented to a flare VAPOR COOLING WATER LEVEL MEASUREMEN1 T CON DENSATE FIGURE 3 2 .72 Overhead System Arraqpnmts a cooler and returned to a spray chamber. The condensed vapor falls into a reflux drum that should have 5-10 minutes’ holdup (relative to condensate rate). vapor in tubes .6. 3 3 ATMOSPHERIC COLUMNS .

many cooling water valves do not: have adequate turndown. At startup time total reflux may be achieved by using the reflux valve to control the level in the condensate receiver. Tyreus. D. n i s is similar to the one recommended later for pressurized or vacuum columns. As shown by a study by B. to minimize product losses.3 Alternative overhead s s e for pressure column ytm .3 Atnmpberic Cohmm 73 or cleanup system. This compounds stability problems. FIGURE 3. Further. Figure 3.3. we need an increasing controller gain and decreasing reset time as total heat load increases.7 also shows a more commonly encountered tank arrangement where the reflux drum is common to both the top product system and the reflux system. however. Note that inerts usually should be added a@er the condenser. it is necessary to add inerts ahead of the condenser. a vent system such as that shown in Figure 3. Finally.7 should be used. A potential and frequent source of trouble with both arrangements is the control of condensate temperature via cooling water." for a constant subcooled temperature. process gain ("C/pph CW) and dominant time constant both decrease as total heat load increases. for pressure control. they are wide open in summer and almost closed in midwinter. subcooling heat load must be a reasonable fraction of total heat load-say 5 percent-or the system will lack adequate sensitivity. A small and a large valve in parallel should often be used. For those columns that must be protected from atmospheric oxygen or moisture. Sometimes.

74 Overhead Spent A m a q p w n ~ FIGURE 3.4 Air-cooled condenser .

) FIGURE 3 5 . use dual. use a cascade temperature water-flow control system. This will have a secondary advantage of reducing the probability of cavitation in control valves and pumps. 2.3 Atmospheric Columns 75 The suggested approaches to a v o i h g these ddKculties are as follows: 1. Select the number of degrees of subcooling so that the sensible heat load will be at least 5 percent of the total heat load. (See a s discussion in Chapter 11. The ratio of cooling water rate for the maximum summer heat load to that for the minimum winter load is often two to three times as great as process turndown. Spray condenser . If summer-winter heat-load variations are sufKciently severe. the temperature detector preferably should . If water-header pressure fluctuations are a problem.) 4 For the horizontal condenser. For the vemcal condenser. lo Section 6.3. The smaller valve should open first and will provide adequate winter cooling. be located in the liquid line just beneath the condenser for maximum speed of response. 3. split-range water valves. (See Figure 3. the temperature detector should be located in a trough at the lower end of a drip collector just below the tube bundle and above the reflux drum.8.

76 Overbead System Arraqgements FIGURE 3.6 Preferred overhead s s e for atmospheric column ytm .

. FIGURE 3.. or under vacuum.10).. 6.. a limitation to this technique: for protection . Another. however. or perhaps adaptive gain and reset. to compensate for changes in condenser dynamics as condensate rate changes.9). This is discussed more fully in the next section. As an alternative to Item 5.-. completely different approach is to run the column at a slight pressure.-. An override from cooling-water exit temperature is also normally needed. 7.. one may use a recirculating coolant system (“tempered” coolant) with condensate temperature control of makeup coolant. Then the condenser cooling water may be manipulated by the pressure controller while subcooling is controlled by manipulating the vent (see Figure 3. say 3-5 psig.7 Alternative overhead system for atmospheric column .. There is. This keeps the condenser dynamics constant and eliminates the problem of retuning the controller as the load changes (see Figure 3...3..3 Atmaspheric Columns 77 5.. The controller should have auto overrides (see Chapter 9).

8 Themowell installation under vertlcal condenser .78 Overhead Spem Awatgemmti FIGURE 3.

Increasingly.3. the cooling-water pump stalled. and sometimes lower. Condensate temperature became too low. The water in the shell began to b i . unaided condensate temperature control is not recommended. it is increasingly common to provide no condensate temperature control. cooling water exit temperature should usually be limited to a maximum of 50-60°C (122-140°F).4. an alert operator shut the column down before any damage occurred. the valve could not pass the required volume of steam. an internal reflux computer is required (discussed in Section 11. Further. An additional problem with condensate temperature control. This saves a control valve. Fortunately. For accurate control of internal reflux. Further. via coolingwater manipulation. As a consequence. the quality of cooling water is sometimes so poor that a minimum velocity must be maintained in the exchanger to minimhe fouling. to minimize the hazard of winter freezeup. As pointed out by Bolles. an atmospheric column with such a control system was running at a very low feed rate. it is necessary to limit subcooling in some columns to avoid foaming on the top tray. but to fzltl with fidl cooling at all times. There should be an override from cooling-water exit temperature. Finally. a limiter should be provided to prevent complete valve closure (see Chapter 9 ) .3 Atmospbetic Gdurnns 79 against fouling and corrosion. discussed in Section 3. FIGURE 3 9 .1). so the controller closed the cooling-water valve located in the exit ol line fiom a vertical condenser. override controls are used to provide this protection. and product vapor issued in great quantities from the vent. relates to column safety. In an instance with which one of the authors is painfully familiar. Tempered coolant system .' however. Another alternative is to use exit-cooling-water temperature control.

For large columns that must be started up and shut down frequently.4 VACUUM AND PRESSURE COLUMNS-LIQUID PRODUCT The preferred arrangement for a vacuum or a pressure column with a large amount of inerts is shown in Figure 3.80 Overhead System Arrunpnents 3. In such cases a manually set vent or bleed valve is ofien adequate and no direct control of condensate subcooling is necessary.12. via condenser cooling-water manipulation. it may be necessary to use a cascade temperature-vent flow-control arrangement. The objections to. . “AC” means air-to-close. for example.* is well suited to either vacuum or pressure columns when the amount of inerts fluctuates over a wide range. A split-range adjustment of the two positioners (see Chapter 11. In this event the arrangement of Figure 3.10 may require an impractically small vent valve. the vent or bleed valves may be tied to the pressure controller to work in parallel with the condenser cooling-water valve. pressure is then controlled by (1) throttling the vapor takeoff if there is a large amount of inerts. condensate temperature control. normal flow of air or gas through the two valves in series is economically small. were stated earlier. on a column in a semicontinuous process that is shut down and started up every day or so.11.13. where column feed rate varies signhcantly. and that must handle severe transients during the startup period. Here the inerts are pulled off or blown out through a vent line in which there is a throttle valve manipulated by the subcooled-condensatetemperature controller. The vent line is connected to a pressure-dividing network with two control valves connected so that as one opens. the other closes. the arrangement of Figure 3. When an expensive inert gas such as N2 must be used. even though each valve is sized to handle a maximum flow equal to five to ten times the average. the low-pressure source is usually a steam jet. This facihtates getting the column on line at startup. For a vacuum column with a small amount of inerts. A more complicated but more flexible arrangement. and split-ranged with the small one. such as that of Figure 3. An example is shown in Figure 3. Section 10) permits both valves nearly to close when the controller output signal is at its midrange value. If the downstream pressure fluctuates too much. For many columns the vent flow functions primarily as a purge and is small enough that moderate changes do not affect column operation. It should be acknowledged that many engineers today prefer to control condensate temperature by manipulation of condenser cooling water.10. It has worked well. For other cases. an additonal large vent valve is sometimes installed in parallel. As far * The symbolism “ A 0 means air-to-open. or (2) throttling an air or inert-gas bleed if there is only a small amount of inerts. Since the sum of the two acoustic resistances is always high. it is common to minimize or eliminate split-range overlap to reduce consumption even further. is better. and difficulties with. For a vacuum column. with a controlled bleed from the atmosphere (or source of inert gas).

10 Overhead s s e for vacuum or pressure column-large amount o inerts ytm f .4 Vacuum an/i P m r e COltrmnr-Lq~id P m d m 81 FIGURE 3.3.

If cooling water is adjusted manually. pressure is controlled by manipulation of makeup and vent valves. They have the advantage of minimizing cooling-water flow rate for any given heat load.82 Operhead System An-anpnenB as we can tell. Section 2). Their use also minimizes subcooling-and there are instances where this is desirable-but at the expense of variable condensate temperature.7. the flow is either insufficient or excessive.11 Overhead system for vacuum column-small amount o inerts f . This kind of control is often implemented as shown in Figure 3. they are equally valid for pressure and vacuum columns as for atmospheric columns. This can cause variable internal reflux d e s s it is compensated for (see Chapter 11. c TO SOURCE OF VACUUM FIGURE 3. So-called "water savers" are cooling-water exit-temperature controls.

should usually be of the "averaging" type. which provides slow. gradual correction. this procedure is recommended for columns with a vapor product (see next section) where condensate temperature is not conuolled.12 Alternative overhead system for pressure or vacuum column-small inerts amount o f . This fits in well FIGURE 3. We have previously indicated that pressure control.4 Vacuum and Pressure Columns-Liquid Product 83 For a large amount of inerts. the engineer should keep in mind that these columns have a narrow range of operation. One last consideration should be noted here-that of dynamics.3. if used. In designing controls for vacuum columns. it is certainly feasible to control pressure by throttling the vent flow. The range in column pressure drop between flooding and tray instability for a perforated tray column in vacuum service may be no more than 10-15 percent.

It should be acknowledged. cannot be changed rapidly. controlling condensate temperature via cooling-water manipulation requires overcoming the condenser dynamics. 3. small changes in pressure can create relatively large changes in driving force. This is so because process-to-process heat exchangers are ofien designed for very small temperature differences.3 Alternative pressure control system .84 Overbcad System Awanpnma~ with condenser cooling-water manipulation since condenser heat loads. On the other hand. the arrangement FIGURE 3 1 . however. controlling condensate temperature via bleed manipulation should be comparatively rapid. like those of most heat exchangers. that ''tighf' pressure control is required in some heat-recovery schemes. If the condenser is external to the column.5 PRESSURE COLUMNS-VAPOR PRODUCT Pressure columns are sometimes operated so that the product comes o f in f the vapor phase.

or cooling-water-supply fluctuations.14 Overhead system for pressure coIumn-vapor product . this may be done as follows. An alternative arrangement. to provide reflux-to-feed ratio control. "Averaging" pressure control should be used. By measuring the cooling-water temperature rise FIGURE 3.14 may be used.3 5 Pressure Columns-Vapor Product . used especially when the condenser is built into the head of the column. which. 85 of Figure 3. A level controller on the reflux dnun balances the rate of condensation against the reflux flow by manipulating condenser cooling water. and when maximum smoothness of vapor flow is desired.15. because of feed flow or composition fluctuations. used successfully is discussed in Chapter 11. pressure control should be cascaded to vapor flow control. This internal reflux arrangement works well if a heat-computation scheme is used for control. Here column pressure is controlled by manipulating the vapor vent valve. in turn. is that of Figure 3. Instead it must be controlled indirectly by manipulation of condenser cooling water. Section 4 If it becomes necessary. Direct measurement and control of reflux are not possible since the flow is internal. A scheme that we have . may be reset by a vapor-composition controller.

In one system the condenser is at a lower level than the receiver by 10 to 15 feet. This means FIGURE 3.15 Alternative overhead system for pressure column-vapor product . 3. we can calculate wR. Several configurations have been employed.6 MISCELLANEOUS PRESSURE-CONTROL TECHNIQUES Hot-Vapor Bypass Another common method for pressure control of p r a a columns involves running with maximum cooling water and bypassing part of the hot gas around the condenser. qc.86 &erhead System Anmgemenk and flow rate. we can calculate the heat transferred. This calculated wR can then serve as the measured variable in a reflux flow control system that uses condenser cooling-water flow rate as the manipulated variable. Knowing the latent heat of the reflux.the reflux flow rate in porncis per hour.

as shown in Figure 3. other examples will be discussed later. It also lowers the operating pressure on the condensing side. as shown in Figure 3. which limits heattransfer capabilities. Dynamic problems with such a system can be severe. Eventually. The liquid in the refiux drum is subcooled. however. There are some practical problems that must be taken into account. thereby lowering the liquid level. Maximum cooling-water rate is normally used. that the column pressure has risen. the liquid level will run high. At low heat-transfer loads. the controller opens either the distillate or reflux valve. violent surging and hammering may ensue. A mathematical analysis is presented in Chapter 15. for example. or by injecting inerts into the incoming vapor partially to blanket the tubes. If the pressure gets too high. the hot-gas line around the condenser has m valve in it.* This may be minimized by designing adequate clearance into the condenser. lower level. thereby dropping the liquid level and increasing the heat-transfer area available for condensation.6 M i c e h m Pressure-Cuntrol Techniques 87 that the condenser runs partially flooded. The temporary “wrong-way” pressure response is commonly called <<. Suppose.16A. for example. The pressure controller pinches the bypass valve to force more vapor into the condenser. At low heat-transfer loads. so there is condensation of vapor at the liquid-gas interface in the drum. the horizontal condenser of Figure 3. Another technique sometimes encountered involves throttling the vapor to the condenser. as suggested by Holland& and shown in Figure 3.3. A valve on the liquid from the condenser floods the condenser to hold column pressure. Consider.1. Another configuration. These hot-vapor bypass systems are not recommended for systems with even small amounts of inerts. Flooded Condenser A pressure-control technique that is growing in popularity involves partial flooding of the condenser without a hot-gas bypass. mverse response”. Common practice. Pressure drop of uncondensed vapor from inlet to the two exits caused a low liquid level in . This results in a temporary increase in pressure since it takes time for the condenser level to drop.17. condenser contents drop to a new. The vapor enters at the center and uncondensed gas exits at the two ends. Another problem was observed by Mueller5 on a partial condenser. As suggested by Chin3. has the condenser mounted above the reflux drum.16B. perhaps as a consequence of increased boilup. This suffers from the drawback of requiring a large valve. The liquid line from the condenser should extend down into the liquid in the dnun so that the cold liquid is introduced near the bottom of the drum. is to bring the condensate into the bottom of the reflux drum (or at least under the liquid surface) and to bring the hot-gas bypass into the top of the drum. If there is insficient clearance between liquid level and the top of the shell. the liquid inventory in the shell was high. A vertical reflux drum is recommended to reduce this interfacial area. which permits a higher rate of condensation and causes the pressure to be restored.

16A Column pressure contml by hot gas bypass .88 fierbead Spem Awaqpun& FIGURE 3.

3.6 Mkcellaneow Presswe-Control Techniques 89 FIGURE 3.16B Column pressure control by hot gas bypass .

In some cases it will be helpll to rotate the tube bundle about its axis just slightly. These levels were so close to the exit nozzles that severe entrainment of liquid in the leaving vapor was observed. In designing a flooded condenser. a ground level condenser FIGURE 3. one must take care to choose a tube pitch that will not cause large changes in exposed tube area per change in condensate level. which favors good overhead composition control. This particular problem was solved by installing a bypass line between vapor inlet and vapor outlet. the vapor and reflux lines can be short. it will be desirable in some cases to provide an override that will open an inert gas valve connected to the vapor inlet. 3 7 GRAVITY-RETURN REFLUX VERSUS PUMPED-BACK REFLUX .90 Overhead System Arranpnenk the center of the shell and high levels at the two ends.17 Column pressure control with flooded condenser . level taps and a level transmitter should be provided. Many arguments have taken place as to whether it is better to locate the condenser overhead and provide gravity return reflux. To protect the exchanger from damage at high liquid levels. On the other hand. For troubleshooting. or to locate the condenser at ground level (or nearly so) and pump the reflux back to the column. With the condenser overhead.

Here a reflux flow measurement is coupled through a controller to a distillate valve. one must be carefd to design the vapor piping and condenser to have a low pressure drop compared with the difference in head between the point of reflux return to the column and the condensate receiver liquid level. unless the drum has a large cross section. the scheme of Figure 3. particularly if the overhead surge drum is also located at the top of the column. For this application a valve with linear trim will have a h e a r installed characteristic if line drop is negligble. and it is harder to remove the condenser tube bundle for maintenance. Overall a properly designed gravity-flowreflux system is sigdcantly cheaper than a pumped-back reflux system. piping and instrumentation can be very simple. To avoid this problem. The support problem can be minimized.3. coolant-in-shell condenser. Here reflux drum level is controlled by throttling distillate flow. For maximum effectiveness the liquid pool in the vapor-liquid disengagement space should have a large cross-sectional area. a more expensive columnsupporting structure is required. Perhaps the most common arrangement is that of Figure 3.19. it backs liquid up into the receiver. Since the surge tank needs a level controller. flow will vary only when the valve position is changed. that is. If the individual valve is shop calibrated. on the other hand. and the overflow weir should permit a wide range of overflows with only a small change in head. Consider. If an overhead condenser is used. Another method of controlling gravity return reflux is shown in Figure 3.21. The reflux flow or flow-ratio controller will usually be fast enough that this will not be a problem for reflux flow. which features a vapor-liquid disengagement space built into the lower section of a vertical-tube. When this valve is pinched. then valve stem position can be accurately related to reflux flow.7 Gravity-ReturnR t f w Venus Pumped-Back R@ux 91 is often easier to maintain. one uses the arrangement of direct return reflux and overflow distillate. one may design a distillate overflow system that provides constant head for reflux. which is especially important if fouling and corrosion are problems. But one a s nseds a higher head cooling-water pump. for example. Since this tank with its contents is often far heavier than the condenser.18. is that variations in level will cause momentary changes in both reflux and distillate flows. however. Use of this technique leads to the arrangement of Figure 3. lo however. For all gravity-returnreflux systems. may have to be cascaded to distillate flow control. . A disadvantage. If. the condenser can be located overhead with only a modest increase in structural requirements over a ground-located condenser. but the equipment that needs to be installed at a high elevation is minimized.20. by building the condenser into the top of the column. a plot of reflux flow versus valve stem position will be a straight line. Reflux Flow or Flow-Ratio Control When reflux is flow or flow-ratio controlled. then the reflux does not come from the overhead surge drum and this vessel can be located at ground level. there is no savings in instrumentation. it is also probably safer since there is no pump to fail. with head across the reflux line fixed. The level controller. Then.

as shown in Figure 3. An elegant way of doing this is to cause reflux to overflow through a Sutro weir. Reversing the controls-that is. employing automatic reflux flow or flow-ratio control and allowing distillate to be the difference flow-provides a positive cure.6 Distillate Flow or Flow-Ratio Control For those columns with gravity return reflux. FIGURE 3. This is commonly called “reflux cycle” and has a typical period of several minutes. It has been observed primarily in columns where reflux flow is the difference between rate of condensation and distillate flow rate. a severe oscillation in overhead vapor flow to the condenser is sometimes encountered.22. that is. where distillate is on automatic flow control or column-composition control. The Sutro weir has the advantage of being a linear weir.18 Gravity flow reflux (flow controlled) and distillate (level controlled) .92 Overheart Syrtem Amaqpnwna causing more reflux to overflow.

7 Gravity-ReturnR $ w Venus Pumped-Back Rejlw 93 A mathematical study of this phenomenon has been published. T. CONDENSER L I QU I0 -VAPOR DISENGAGEMENT SPACE FIGURE 3. -Select the flow-metering orifice to hold liquid head in the reflux line at (. there will be no reflux cycle. is the condensing temperature. -Use a condenser with a recirculating coolant.3. -Increase column operating pressure. -Use a horizontal condenser with vapor and a generous fiee volume on the shell side. This increases vapor density and decreases dT.19 Liquid-vapor disengagement space built into condenser . &a at the bottom of the pot with the Sutro weir. vertical condenser with many tubes (vapor inside tubes). If subcooling is zero./dP. -Keep condensate subcooling to a minimum. -Use a large-diameter vapor line to reduce acoustic resistance and to increase acoustic capacitance. thereby fUrther improving stability.’ It was found that the following measures are helpll in increasing stability and minimizing cycle amplitude: --Provide a difference in head between the liquid level in the reflux drum (or reflux accumulator) and point of reflux return to the column at least ten times as large as the average pressure drop across the vapor piping and condenser. or use a short.

94 o~erbead Spem Arranpnents FIGURE 3.20 Gravity-flow reflux system with ground-located surge tank for distillate .

7 Gravity-ReturnR $ w Versus Pumped-Back R$w 95 FIGURE 3.3.21 Control o gravity reflux flow rate by throttling top product flow f .

22 Control o gravity reflux flow rate by overflowingthrough Sutro weir and by throttling f distillate flow .96 Overbeead System Awqqement~ FIGURE 3.

fiom best to least desirable. i ? a Qz = = A The last scheme is shown in Figure 3. proportional level control. proportional level control cascaded to reflux flow control. ft3/min. thereby causing a reflux flow oscillation as a result of an intermittent siphon action. 2 10 psi. Plant experience indicates that it does not completely eliminate the cycling. when reflux flow must be the difference flow between rate of condensation and distillate flow: -Pumped-back reflux. proportional level control. Another important point for gravity return reflux is the method of connecting the reflux piping to the column. Each of the two piping arrangements of Figure 3. -Gravity-flow reflux. Ap.23. from Sutro weir inflow from condenser. Ap. -Gravity-flow reflux. TH 3 3-5 minutes. Note that T~ is the hydraulic time constant of the surge tank with Sutro weir: where aQ dH = = constant for a sutro weir outfIow. T~ > 3-5 minutes. cylindrical design assumed). TH is the level control time constant (see Section control. proportional level control cascaded to reflux flow 3-5 minutes. T H 3 3-5 minutes. .10). fi3/min. surge tank with Sutro weir. -Pumped-back reflux. but reduces the amplitude by a factor of ten or more to an acceptable value. TH 3. TH > 3-5 minutes. Inerts sometimes accumulate in this pocket.3. There have been cases where hot vapor was sucked back into this pocket and caused such severe hammer that the reflux line and column nozzle were ruptured. it is sometimes necessary to take stronger measures. Make sure the reflux line has a sufKciently high hydraulic resistance. cross-sectional area of vessel. -Gravity-flow reflux.24 has an undesirable upward loop just before entry into the column. It is simple and inexpensive to fabricate and permits locating the condenser at a lower elevation than do any of the other techniques. 3 5 psi. f (vertical.7 Cravity-ReturnR$wc Vmw Pumped-Back R$ux 97 Although the preceding are helpll. The following list of overhead schemes is in order of preference.

surge tank With Sutro weir.98 Overhead S j m m An-aqpncn& FIGURE 3.23 Gravity-flow reflux. T~ > 3-5 minutes .

4 Undesirable piping arrangements for returning reflux to column FIGURE 3 2 . the piping may enter horizontally. They have demonstrated.8 Gmtrol Tecbniqua with Air-Cmlcd condcnrm 99 This phenomenon is particularly troublesome with vacuum towers where some slight air leaks are unavoidable. or hs with a slight indination as shown. certain control problems.3. In recent years air-cooled heat exchangers have gown enormously popular. The preferred arrangement of Figure 3.25 avoids ti kind of flow instability. 3 8 CONTROL TECHNIQUES WITH AIR-COOLED CONDENSERS .5 Preferred piping arrangement for returning reflux to column . however. They are far more FIGURE 3 2 .

Small holdups favor good composition control. 6 . 3. 7. requirements may be much greater. This is discussed in more detail in Chapter 5. low-head pump therefore is required. 310 LEVEL CONTROL OF CONDENSATE RECEIVER . In extreme cases tempered coolant is taken &om and returned to a supply tank that is temperature controlled. and the condenser must be designed for a small pressure drop on the coolant side to minimize pump horsepower requirements. But when the holdups are part of a feed system for another process step. or at least adequate. It eliminates problems with high-freezing-point condensate that might p l q the condenser if once-through coolant were used. . Toplocated fans provide much better protection against rainstorms (see Figure 3. Tempered coolant is employed for either or both of two reasons: 1. Internal reflux computers (see Chapter 11). Variable-pitch fans. Partial bypass of hot liquid from upper section of the exchanger and mixture with cold liquid leaving at the bottom. comments or suggestions regarding required holdup will be primarily fiom the standpoint of getting good.9. Condensate-temperature and column-pressure control are easier. control of the column with which the holdups are associated.4). 2. The temperature rise per pass is kept small. 5. Various techniques have been devised to control the rate of heat transfer or to compensate for condensate temperature changes: 1. rapid temperature control. This permits sensitive. Adjustable louvers in the exchanger housing to control suction air flow. Use of induced-draft rather than forced-draft exchanger designs. The two control valves may be replaced by a single three-way valve if the recirculating flow is not too much larger than the return flow.100 Overhead System Arraqqemenk sensitive to atmospheric changes such as rainstorms or even changes in wind velocity &an are liquid-cooled exchangers. 2.9 “TEMPERED” VERSUS ONCE-THROUGH COOLANT The term “tempered” has been applied to coolant systems that feature a high circulation rate through the condenser as shown by Figure 3. 4. A high-flow. 3. Variable-speed fan drives. Condenser dynamics are radically improved over those achieved with once-through coolant. AND REQUIRED HOLDUP In this section. Flooded operation. Speed of response is greater and condenser dynamics do not change with load changes.

26.-. one must also have overrides on reflux for the same reason. Less commonly. For level control via reflux-flow manipulation.) Since. (With electronic analog or microprocessor controls.fCondensate Receiver and Reqzrired Holdup 101 Level conrrol in condensate receivers or reflux drums is commonly achieved by manipulating either top product flow or reflux flow. For maximum flow smoothing. For this example level is maintained by throttling distillate flow. a relatively simple control system can be used. an alternate design with nonlinear gain and reset may be used-see reference 12. however. If a PI controller .- FIGURE 3.. . Note that the PI level controller must be enhyced with highand low-overrides (called “auto overrides”) to keep level within the vessel. it is necessary to sacrifice flow smoothing in the interest of good composition control.3.26 Condensate receiver level control via distillate . Cascade control is used to eliminate flow changes caused by control-valve upstream and downstream pressure variations. it uses the cascade PI level-control to flowcontrol scheme of Figure 3.. For the first two cases..10 Level Control . Note that the flow measurement must be linear (or linearized) for stability reasons. overhead level control is accomplished by adjusting boilup or by adjusting condenser cooling water. there are two outflows. The quantitative design is discussed in Chapter 16.

* If there is sigdcant level self-regulation. For pneumatic systems inexpensive fixed-gain relays are available for this application. dimensionless Kc = = = k.For pneumatics the bias is so set that the output is 9. (psi for pneumatics) controller gain. and the bias adjustment therefore should be treated as a calibration adjustment rather than as a "tuning" adjustment. Chapter 9 discusses overrides further. As indicated. it should be tuned for tight control of level.27. If level gets too high.0 psig.27 also shows simple ovemdes that act on the distillate valve. if level becomes too low. TH is also important in the design of PI level controls.102 Ovcrhead Sjsm Alrangemenn is used (usually without overrides). If the manipulated valve has a linear installed flow characteristic (preferred). not a v e r a p g level control. the distillate valve is opened. It n is expressed a little differently. This means that the control valve is closed at the 25 percent level and wide open at the 75 percexat level. and if there is no level self-regulation (if ApI does not change appreciably with change in level). Figure 3. A& is input span of valve positioner corresponding to full valve travel as flow-sheet value of manipulated flow. one should use cascade level-flow control. the &stillate valve is closed. for cascade controls. it uses a controller with gain 2 (50 percent PB) . For this application it is probably more appropriate to use a proportional-only controller as shown in Figure 3. however. ." then the dynamic response of the proportional-only7 level control system may be defined by a first-order time constant: A (3-3) where TH is in minutes = = A cross-sectional area..0 psig when the input is 9. cylindrical design assumed) level transmitter gain Kd - -- A&& AZ3T is the level transmitter span corresponding to the output AHT' signal span A & .(eEs/ABc) for valve with linear installed flow characteristics. ft2. ft3/min t As will be seen i Chapter 16. These numbers shouid be regarded as part of process design. of seal pot (vertical.

27 Proportional-only condenser seal pot level control via reflux flow .10 L m l G m t d OfcOnaCnrate Rtctipcr and Required Holdup 103 FIGURE 3.3.

the available holdup must be very small. by the use of special techniques. that is. If. the various control schemes usually will require a rH 2 2 minutes. High-performance valve positioners probably will be required. Holdup time is usually considered to be equal to volume divided by throughput. and A$& = A$L: = ~QFS AHT = Kc keQFs A rH (3. Kc = 2.104 Overhead S s e A m n p n e n B ytm During the early stages of a design project. If pneumatic instruments are involved. 7. the preceding is adequate for twopipe designs with up to a 1200-foot one-way distance for 1/4-inch OD plastic tubing or a 2000-foot one-way distance for 3/8-inch OD plastic tubing. (See discussion on valve sizing in Chapter 11. A AHT and . for a linear installed valve characteristic. and for pneumatic instruments various methods are available for minimizing lags and improving speed of response. the level nozzle spacing usually must be determined before control valves are sized.5) In the discussion that follows. Typically. = 4. however..) Then. to design for rH or [rH]oR less than 1 minute. Process engineers often think of “holdup time” rather than a time constant. it should be preferably in a separate vessel outside of the reflux path. for k. For any given scheme. is a multiplying factor typically in the range of 2-6. for process reasons. it is sometimes possible. This may require a rH much greater than 2 minutes.’ Experimental data for long pneumatic tubing runs are gven in reference 10. horn equation (3. System performance then should be checked by frequency-response methods or computer simulation. however. For electronic-analog or microprocessor controls. As noted earlier. [ T H ] o R 2 1 minute. if additional volume is required for smoothing out feed to the next process step.3). valvesizing procedures lead to: (3-4) where k. If we think in terms of the volume corresponding to the level transmitter span. the limiting factor will be the speed of response of the valve-positioner valve-actuator combination. one should make sure that override time constants are at least 1 minute.

= AtIL. larger reflux drum holdups have proved particularly desirable for columns that occasionally experience slugs of light ends or inerts in the feed. if level goes too low. we will usually use K. this is accomplished by a relay with a gain of 4 and a h g h selector (HS). may be minimized by appropriate use of overrides. This is accomplished by another gain 4 relay and a low selector (LS). As shown in Figure 3. if rH 3 2 minutes. Without several minutes of reflux holdup. . = 4 Therefore..) Let us now see what effect the difference between top product flow and reflux flow has on rHand [rHIoR. we want to pinch reflux flow.10 Level Control of C k a t e Retziver and Required Holdup 105 A8. For the same rH.only one eighth the volume is required. = 2. we want to increase the reflux flow. This permits faster and tighter composition control. Larger reflux drum holdups (10-30 minutes in terms of total condensate rate) are favored by some designers because they provide more liquid-surge capacity.3. There is some disagreement about optimum reflux drum holdup. us suppose Let where & is reflux flow and & is top-product flow. Level Control Via Top Product (Distillate) If level goes high. In the experience of one of the authors. k . (See Chapter 9 for further discussion..25 x 4 3 This illustrates an advantage of PI controllers in averaging level-control service. Then: that & = 5 a. This enables the column to ride through larger disturbances without losing reflux flow. This problem. Then: 2 minutes 2 minutes X 0. and consequently internal liquid flows (which take some time to reestablish). rwv 2 2 minutes x 2 x 4 16 minutes For proportional-reset controllers.26. liquid flows would be lost and the time for the column to recover from this upset would be appreciably lengthened.25. however. K. indicating that the sum of the distillate and reflux flows is less than condensate rate. On the other hand. 3 THU 3 = 0. The condenser is essentially blanketed during the period it takes to vent these noncondensibles off. Small holdups of liquid are desirable fiom the standpoint of reducing time constants in the overhead composition control loop.

Here it is assumed that both the distillate. x 12 ~ 3 2 Q ~ QmfD (3.7.27.then: -[‘HIOR [‘HI - 32 x 5 = 160 (3. it may upset the process.106 Overhead S s e Arraqements ytm and where Qmp Qmfl = distillate flow-meter span.12) This indicates that override action may be extremely rapid compared with that of normal level control.11) If the flow-meter spans are in the same ratio as the average flows ( 5 : l ) .10) Therefore. if possible. this is very similar to the previous case except that the controller must have a gain of . and KOR = 4: and [‘HIOR = A AHT X 12 4 X 2 x Q .< = 2. K. level control via reflux should be avoided.and reflux-flow control loops are fast compared with the level control loop. since such designs are often plagued by a “reflux cycle” as mentioned in Section 3. . For this reason we mostly choose KOR = 2 if possible. = 0.2 for an AC reflux valve and a level above 75 percent must open the distillate valve. This is not usually desirable. Level Control Via Reflux Flow For the system of Figure 3. For level control via reflux. ft’/min = reflux flow-meter span.13) . ft3/min and K R is the override relay gain and K.7. is the subcooling factor discussed O in the next section. For gravity flow reflux. If this design cannot be avoided. the entire condenser-reflux system should be designed according to the recommendation in Section 3. . while a level below 25 percent must close it. ‘H [‘HIOR = (3. the characteristic time constant is defined a little differently: (3.25. If K.

21.. Note that reflux valves should be sized to handle the maximum rate for total reflux operation. Hollander. H. P Y ~ . "C To allow for condenser dynamics. 7. "C TR = external reflux temperature. Cbem.Eng. E g n e i g Hydradiu.132-134 (Apr. Cbem. 4. one should use a proportional reset level controller..Eng. 48-52 (Sept. 6. "Reflux Cycle i Disn tillation Columns. L. 3. pcu/lbm "C A = vapor latent heat.Referems 107 where K. 145153 (Oct.. this should be followed by a square root extractor. P Y ~ 63: . Hydrocarbon PYOC.. If level is cascaded to flow control. Wild. N. 1966. P. ISAJ. Bolles. H. C.. proportionalonly level controller (usually because available holdup is very small). The proper design is discussed in Chapter 16. Mueller.. 2. a dynamic analysis should be made to determine proper holdup and controller type. Level Control with Small Seal Pot Volume If satisfactory flow smoothing cannot be achieved with a gain 2. 1949. A. 7 H should be at least 5 minutes and [7H]0R at least 2 minutes. Cbem. = subcooling constant. Buckley. 1978).. If an orifice AI' transmitter is used. 1967).Eng." presented at IFAC Conference. = 1 + A ( To ' lbm internal reflux flow lbm external reflux flow- - TR) (3.. level control via reflux has the advantage that external reflux temperature changes do not change internal reflux. Level Control Via Boilup For overhead level control via boilup. . W.. 185-187 (May 1957). 1969). Chin. niern Wiley. For columns with simple controls. REFERENCES 1. L.. 5. Rouse. London. the flow transmitter should have a linear output with flow..14) where cp = reflux specific heat. New York. T. S.G. pcu/lbm To = vapor temperature. 8 (July 1974). .

Shunta.. L. J. “Modelling and SimControl. P. INTECH. Fehervari. Buckley.. P. 12.” INTECH. Jan.” INTECH. New York. P. ccNontinear Control of Liquid 10. Apr. Technzques of Pmce~s 11. Buckley. June 1975. Tyreus. P. and W. 1976. 1969. Quantitative Des@ o Condensers. 43“Designing Long-Line Pneumatic 48. San Francisco... Wdey. part 1.108 Overhead System Arranpnem 8. S. 1964. Luyben. ulation of Vertical Subcooling f 9. Apr.” Automatic Control Pneumatic Control Loops. 61-66. B. and W. S. . Council. Control Systems. S. 1974. 1983.. D. part 2. 33-40. 39-42. Buckley. Level.

There should be a minimum liquid level depth above the nozzle to the drawoff line from the column base or vaporizer separator. Spacing between the vapor-return nozzle and the lowest tray should be large enough to minimize entrainment of liquid drops in the rising vapor. The maximum liquid level should not be too close to the vapor nozzle as this will promote turbulence in the liquid surface and liquid entrainment into the rising vapor.4 t ~ Column-Base and Reboiler Arrangements 4. 3. This is required for two reasons: first. forced-circulation. It requires simultaneous consideration of fluid mechanics. This difficulty. and second.1 INTRODUCTION he design of a column base with its associated reboiler can be a comdex ” matter. or kettle-type reboiler is used: l. Vortexing (like the swirl at a bathtub or slnk drain) is undesirable since it promotes entrainment of vapor into the drawoff line. heat transfer. Minimum liquid level should not be too far below the vapor nozzle. This is typically one vapor nozzle diameter and normally will be specified by the column designer. or the falling liquid drops will entrain too much vapor into the liquid pool. For example. 4. Vortex breakers should be installed routinely. 109 . the following must be considered when a vertical-thermosyphon. and process control.’ 5. they should have an orientation no more than 90” from the vaporreturn nozzle and should not be located under the last downcomer. however. mass transfer. can be largely offset by item 4. In severe cases this may cause foaming in the column base and “gassing” of the reboiler. to minimize the likelihood of vortex formation at the drawoff nozzle. to permit entrained vapor bubbles to rise and separate from the liquid pool. If level-measurement nozzles are not protected by an internd damping chamber. which may cause “gassing” of the reboiler or bottom product pump or circulating pump. 2.

the following calibration procedure is recommended. -The transmitter output span should be 3-15 psig for pneumatics. 4. heat-transfer f u will be a maximum when column-base liquid level relative to the bottom tube sheet is about one third of the distance between the two tube sheets2-see Figure 4. on a volumetric basis. or 420 mA for electronics. for 44 . The top nozzle center l should be at least 1 inch below the lip of the vapor-return nozzle. flooded-bundle kettle. This allows 6 inches below the top level nozzle to accommodate changes in liquid-specific gravity. if shell-side steam pressure.1A. This means that only about 5-25 percent by weight of reboiler effluent is vapor. A level nozzle spacing of 44 inches is suggested for A' transmitters and 48 inches for displacer transmitters. -Zero level should be 6 inches above the KRL. In view of this. it is a conservative practice to maintain column-base liquid level below the top tube sheet and above an elevation corresponding to the midpoint between the two tube sheets.2 VERTICAL THERMOSYPHON REBOILERS* lsrpical Operation and Design Vertical-thermosyphon reboilers or calandrias are commonly designed for a mass circulation ratio of about 5-20 parts of liquid to one part of vapor. however. For most applications. and lx composition are fixed. horizontal-thermosvphon reboilers are sometimes encountered.? The tubes are filled with froth. For many applications the practical implementation of the preceding is as illustrated by Figure 4. heat transfer declines only slightly. at a liquid lx elevation corresponding to the top tube sheet. t A literature review of thermosyphon reboilers is given by McKec4 . tube-side pressure. If liquid level goes below this point. * Particularly h the petroleum industry. forced circulation.3 the above may be influenced by column pressure. For either type of transmitter. and pure liquid exists in the tubes only for a few inches above the bottom tube sheet. or 1 inch below the elevation of the bottom lip of the internal vapor downspout (if used). As pointed out by Smith. The bottom nozzle should be centered 6 inches above the KRL (knuckle radius line) if a AP transmitter is used.6 = 38 inches of process fluid. As liquid level goes above this point.110 Column-Base and Reboder Arranpnents In this chapter we consider primarily five types ofreboilers:vertical thermosyphon. f u is perhaps 10-15 percent less than maximum. and internal reboilers. heat transfer falls off rapidly. flooded thermosyphon (steam side).1B. the numbers are almost reversed. and 2 inches above the KRL if a displacer instrument is used.

4.% DISTANCE BETWEEN TUBE SHEETS FIGURE 4.1A Vertical thennosyphon-heat flux vs.2 Vertical Themtoryphon Reboilen 111 0 25 50 75 100 SUPPLY SIDE LEVEL. supply side liquid level FIGURE 4.1B Distillation column base with thennosyphon reboiler .

In one set of experiment^. B. Corrective measures include reducing the vapor exit restrictions and increasing the liquid supply restriction. an abrupt increase in steam flow causes liquid to be displaced back into the column base. Manometer theory says that damping is increased as the liquid flow restriction is increased. 1 = 15 feet. Change in Tube Vapor Volume with Change in Steam Rate For a fixed liquid level in the column base. a reboiler and column base function crudely as a U-tube manometer whose natural period in seconds is approximately d / 1 . the average level will be 6 + 19 = 25 inches above the KRL. The program calculates the weight fi-action vapor at the end of each of a number of sections of tube length. Level-measurement techniques. There may be a connection between this and choked-flow instability. The cubic feet of vapor in each section of tube are then calculated fi-om the following equation: . are discussed in Chapter 11. with a resulting temporary increase in column base level. Figure 4.^ a period of 3-4 seconds was observed. Reboiler-Column-Base “Manometer” From a hydraulics standpoint. The end result is similar to boiler swell (Chapter 10. Choked-Flow Instability When the reboiler heat flux pcu/hr ft? or Btu/hr ft? becomes too high. Control and Operation Difficulties A typical thermosyphon reboiler has several additional features that sometimes cause or contribute to control difficulties: A. including baffle design. for a period of 3 seconds.2 is derived by running the reboiler computer design program for at least three values of base level and five values of heat load. C. this effect is very pronounced at low heat loads but tapers off at higher loads. the volumetric percent vapor in the tubes increases as steam (or other heating medium) flow rate is increased.2. 3 where 1 is the length in feet of the liquid column. reference 6). it is not known how general this is. there ensues a violent oscillation of process flow through the reboiler tubes and of reboiler AP.1l 2 Column-Base and Reboiler Arrargmnts Since we normally design the controls to work in the middle of the level transmitter span. As shown by Figure 4. which is fairly typical for industrial installations.

liquid density. this effect is small compared with that due to changing steam rate.2 Vertical Themtosyphmz Rebollen [V. tube vapor volume decreases with increase in liquid level. fi3/fi V: = The total vapor volume in the tubes is then E[V.]. x n where n is the number of tubes and E[V.]. heat load.4. FT3 I HEATLOA0. Change in Tube Vapor Volume With Change in Column-Base Liquid Level If heat load is held constant but column-base liquid level is varied. and base liquid level .2 Relationship between vapor volume in tubes of thermosyphon reboiler.F w r f + Fw* 113 x L x v: vapor density. in one tube. of tube section inside volume of one tube section. fi3. in feet.PCUIHR FIGURE 4. Ln cases examined so far. = total vapor volume. D.-l = + [FmrIz 2 [FmII = weight fraction vapor at exit of tube section i L = length.= where pv = * Pv 1 . VAPOR VOLUME IN TUBES.]. lbm/ft3 pL = F. lbm/fi3 [Fm].

High Concentration o Nonvolatile Components f Here the reboiler is fed with a wide boiling range mixture. which may cause either fl boiling or choked flow. a large amount of material with a much higher b % or exceeding maximum steam temperature. low boilers are completely stripped out in the tubes. Fortunately new methods developed by the ISA (see Chapter 11) permit much more accurate prediction of flashing and cavitation. . If.114 Column-Base and Reboilw Arrangements E. as sometimes happens. control may be very sluggish at higher loads. A steam-supply pressure regulator is often used to protect against this. the reboiler should be designed to operate over its normal range in one flow regime or the other. and boiling stops. condensate subcoolmg is small and there is some cavitation in or flashing across the condensate valve. For the case where the column-base liquid contains only a small amount of low boilers and o p i t perhaps approachmg on. a flooded thermosyphon reboiler operates by throttling condensate rather than steam flow. For noncritical flow the steam flow or flow-ratio controller can have much higher gain (often by a factor of 5-10) than when steam flow is critical. The increase in level and low boilers then permits boilup to resume. one may maintain constant head on it by the overflow and external tank design of Figure 4. however. the control loop may be unstable at a slightly lower load. The basic principle is that of varying the heat-transfer surface. thermosyphon action and heat transfer may fill off as liquid level in the column base f d s much below the elevation of the top tube sheet. if the controller is tuned at a lower load. The column contents then drop into the column base. and many valve manufacturers now can provide anticavitation trim. the switching point from critical to noncritical flow occurs at a heat load close to normal operatug heat load. thermosyphon action ceases.3 FLOODED THERMOSYPHON (STEAM-SIDE) REBOILERS As shown by Figure 4. it is questionable whether a thermosyphon reboiler should be used for such applications. Care must be taken to avoid im excessive reboiler AT. A forced-circulation reboiler would be better. If. controller mung can present perplexing problems. 4. a thermosyphon is used. Critical Versus Noncritical Steam F o lw When the steam valve pressure drop is high enough that flow is critical. What often happens is that either base level or low boiler concentration gets too low. At a high heat load such that shell-side liquid level is low. Conversely.3. therefore. The reboiler has process dynamics different fiom those obtained with noncritical steam flow. In view of what is often a rather narrow range of acceptable operating conditions. F. The alternate cessation and resumption are violent enough to merit the term burping.17. a condensate valve is much smaller than a vapor supply valve. If the controller is properly tuned for a higher load. reboiler shell pressure has no effect on steam flow rate. The chief advantage is that for a given steam flow in pounds per hour. Preferably.

performance tests. startup. The theory and design equations are presented in Chapter 15. overrides. For troubleshooting. sometimes by a condensate-flow measurement (more accurate and cheaper).4. and.3 Flooded Tbemmypbon (Steam-Side) Reboilers 115 Control of the condensate valve is sometimes achieved by a steam-flow measurement. No trap is required. The height FIGURE 4. Steam to the reboiler is throttled in a conventional fashion-usually flow controlled or flow-ratio controlled-but there is neither a trap nor a condensate pot.4. Instead condensate is removed through a loop seal whose top is vented to atmosphere. Response is more sluggish than when steam flow is throttled. a liquid-level transmitter should be installed on the reboiler shell side. A special version of a flooded reboiler designed for low-boiling materials (requiring low-temperature steam) is shown in Figure 4.3 Flooded reboiler . in some cases. and sometimes by another process variable.

a restriction downstream of the reboiler is sized to prevent vaporization in the reboiler tubes. Ibf/fi? absolute condensate density.5. 4. To minimize pressure drop. This arrangement has the equation: H L P BL = H s p B L + Ps + AI'. FIGURE 4. This restriction may be in the vapor line to the column or right at the vapor nozzle outlet. but also occasionally are vertical.4 FORCED-CIRCULATION REBOILERS Forced-circulation reboilers are commonly horizontal as shown in Figure 4. feet = condensate level in = = the shell. Ibm/fi3 Now the absolute value of Ps does not change very much so boilup rate is mostly controlled by variation in exposed tube area.. feet Ps p shell pressure.ne B C B C (4-1) where HL Hs = loop-seal or standpipe height.116 Cohmn-Base and Reboiler Awanpnents of the loop is typically 5-10 feet.4 Flooded reboiler for low boiling point materials .

5 FLOODED-BUNDLE KETTLE REBOILERS Kettle-type reboilers are sometimes used with vacuum columns. The usual arrangement is that of the flooded-bundle type FIGURE 4. They eliminate the need for the circulating pump required with a forced-circulation reboiler. for applications where a thermosyphon would be expected to foul. and for applications where thermally sensitive materials are being distilled.4.5 Fhkd-Bundle Kettle Rebuih 117 Forced-circulation reboilers are widely used for vacuum columns because of their lower pressure drop.5 Column base with forced-circulation reboiler . and avoid the temperature elevation encountered in the lower end of a thermosyphon reboiler. 4.

6 Kettletype reboiler wlth internal weir . which provides a high signal the first time the reboiler is filled with liquid. One has a latch-up circuit with a gain 25 relay. a head measurement must be made on the tube bundle chamber.6. Once latched up this circuit does not function again until it is unlatched by switching to “shutdown.” The other circuit is intended to shut the steam valve if the total head drops below a certain amount &er normal operation has been in effect. The column base runs empty and there is just enough liquid head in the line connecting the column base to the reboiler to overcome the pressure difference between the kettle reboiler and the column base. as shown on Figure 4. or. the steam valve is held closed u t l the level ni covers the tubes. This can be connected to an interlock. This is very similar to an internal reboiler with an isolating baffle or chamber. as will be discussed in the next section. This requires FIGURE 4.7. to suitable overrides. Shown here are two overrides (pneumatic devices illustrated).118 Column-Base and Reb& Arrangemats shown on Figure 4. To protect the tube bundle in the event that boilup temporarily exceeds downflow.

cylindrical bundles of U-tubes with the heating medium in the tubes. 4. pumping and pressure-drop considerations are eliminated. is the fact that the column base must be longer (taller) to provide FROM STARTUPSHUTDOWN SWITCH FIGURE 4.4.6 Internal Reb& 119 field calibration since the presence of froth between the tubes produces an average density well below that of clear liquid. Offsetting these in part. In comparison with external reboilers.7 Protectbe circuits for tube bundle chamber in kettle-type reboiler .6 INTERNAL REBOILERS Internal reboilers. however. as are piping and reboiler shell. usually in the form of horizontal. have been used to a modest extent.

Note that L1 and L3 are the nozzles used for level control (see discussion in Chapter 11). 2. the column must be shut down to clean the tubes. sometimes it is a two-pass bundle extending through the column with the intermediate head on the far side. the top 24 inches it between Lz and L3 are clearance. Tube bundle with isolating chamber.3) v = vapor velocity. If. for example. this is no trivial p r ~ b l e m . Ibm/ft3 (4-2) The Hepp7 correlation may be used to estimate 4: 4 where = 1 . Thus this type of internal reboiler probably should not be used in vacuum towers since the required base froth volume would be too great. Some bubbles. The latter design is better than the former for flooded-rubeoperation.8 is suggested. we need twice the volume that would be required for clear liquid.62 Y GV (4.Here the tube bundle* is submerged in a pool of liquid. Tube Bundle Without lsolating Baffle or Chamber This case is illustrated in Figure 4. To allow for variations in f o h density. Although liquid level can come a little way below the top of the tube bundle without causing either fouling or a reduction in heat transfer. A indicated by the s literature. not clear liquid. 4 = 0. For the particular case where C#I = 0. Since most level instruments really measure head. Two basic arrangements for installing and operating internal reboilers are employed: 1. there is the practical problem of measuring or predicting true level accurately enough to prevent the froth from rising high enough to flood the lower tray or trays. of controlling the level of a variable-density f o h In addition. lbm/ft3 To get adequate holdup for level control. of course. exist in the spaces between the tubes. ~ From the column or reboiler designer it is necessary to get an estimate of relative froth density: = Froth density. * Sometimes the bundle consists of U-tubes.120 Column-Base and Reb& Arrangements adequate space for controlling the fioth level.5. not level. the nozzle spacing of Figure 4. Above the bundle we normally find fioth. it is more conservative to keep the bundle submerged.5. Tube bundle without isolating baffle or chamber.L3 and Lz are used to determine that the bundle is covered by liquid.0. lbm/ft3 Clear liquid density. ft/sec pv = vapor density. There is also the added difficulty it.8. we must take 4 into account. the level transmitter is calibrated such that “100 percent” level is 76 inches above L2. .

6 Internal Reboilers 121 FIGURE 4.4.8 Column base with internal reboiler-no baffles o weirs r .

10. a low level should pinch this valve. and valve. In the event that boilup exceeds downflow. There are now two problems: (1) controlling level in the column. reduce reboiler heat-transfer capacity. connected to nozzles L3 and L4. In some cases there is also a desuperheater. This drops pressure in the reduced-pressure header. but is often a self-contained. If all of the controllers are tuned tightly.” The downcomer fiom the last tray dumps into the chamber and the excess overflows into the column base. lower. as sometimes believed. This last item is sometimes installed because of a common misunderstanding about the effect of superheat on reboilers.11 may . and ( 2 )protecting the tube bundle if boil-off exceeds downflow. Thus it is recommended that the level be measured and controlled below the isolating chamber. 4. there is often a severe interaction between the loads and the pressure reducer. From a high-pressure header. there may ensue “fighting” among the controls and substantial swings in header pressure and steam flows to the various reboilers.” This may feature a transmitter. In theory it is possible to run base level at elevations above the bottom of the isolating chamber (in the space between L2 and L4 in Figure 4. A system such as shown in Figure 4. or “bathtub. but in the meantime the lower pressure may have caused another reboiler supply valve to open.The output of this transmitter should be connected to an override in the heating medium control valve circuit. self-actuated pressure regulator. one should provide a second level transmitter. it may be more logical to eliminate the pressure reduction. For a header with many loads. that steam flow to one reboiler is increased.9). are limited. it simply reduces slightly the amount of condensate’ and usually does not. Suppose. and therefore holdup at this elevation. With many organics. steam is let down to a lower pressure through a “reducing station. In practice it is undesirable because crosssectional area. this causes a type of fouling To called “varni~h.7 STEAM SUPPLY AND CONDENSATE REMOVAL Steam supplies for distillation columns are commonly arranged as shown in Figure 4. The outlet weir from the chamber is generally set high enough to flood the tube bundle. Some columns have been built this way. header pressure is really constant at only one point. controller. chamber. the level in the isolating chamber drops and exposes tubes. at the far end of a distillation t a n header pressure may be much ri. for example.122 Column-Base and Reb& Arranpnmts Tube Bundle with Isolating Chamber As shown by Figure 4. The pressure controller opens the upstream valve to restore pressure.9. internal reboilers are sometimes isolated fiom the column base by an isolating b d e .~’ prevent this from happening. Particularly for a small number of loads. The reduced-pressure header may serve one load or many. For a design with only two or three loads.

9 Column base with isolated internal reboiler .4.7 Steam Supply and Condenrate Remmal 123 FIGURE 4.

124 Column-Base and Reboiler A r r a y w n t s be used instead. A suitable scheme for temperature and pressure compensation is gven in Chapter 11. Here each reboiler is supplied fiom a valve connected directly to the high-pressure header. FIGURE 4. and less probability that flow regimes will switch during normal operation (see Section 4.2). This figure also illustrates temperature and pressure compensation of the steam flow meter. This saves one valve and controller and the higher pressure drop means smaller valves. Errors in steam-flow metering without such compensation are also discussed. There is more likelihood of critical pressure drop. Header pressure and temperature variations severely impair the accuracy of many uncompensated plant steam flow meters.10 Steam header configuration .

some reboilers. This means that the installed valve flow FIGURE 4. temperature compensation is sometimes omitted since temperature changes more slowly than pressure. in effect. Two practical problems are associated with the selection of condensate removal valves: (1) in sizing calculations allowance usually must be made for some flashing. These pots then have conventional level controllers. and (2) the pressure drop increases with flow rate since reboiler shell pressure increases with load. As a result. Many plants have experienced the need for high maintenance with these. as pointed out by Mathur: a rapid closure of the steam valve may cause a vacuum in the shell and p d back condensate through the trap with possible hammering and vibration in the shell.4.7 Steam S u . are now equipped with condensate seal pots. are very simple level controllers. Sometimes enough steam leaks through to impair the accuracy of steam-consumption estimates based on steam flow metering. In addition. Steam condensate removal is most commonly achieved by traps-devices that.11 Improved steam supply and flow control system . particularly large ones. 4 and Condenrate Removal 125 When header pressure variations are small.

12. while a third pinches bottom product flow if column base composition has too high a concentration of low boilers. Base-Level Control via Bottom-Product Withdrawal In addition to the auto overrides of Figure 4. if at al possible. Anticavitation trim may be required. auto overrides to keep the liquid level between the level transmitter taps. if bottom-product flow goes to a large storage tank). however. Avoid scheme 2. If flow smoothing is noncritical (as. If possible. As mentioned earlier. control of the column. as indicated by Figure 4. intermediate buffer tanks to feed another process step. or (3) feed flow. This is discussed in Chapter 5. f? i AH. (2) steam or other heating medium flow. 4.” mentioned earlier. In the discussion that follows. or at least adequate. the basic criterion for choosing column-base holdup will be good. But if the column base serves as a feed vessel for another step. if the bottomproduct flow is very small. One pinches the bottom product flow if a downstream level gets too high. A second pinches the feed flow if column base level becomes too high. because of reboiler and columnl base “swell. feet K* = level controller gain QmfB bottom-product flow-transmitter span. good composition control is favored by small holdups. It is sometimes required. we recommend a PI level controller cascaded to flow control. we will avoid the use of separate. The PI level controller should have. The characteristic time constant for this system is: where A B = column-base cross-sectional area. Quantitative design is discussed in Chapter 16.13. for example. its required volume may be influenced by downstream requirements.12.8 REQUIRED HOLDUP FOR LEVEL CONTROL Base level is most commonly controlled by manipulating one of three variables: (1) bottom-product withdrawal. For maximum flow smoothing. = level transmitter span. we commonly have three more overrides.126 Column-Base and Reboiler ArranpnmB characteristic should be between linear and square root. it will be satisfactory to use a proportional-only level controller as shown in Figure 4. fi3/min = For this system it is usually satisfactory to make TH 2 10 minutes. .

8 Required Holdup fbr Level C0nh.4. Q c E Q C cn e L m s m a 2 3 d E 5 E! 5 2 -0 e. g E fi * m Q n m I E - ¶ 8 rc 0 2z -is woo a3 Q E3 .01 127 c .

f ? . averaging level control system may be defined by a first-order time constant: TH = A AHT B K. where TH A B is in minutes = cross-sectional area.4a) de. x 12 x 42 o (4.13 Proportional-only level control system for column base .128 Coiumn-3me and Re&& Arrayemmk The dynamic response of the proportional-only. of the column base or vaporizer separator AHT = level transmitter span. in feet. of process fluid FIGURE 4.

8 Required Holdup jiw Level Control 129 K. = controller gain fi3/min = manipulated flow. 6 . Base-Level Control via Steam-Flow Manipulation In this case rHshould be 20 minutes or more. we now lean to using a factor of 4 for design calculations early in a project. one should make the volume AB X AHT at least ten times the volume inside the reboiler tubes. that is.4. 5 3 Q ~ 7~ s This is the required column-base cross-sectional area. Further. maximum steam flow . and square-root extractor). Then if the valve positioner input span is 12 psi: A B 2 x 4 x AFS 38/12 = 7H (4-5) 2 . however. Earlier it was suggested (Section 4.= input span of steam flow transmitter. = signal to valve positioner. A mathematical analysis is given in Chapter 16. For this design: where ( w ~ ) . psig For a linear installed flow characteristic. the discussion in Chapter 11. As a result of some as adverse experiences due to undersizing holdups by using a factor of 2 to multiply to get @ J m M . pounds per minute. In addition.1) that a generally useful level transmitter span would be 38 inches of process fluid. @o)max ~ Q F S QFs = flow sheet or average flow. for a thermosyphon reboiler. Base-Level Control via Feed-Flow Manipulation Here rH should be 20 minutes or more and base-level control should be cascaded to feed-flow control. fi3/min There is some uncertainty in valve gain because of Merences in the valve- sizing philosophy followed by various instrument engineers. the level controller should be cascaded to a steam-flow controller with a linear flow transmitter (or orifice AP uansmitter. See.

) If this arrangement is used.gemenB = = A .16.. If a thermosyphon reboiler is used. It therefore is recommended that all such applications be subjected to a dynamic analysis such as presented in Chapter 16. care should be taken to ensure that the lower level nozzle is not below the midpoint of the tube length. (See Figure 4. An occasionally encountered design for this purpose is illustrated in Figure 4.17. pcu/lb = process fluid density.’* and it is our understanding that there are various proprietary designs that have never been publicized. which extends h o s t to the base. As can be seen. lbm/ft3 PP Note that the term in brackets (Wst>m S PP AP is the process vapor flow corresponding to (w~). Much difficulty has been encountered when trying to control base level by adjusting steam flow to a thermosyphon reboiler.A typical control system arrangement for level control via steam-flow control is shown in Figure 4. pcu/lb process fluid latent heat of condensation. it may be necessary to send the bottom product to a cooler and surge tank to get adequate flow smoothing for the next process step. it is necessary to keep the amount of hot-liquid storage to a minimum. When this valve is pinched. The preferred design of Figure 4. the thermosyphon reboiler suction is taken fiom behind the outlet weir for the last downcomer. If a cooler is used. This means that reboiler feed is richer in low boilers than is the column-base contents.9 MISCELLANEOUS COLUMN-BASE DESIGNS A large number of column-base designs have appeared in the literature.” Level nozzle spacing should be chosen as before for the same reasons. This may or may not be undesirable. Minimum Holdup Design For those distillations involving thermally degradable materials. the bypass valve opens to send material to the column base. .15 involves maintaining liquid level in a small-diameter pot or “peanut. it should be noted that the surge tank may act as a condenser via the equalizing line. Many of these include baffle arrangements intended to make the column base and reboiler function as an efficient stage of separation.130 Column-Base and Reboiler Awa?. the bottom product valve may be used with an override to maintain a low level in the column base.14. 4. hP steam latent heat of condensation. Note that for overall process control.

d m a 3 $ G g E m s 2 d 8 w L 0 q w s 8 g - C 8 gg gz .4.9 Mhcehneous Column-Base Desgns 131 c . E W m E L C m U W 9 m J i 0 *I .

15 Column base design and arrangement for minimum holdup .132 Column-Base and Reboiler Awangements FIGURE 4.

an external surge tank must be used. FlGURE 4.17.1 0 MISCELLANEOUS REBOlLER DESIGNS Two other types of reboilers are sometimes encountered: (1) reboilers with hot oil as a heating medium.10 M U c e h m Reboiler Des$ns 133 When the column base is too small for smooth level control. The bottom product receiver. 4. should have enough holdup to contain the entire column contents.4. Both are more common in the petroleum industry than in chemical plants. as a minimum. A recycle line back to the column base (or to a feed tank) facilitates startup and can be used to protect the column base-and therefore also the reboilerfrom excessively low level. and (2) direct-fired reboilers. One way of connecting the two is to overflow from the column base into the bottom product receiver as shown in Figure 4.16 Reboiler piping arrangement for preferential boiling of reflux from lowest downcomer .

134 Column-Bare and Reboiler Arranpnens I 3 V I $! Q F i 2 + m Q 3 ' ' S * c Q E 0 F $! 0 30 n 3 m z - s 2 L 4 2 T Q bE .s E = sU g L .

J.. Mathur. 5 . 1973).. For higher temperatures than available with steam. F. K. 146 (July 31. H. Hepp.. Chem. Frank. 1961.. S.Rejb-ems 135 Hot-oil reboilers may be like steam-heated reboilers except that there is no phase change on the oil side. R. 4. Prickett. Computation and Control. Wash. 8. Chem.. 3. 118-120 (Aug. Oil GasJ. Shellene. . V. Hydr. C.. Snyder. 107-110 (Sept. Chem. 70 (July 1974). 189196 (July 1961). Chem. 1969). 1970). J. direct-fired reboilers may be used.. Seattle. 1967. 1967). 2. 4. use is made of other condensing media such as Dowtherm or p-cymene. For even higher temperatures. 9.. 1963). Vol. E. 40: Jacobs. J. 3. and D. K. Chem. Aug. Best practice is to estimate boilup wBU from the equation: This requires measuring oil-flow rate and inlet and outlet temperatures. Stemphg. 1973). and C. 59: 66-69 (Feb.IEC. Eng. 101-106 (Sept. New York.. Handbook o Autof &ion. PYOC. F. and R.. 6. D. Prog. 687. P. O. Patterson. J.. Dobratz.. Eng. V. Eng. M. Smith. Grabbe... Oldershaw. R. 3. Eng. at Ninth National Heat Transfer Conference AIChE-ASME. Eng. McKee. “ExDerhentd Studv of a Vertkal Thekmsyphon Rebkler“ presented 10. H.. M. N.. Church. 3. Prog. 62: 76-82 (Dec. M. REFERENCES 1. C...

and (3) there is good mixing in the tank. as shown by Figure 5. 5. When the upstream pressure is higher than the column pressure. we will assume that (1) process material-balance control is in the direction of flow.2A or by adjustment of a variablespeed drive of a pump with a fixed stroke as shown by Figure 5. For small feed rates. it provides the most protection and offers the operator the maximum flexibility. however. This (Figure 5. it should also send the feed to the proper feed tray. For minimum energy consumption operation. And for startup/shutdown of the column being fed.1A.1 shows three commonly encountered feed flow schemes. The flow signal should be linear-one should use a linear flowmeter or a square-root extractor with an orifice and AI' transmitter.1B.5 t Feed System Arrangements 5. then a pump is required. then only a letdown valve is required. (2) feed-tank level control is of the averaging type.1C is required.2B. feed composition. Some care should be taken in locating the feed control valve if flashing can occur. If column pressure is greater than upstream pressure.1 GENERAL COMMENTS he feed system for a column should h c t i o n as a filter for incoming disturbances in feed flow rate. it may serve to receive recycled column product streams. then a cascade level control/liquid flow control system such as that of Figure 5.lC) is really the preferred overall design. If the minimum stroke rate is at least three times the reciprocal of the feed-tray holdup time.2 FEED FLOW CONTROL Figure 5. upstream or downstream pressures can vary signdicantly. and sometimes feed enthalpy. Hydraulic problems have been experienced in columns where the feed 137 . In the following discussion. Control may be by stroke adjustment on a constant-speed pump as shown by Figure 5. If. it may be considerably cheaper to use a positivedisplacement pump (piston type). as shown by Figure 5. no pulsation damper is required.

1 Feed system for distillation column .138 Feed System Arranpmnts FIGURE S.

5.2 Column feed systems with positive-dispIacement pumps .2 Feed F h Control 139 FIGURE 5.

One source of heat disturbances is the column feed. Slugs of liquid can be dumped intermittently onto the feed tray as vapor slugs push up through the feed line. particularly if a heat economizer such as that shown in Figure 5. The objective-to minimize heat consumption-is a worthy one. where feed flashing can occur.3 FEED TEMPERATURE CONTROL Fluctuations or variations in the column heat balance can be a major factor in interfering with good composition control. Shown in Figure 5.3 Column feed preheat via exchange with battom product . 5.3 is used. one can achieve temperamre control.4 is such a system with a temperature control scheme FIGURE 5. but h s system will not.140 Fced System Arvangemerrts valve is located at a lower elevation than the feed tray. the feed valve should be located as close as possible to the column inlet feed nozzle. provide constant feed temperature or enthalpy. Accordingly this arrangement is occasionally modified by the addition of a heater. without a bypass. Therefore. With a bypass.

2 Since FIGURE 5. Overhead vapor from the column. and other process streams are sometimes used to minimize energy consumption. or if it is partially vaporized. some of which are quite complex. sidestreams (liquid and.5. particularly. 5.4 Column feed temperature control with economizer and preheater .4 FEED ENTHALPY CONTROL If the feed is to enter at its bubble point. Column feed is sometimes preheated in other heat-exchanger systems. This fixes the feed enthalpy if the feed is not partially vaporized (see the next section for a discussion of feed enthalpy control).5.’.4 Feed Enthalpy Control 141 that holds feed temperature constant by thmttling steam flow (hot oil is sometimes used). then we need an enthalpy control scheme such as that shown in Figure 5. vapor).

it must be calculated as follows: JrF where = cpf = TB. lbm/min feed rate. = TB.P . of bottom product leaving economizer feed iiil temperature.142 Feed System Arrangements enthalpy cannot be measured directly.TB2) fps + WF I!~ . degrees Kelvin ( = "C + 273) nta bottom-product flow rate.5 Column feed enthalpy control with economizer and preheater . degrees Celsius. pcu/lbm "C temperature.t f T1 WB + -cpb wF (TBl. = TI = wB = wF = . degrees Celsius. pcu/lbm feed specific heat.J (5. lbm/min FIGURE 5.1) JrF enthalpy of feed entering column. of bottom product entering economizer temperature.

Many columns have more than one feed. since the base of the column is at the highest temperature in the column. For a given feed composition and enthalpy.5 Feed Tray Lo& = bottom-product specific heat. If air or cooling water is used in the condenser. pcu/lbm The manipulation of feed enthalpy can be very important for some columns in utility optimization schemes and in maximizing column capacity or separation. Sometimes unexpected results occur. a column should generally be equipped with a number of alternative feed trays to handle changes in operating conditions from those assumed for design. 143 pcu "C ws JlS = = lbm steam flow rate to steam preheater. This increases the heat load on the condenser. less expensive heat source than must be used in the reboiler. it is often economical to increase feed preheat. causing the column to f o d As shown in Figure 5. Intermediate reboilers in the stripping section of the column are often used for the same reason. Luyben5 has shown that the optimum feed-tray location in some columns rises higher in the column as the feed becomes lighter (increase in more volatile component concentration). there is an optimum feed-tray location that permits making the specified separation with the least energy consumption. it is usually economical to minimize feed preheat. Feed preheat should be minimized when the column is condenser limited or is limited by flooding above the feed tray. The effect of feed preheat on column capacity can be significant. These feeds should not be mixed together and introduced onto the same feed tray if their compositions differ. It is also the tray that will permit maximum feed rate without lo. 5 5 FEED TRAY LOCATION . Adding more heat to the feed reduces the amount of heat that must be added to the reboiler (but not in a one-to-one ratio) and increases the reflux ratio required to make the same separation. but in other columns the effects can be small. It is often possible to transfer heat into the feed stream using a lower temperature. If refrigeration is required in the condenser.5. lbm/min difference in enthalpy of steam entering preheater and condensate leaving it. The magnitude of the energy savings to be realized by changing feed tray location can be very significant in some systems (10-20 percent reduction in heat input). Feed preheat should be maximized when the column is reboiler limited or when it is limited by flooding below the feed tray. . Each system must be examined to determine the strategy and the incentives for controlling to an optimum feed-tray location. while in other columns exactly the opposite is true.6. Each feed stream should be introduced onto its optimum feed tray if energy consumption is to be minimized.

F h v Smoothing. fi3/min. FIGURE 5.6 FEED TANK SIZING There are a number of possible criteria for tank sizing.144 Feed System Arrawemenk S.6 Column with multiple feed trays . They include the following. Here P = average vessel holdup. and& = average feed rate. Mking. The level control time constant rH should also be large compared with the reciprocal of the closed-loop natural frequencies of the downstream column-composition control loops. The mixing time constant of a well-mixed tank TM = QF : V should be large compared with the reciprocal of the closed-loop natural frequencies of the downstream column-composition control loops. fi3.

If this cannot be achieved with the proposed vendor's hardware. On the other hand. As a rough rule of thumb. If the engineer is confionted with one of the following situations. Column to be fed is involved in a heat-recovery scheme where the vapor from one column furnishes part or all of the reboil heat for another column or columns. Column to be fed has a sidestream drawoff that is a major fraction of the feed.1% (mole or weight) impurities or less. 5. the use of intermediate tanks with supports. one should consider using a dedicated small computer as a level controller. If it is decided that a separate surge tank is needed. control of the succeeding process steps will be easier if the feed holdup is large. proportional-only control is best since output flow changes are never larger than input flow changes. (very close to flooding) for a substantial fraction of on-stream time. The surge tank should also have good mixing.7 to 5. therefore. Then A M .7 Feed S p e m fm Double-Column System 145 The closed-loop natural frequencies of the composition control loops are n n and rH should be at not easy to calculate early i a project.25 7 H X @ff)m (vertical cyhdricd tank assumed). = 2 rH x @ff). surge tank volume between the level taps should correspond to a 711 2 1 hour. but i general least 15-20 minutes.7 FEED SYSTEMS FOR DOUBLE-COLUMN SYSTEMS Figures 5. Tank (holdup) size is also a function of whether intermediate tanks will be used in a train or whether product will be transferred directly from one column base to another column. the recirculation rate should equal or exceed l o a s .or bottom-product quality specification of the column to be fed is 0.5. usually because of structural advantages. pump. Generally speaking. The reset time for unity damping ratio and 7H = 1hour would then be 4 hours. then a separate surge tank for column feed is recommended: 1. 4 Column to be fed is expected to operate at or close to maximum capacity . Otherwise one must go to a proportional-only level controller.7 shows a simple split column where two columns are employed instead of one tall column.25 (see Chapter 16). composition control of a column will be easier (faster) if the reflux drum and column base are kept small. Compromises. a separate surge tank should be used for column feed. and instruments may increase investment substantially. A AHH. are frequently necessary. If item 4 is a significant consideration. electric switchgear. 3. However. For a PI controller with auto overrides and Kch = 0. 2.= 0. then.. 5. This scheme . Columns in train have primitive controls (see Chapter 1). Top. Figure 5. preferably with gain 2.9 illustrate several systems where two (or more) columns are operated in series or in parallel. if required 7 H or rM is much greater than 15-20 minutes. as a starting point.

146 Feed System Arrangements E s - 3 c . P r n m 8 k $ L 2% *x ss 3 U .

5.0 E E8 .7 Feed System far Double-Column Systems 147 s $ s E s 3 U B w w B e ooE q L 3 g.

148 Feed System Arranpnents FIGURE 5.9 Feed systems for columns in parallel .

permits both feed rates to vary.8 shows a similar but distinctly different two-column system. the smaller column is normally the one that would be base loaded since swings in feed rate to the larger column are smaller on a percentage basis. Each column has its own reboiler and condenser. Vapor from the top of the second column is condensed and pumped into the base of the first column. depending on the optimum feed-tray location. 2. Al the vapor is generated in the base of the second column. since this typically operates at a lower temperature than Process Unit No. at the same time.9 illustrates feed systems commonly used for two (or more) columns operating in parallel. Most of the flow from Process Unit No. It is usually undesirable to put everything through the feed tank. . The level in Process Unit No.8 FEEDS WITH MAKEUP/PURGE TO TANKAGE Figure 5. The tank is connected in series with the main process flow. Liquid reflux for the second column is provided by the liquid from the base of the first column. The lower codguration. Feed can be fed to either column as dictated by composition. Figure 5. Note that this system uses about twice as much energy as a single column would require.9A.5. 5. Figure 5. The two valves on the makeup and purge lines from and to the feed tank are split ranged so that both valves cannot be open at the same time. If the two columns are of unequal capacity. Feed can be introduced into either column. 2.8 Feeds with Makeup/Puvge to Tankage 149 is also sometimes used when revamping an existing process to udlize existing l columns. 1 must be controlled. 1 goes directly to Process Unit No. Liquid from the base of the first column is fed onto the top tray of the second column. It is sometimes used in vacuum systems if maximum base-temperature limitations and tray pressure drop prevent the separation from being achieved in one tall column. The other column then takes the swings in feed rate.9B. The control system shown in Figure 5. A bias signal is usually needed to balance the feed rates between the columns. is used when there is some advantage in keeping the feed rate to one of the columns constant. Figure 5. The upper configuration.10A is a simple technique for minimizing energy consumption. The vapor from the top of the second column is fed into the base of the first column. This guarantees that a minimum amount of material is going to or coming from the tank. Figure 5. 1. Cooling all the process stream and then feeding it to Process Unit No.10 shows a system sometimes encountered when a lqge feed tank is used to balance load and demand at some point in the process. 2 significantly. The top of each column is maintained under vacuum. 2 normally increases energy consumption in process Unit No. it is desired to fx the flow rate into i Process Unit No.

150 A.10 Makeup/purge feed systems . SINGLE SYSTEM Feed System Arrarpynwnts FIGURE 5.

Eng. This problem can be eliminated by the more sophisticated and costly system shown in Figure 5. Speicher. Lupfer. Luyben. W. D. 53:963-969 (Dec. E. D. Material flows through these sequences of columns. IEC Fundamentals 10:147 (1971). should check all level controls in a sequence of process equipment.Rt$mences 151 The system has some dynamic problems. To prevent the level in Process Unit No. 1962). and small fluctuations in flow rates in upstream columns can be amplified as they work their way down through the train if two or more level controls in sequence have unenhanced PI controllers and have the same rH and rR. the valve i the direct line connecting to two Units is pinched). If necessary.End. W.10A also requires overrides: 1. J. The system in Figure 5. . and M. Although this is uncommon. 5. this problem is greatly deviated. 87-88 (Feb. that if one uses PI controllers enhanced with auto-overrides. in an unpublished study. some of the tank sizes or controller parameters may have to be changed. 2.. L.9 FEED SYSTEMS IN SEQUENCES OF COLUMNS WITH AND WITHOUT RECYCLES Trains of distillation columns are often encountered. Lupfer. and E. The level loop is affected by the flow loop. 2 from getting too low (a low level pinches the valve in the line to the feed tank). W. E. Cant. I d . A pressure controller on the "header" is used to balance supply and demand flow. Cbem. REFERENCES 1. 3.10B. therefore. 1961). it may happen.. Oglesby. It has been shown by one of the authors. 2. 2 (if the n flow gets too high. To prevent too much material from entering Process Unit No. Oglesby. and M. The engineer..

provided the proper installed valve flow characteristic is used. first to lay out all of the materialbalance controls. either technique is satisfactory. as discussed in Chapters 12 and 20. We will begin with combinations of level control and feedforward compensation for applications where material-balance control is in the direction opposite to flow.1 INTRODUCTION s indicated in Chapter 1. It has been shown 1. Columns with sidestream drawoffs are discussed in Chapter 7. Note that 153 . These are usually larger and more rapid than composition or thermal changes. Then we will consider schemes in which material-balance control is in the direction of flow.6 a Level Control and Feedforward Options 6. Unfavorable schemes-those that are hard to design or to make work-will be pointed out. it is usually advisable to use an impulse feedforward technique unless the wild flow rate changes less than 2 : 1. we will use ratio controls in most of the illustrations in this chapter. there is a serious variation of control loop gain fiom low to high production r a t a 2 Consequently. In this chapter we will consider only feedforward compensation for production rate changes. however. In the absence of feedback composition controls-usually because adequate composition measurements are lacking-feedforward compensation is almost mandatory. It was also indicated that feedforward compensation could be used to supplement feedback composition controls to achieve more constant compositions.2 that for single-loop (no cascade) control. For cascade loops. These are mostly liquid level controls. Only conventional columns with top and bottom drawoffs wdl be considered here. however. either open loop via a multiplier or closed loop via a divider. where the flow-ratio control is the secondary or slave loop. their use should be avoided unless no suitable option is available. For simplicity. it is most convenient. when starting the design of controls for a new or modernized plant. The most common compensation technique is that of ratio controls.

Base composition may be controlled by trimming the steam/bottomproduct ratio. appropriate dynamic feedforward compensators should be provided. one should take care to account for reflux subcooling (see Chapter 16. Reflux Drum Level Control Via Reflux In this case. Section 6. Overhead composition may be controlled by trimming the reflux/bottomproduct ratio. Section 2. including overrides for bottom product and steam flows.2. however. Overhead Level Control Via Top Product There has been considerable controversy in the literature about whether to have reflux drum level control via top product or via reflux flow. On the average. in most cases. Base composition may be controlled by trimming the steam/bottom-product ratio control. In calculating reflux drum level controller settings. level control must be as "tight" as possible for best composition control. Since there is a dead time involved-the time for liquid to flow from the feed tray to the column base-the base level controller settings should be determined by the method of Chapter 16. Then there will be two options for reflux drum level control. Section 4). 6. or if base level sets steam.1. Since the bottom-product flow is the demand flow.u if reflux drum level control sets reflux.154 Level Control and Feedfmard 0pth. We can now visualize several possible designs. . that base level be controlkd by feed rate.2 MATERIAL-BALANCE CONTROL IN DIRECTION OPPOSITE TO FLOW This approach implies that either the distillate or the bottom product is the demand flow and that the column must be operated at a rate to satisfjr that demand. For all ratio loops. we have steam/bottom-product and distiUate/bottom-product ratio controls. whether proportional only or PI. Our own studies f d to show any overwhelming advantages one way or the other. The reflux drum level controller settings (if a PI controller is used) should be determined by the method of Chapter 16. we find top composition control somewhat more straightforward when reflux is manipulated by a ratio controller or by a direct composition controller as shown by Figure 6. Bottom-Product Demand: Base-Level Control Via Feed Bottom-product demand requires. as shown in Figure 6. This section also gives a complete design. Overhead composition may be controlled by trimming the disdlate/bottom-product ratio with perhaps a feedforward compensator connected into the overhead level control loop. steam and reflux are ratioed to it.

overhead level control via top product. base level via feed .6 2 Ma$eriul-Balana Control a Directwn Opposite to F h n 155 DEMAND FIGURE 6.1 Bottom product demand.

base level control via feed . overhead level control via reflux.156 LeueL Contml and Feedfmard 9th FIGURE 6.2 Bottom product demand.

A buffer tank in the top-product line is recommended. Practically speaking. this column control scheme is not desirable. Preferably one should make T H 2 5 minutes. If steam flow is metered by an orifice. Since reflux drum holdups are usually small compared with base holdups. material-balance control in the direction of flow means that the column must take whatever feed is supplied. base composition may be controlled by trimming the bottom-product/distillate ratio. As indicated earlier. in this case. difficult unless a large reflux drum holdup is available. Overhead composition control may be accomplished by trimming the top-product/bottom-product ratio. either reflux drum or base level control must manipulate feed rate.5. it should be linearized with a square root extractor. bottom-product/top-product and steam/top-product ratio controls. Section 6. 6 3 MATERIAL-BALANCE CONTROL IN DIRECTION OF FLOW . Note that if either distillate or bottom product (or side product) is a demand flow. Top product composition may be controlled by trimming the steam/distillate ratio. Base level controller settings may also be determined by the method of Chapter 16. the calculation of base level controller settings follows that of Chapter 16. reflux drum level control via boilup is difficult unless a lot of holdup is available. this case requires reflux/bottom-product and distillate/bottom-product ratio controls. Because of interactions between composition controls. level controller tuning will require a dynamic analysis of overall column material balance such as discussed in Chapter 14. subject . Base composition may be controlled by trimming the reflux/bottom-product ratio. bottom composition may be controlled by trimming the bottom-product/distillate ratio. however.3 Material-Balance Contra1 in Direction af Flow 157 Reflux Drum Level Control Via Boilup As shown in Figure 6. Overhead composition may be controlled by trimming the reflu/distillate ratio. although probably not impossible. as shown by Figure 6. Section 4. Distillate (Top-Product) Demand: Base Level Control Via Feed Reflux Drum Level Control Via Reflux This case requires. a buffer tank in the top product h e (notin the reflux line) is highly recommended.6. Section 6. Determining settings for the reflux drum level controller is.3. The calculation of reflux dnun level controller settings follows the method of Chapter 16. This is. this scheme requires reflux/distillate and bottomproduct/distillate ratio controls. Reflux Drum Level Control Via Boilup As shown in Figure 6.4.

158 Level Contvol and Feea!jimard Optimu FIGURE 6. base level control via feed .3 Bottom product demand. overhead level control via boil up.

reflux drum level control via reflux.3 Matffial-Balunce Conml i Direction o F h n f 159 FIGURE 6. base level control via feed .6.4 Distillate demand.

reflux drum level control via reflux. base level control via feed .5 Distillate demand.160 Level Control and FeerSfmrttard Optimr FIGURE 6.

a square root extractor should be used. care should be taken to include the subcooled reflux effect on overhead level control. reflux drum level is controlled by manipulating top product. It is normally used only when the average bottom-product flow is very s m d . Reflux/feed and bottom-product/feed ratio controls should be provided. Reflux Drum Level Control Via Distillate. Section 3. Both level controls may be calculated by the method of Chapter 16. Base Level Control Via Boilup. and base level is controlled by manipulating boilup (see Figure 6. Top composition may be controlled by trimming the distiUate/feed ratio and with perhaps additional feedforward to the reflux drum level controller. Base composition may be controlled by trimming the steam/feed ratio. Reflux/feed and steam/feed ratio controls should be provided. can restrict feed flow. Reflux Drum Level Control Via Reflux. Reflux Drum Level Control Via Distillate. Top composition may be controlled by tnmrmng the rdux/feed ratio.^ If. Reflux Drum Level Control Via Distillate. unless the top product is going directly to storage. Reflux Ratioed to Top Product It is sometimes recommended that reflux be ratioed to distillate to diminish interactions between top and base feedback composition control^. Top composition may be controlled by trimming the refldfeed ratio while base composition may be controlled by adjusting the bottom-product/feed ratio. however.7) This is also a commonly encountered control scheme. Base Level Control Via Bottom Product (Figure 6. The characteristic time constant ‘TH should be at least 15 minutes and other design factors should be as indicated in Chapter 16. Distillate/feed and steam/feed ratio controls should be provided. particularly high base level. bottom composition by trimming the steam/feed ratio. Base Level Control Via Bottom Product (Figure 6. there will be a dynamics . Both level control systems may be calculated by the method of Chapter 16.6) This is one of the most commonly encountered column-control schemes. In most cases base level control by boilup requires a dynamic analysis. If steam flow is measured with an orif?ce. and perhaps supplementary plant tests.6.3 Matetid-BalanCe Control in Direction o Flow f 161 to the restriction that column overrides.9). but base level control by boilup is very difficult.8) Overhead level control may be calculated simply by the method of Chapter 16. Section 7. In many cases this scheme will require a buffer tank in the top-product line to the next piece of process equipment. Base Level Control Via Boilup (Figure 6.

reflux drum level control via distillate.6 Material balance control in direction of flow.162 Level Control and Feedjimard Options FIGURE 6. base level control via bottom product .

7 Material balance control in direction offlow. base level control via bottom product .6. reflux drum level control via reflux.B a l a m Contml in Direction o F h f 163 FIGURE 6.3 M a t d .

reflux drum level control via distillate.8 Material balance control in direction of flow.164 Level Control and Feedfmard OptMns FIGURE 6. base level control via boilup .

6.3 Material-Balance Control in Direaim o Flow f 165 FIGURE 6.8 but with reflux ratioed to distillate .9 Like figure 6.

6. To a considerable extent. This configuration produces a positive feedback loop that either will shut the column down or open the steam and reflux valves wide. One of the worst combinations of controls is reflux drum level via reflux and base level via boilup. that an “inverse response compensator” can be designed and implemented on a computer or with some microprocessor controls. An increase in bottom product will cause a decrease in distillate. neither of which is a desirable situation. with no change in feed rate. however. To minimize difficulties the design recommendations of Chapter 16.4 UNFAVORABLE CONTROL SCHEMES Perhaps the most common system that has given trouble is base level control via steam. Operator intervention is eventually required. Section 7. One of the authors has shown. We have seen it used on columns with very large base holdups-8 24 hours. or both. these problems can be mitigated by using proportional-reset instead of proportional-only overrides (Chapter 9). in an unpublished study. This scheme is analogous to the preceding scheme. this was corrected by converting the steaxdfeed ratio control to steamhottomproduct ratio control. sieve trays give the same trouble. Changes in the demand flow rate are not accompanied by appropriate changes in the other flows. Since bottom product was the demand flow. In effect the level controller adjusts both feed and steam.5 UNREASONABLE CONTROL SCHEMES The control schemes in this section have one common fault-the inability to control the overall column material balance. High base-level overrides on steam have also given trouble for the reasons mentioned. If the bottom product is the demand stream. 6. to . A high base-level override on steam should not be used on the same column with either reflux drum level control on reflux or a high reflux drum level override on reflux. It is the result of “inverse response” (see Chapter 13).166 Le91 Control and Feea!jimvard Options problem since reflux is a fknction of boilup. This is particularly true if a thermosyphon reboiler is employed or if the column has valve trays. should be followed. This is in addition to other problems of controlling base level by boilup (see Chapters 4 and 16). If the distillate is the demand stream. In a recent project we had considerable difficulty with base level control via feed accompanied by steam/feed ratio control. then base level control via bottom product and overhead level control via boilup will not control the column properly. then base level control via boilup and reflux drum level control via distillate will not control the column properly. At low boilup rates.

s. O. REFERENCES 1. ISA. and A.. . S. one of the level controllers mwt manipulate feed rate. New York. Elsevier.. J. P.. Budey. Maarlereld. 1975.Referemes 167 If distillate. Rademaker. P.” Proceedin. S. 2. E. 3. 1964. Texas ACM Symposium. “Selection of Optimum Final Element Characteristics. Rijnsdorp. “Design of Pneumatic Flow Controls. Buckley. Fifth National CHEMPID Symposium. bottom product. Jan. or side product is a demand flow. 1976.”Proceedinfls. L?ymma and G m m l $Continuow Distillution Unztr.

For the chemical industry. These columns not only are very large in diameter. 2. but also have a large number of product drawoffs and auxiliary heat exchangers. column designers sometimes tackle the separation of multicomponent mixtures with one sidestream drawoff column instead of using two or more conventional columns. 7. Whether to take side draw as a vapor or as a liquid is usually decided as follows’ : 1. Solvent-recovery systems in plants manufacturing plastics or synthetic textile fibers often utilize such columns.1 INTRODUCTION o reduce investment and energy consumption. Further. The objective here is to minimize low boilers in the sidestream. If side draw is taken from a point abuve the feed tray. If side draw is taken be& the feed tray. 169 . each with two product drawoffs. commonly 25-40 feet. The most extreme example of this philosophy is the crude column or crude still of oil refineries. it is usually taken as a liquid. This is done for the purpose of minimizing high boilers in the sidestream.7 t Control of Sidestream Drawoff Columns 7.2 SIDE-DRAW COLUMNS WITH LARGE SIDESTREAMS Sidestream drawoff columns sometimes have the task of removing small amounts of both low boilers and high boilers from a large intermediate boiler. Even the simplest side drawoff column is usually much more difficult to control than conventional two-product columns. it is usually taken as a vapor. such columns generally have much less flexibility and less turndown capability. In this chapter we will concern ourselves with this type of column. it is more common to see columns with a single side draw.

exact composition control of product streams requires as many streams as components. The overhead level control loop that marlipdates reflux flow is nested within the base level control loop. Since the top and bottom products are small. instead of emptyrng into the tray below. If this is not done. compounds that are heavier than the light key but lighter than the heavy key. As will be shown in Chapter 9. This scheme has been used successfully on a nonequal molal overflow distillation. When the side draw is a vapor.and bottom-product flows (or their ratios to feed) are available for composition control. Changing the side draw changes vapor rate t the o top of the column. base level is controlled by throttling side draw.4 COMPOSITION CONTROL OF SIDE-DRAW COLUMNS As indicated in Chapter 1. 7 3 SIDE-DRAW COLUMNS WITH SMALL SIDESTREAMS . which finally change reflux back down the column. Ideally the side draw should be ratioed to the feed. and steam flow are shown ratioed to feed.1 shows a typical control scheme for a column with a vapor side draw. An alternative arrangement is shown in Figure 7. this changes the rate of condensation.2. Let us consider how to provide composition control of the system in Figure 7. if it is very small. and as many manipulated variables as components.3. but the side draw is needed for column materialbalance control. Base level is controlled via side draw while surgetank level control sets reflux back down the column. Let us assume that the feed consists of lumped low . otherwise the base level controller on occasion may cause so much side draw to be taken that reflux temporarily will be inadequate. however. and steam will be wasted. column capacity will be limited.fSidestream Drawof Columns Figure 7. The downcomer.170 ControI . As shown in Figure 7. composition control will be poor. As mentioned in Section 1. column base holdup should be generously sized. 7. empties into a small surge vessel. but in practice. bottom draw (bottom product). composition of the sidestream is of primary importance. For this column. a distillation column sometimes tends to collect intermediate boilers. By measuring the reflux flow being returned from the surge tank. Here top. As an alternative one might control surge vessel level by side draw and base level by reflux from the surge vessel. the column-control scheme is very similar to that of a conventional two-product column with the addition of sidestream controls.9. we can use an override on the side draw to ensure adequate reflux. In this case a small sidestream is required. simple flow cone01 is sometimes used. Top draw (distillate). Here the side draw is a liquid.1. this requires overrides or limiters for the side draw. A practical problem is that of maintaining adequate reflux down the column.

7.4 Composition Control of .$&-Draw Columns 171 FIGURE 7.1 Basic control scheme for column with sidestream draw& .

2 Controls for liquid sidestream drawoff column .172 Control of Sidestream Drawof Columns FIGURE 7.

7.4 Composition Control o Side-Draw Columns f 173 FIGURE 7.3 Alternate control scheme for column with sidestream drawoff .

6B. See Figure 7. If turndown of more than 2: 1 is anticipated. . there are four specifications for the three product streams (three components assumed). It should be noted that many existing side-draw columns have neither composition controls nor ratio controls. This increases the reflux ratio. In some cases the purge flows are so small that the loss of product in them is deemed negligible. the valves for trays higher in the column open as valves for trays lower in the column close. thereby decreasing the concentration of intermediate boiling component B at both ends of the column. and lumped high boilers C.4. product B. For pneumatics the two bias relays are each biased such that relay output is 9 psig when component B reaches its maximum permissible concentration. If the signal from either analyzer becomes too high. 2. Figure 7. The top-product stream will consist mostly of A. See Figure 7. which are controlled as follows: 5 1. As shown in Figure 7.174 Contml o SirieStream Drawof Columns f boilers A. this easily can be implemented by conventional analog hardware. 3. Thus as the signal &om the XS1 composition controller increases. See Figure 7. 2. and the side product mostly of B with some A and C. increase the bottom-product/feed ratio. 7. Doukas and Luyben came up with the idea of changing the location of the sidestream drawoff tray.4. If there is too muchA in the side product. and if composition control is used.6A.5 AN IMPROVED APPROACH TO COMPOSITION CONTROL OF SIDE-DRAW COLUMNS To ain one more degree of freedom for composition control. many columns can make good use of steam/feed ratio controls. steam flow is increased. increase the steam/feed ratio. The concentration of the lightest component in the sidestream product is controlled by the location of the sidestream drawoff tray. If there is too much side product B in either the top or bottom product streams.5. impulse feedforward compensation (see Chapter 12) for feed flow will do a better job than ratio controls. If there is too much C in the side product. As shown in Figure 7. The concentration of the intermediate component in the top product (distillate) is controlled by manipulating the reflux ratio. the bottom-product stream mostly of C. increase the top-product/feed ratio. however. The fixed-gain relays are calibrated so that only two of the drawoff valves can be partidy opened simultaneously. The following control strategy therefore is suggested: 1.5 illustrates a method of minimizing the concentration of B in either the top product or bottom product. In view of the rising cost of energy.

7.5 A n Impwed Approach to Composition Control o St&-Draw Columns f 175 FIGURE 7.4 Scheme for control of sidestream composition .

fSidestream Drawof Columns 3. The concentration of the heaviest component in the sidestream product is controlled by the sidestream drawoff rate. 7.7 and 7. together with part of the FIGURE 7.6 PREFRACTIONATOR PLUS SIDESTREAM DRAWOFF COLUMN Doukas* studied two schemes involving a prefractionator as shown in Figures 7. controlled by heat input to the reboiler.176 Control . 4 The concentration of the intermediate boiler in the bottom product is . In both cases the functions of the prefiactionator were (1)to remove essentially all of the low boilers out of the top.8.5 Control of terminal composition .

0 c 3 L 0 yp 00:sg m m + .7. a) *c mm'E: = a ) = f g.s 5oyc iG-02 c g zc g % .6 Pr@&*w Plus Sidestream Drawof Column 177 x -EZ asp' 9 ) . u-.g cu -2:s a l 323 = 00 % E: -3 9) i3a) I U 3 m C 3.Em + . $!!ES a ma) a) 2a)Z g m U EB& %=* 2 -09) E 0.E "3.O%g SSU** .cl_o + m a n o.5 cc a e+. 2s % c a)ni % E 25 a) P ) a 8 a)':%. E ws2gg m CSP5+.

7 D-scheme .178 Cmrtrol of Sidestream Drawof Columns FIGURE 7.

6 Pr@-actim~or Plus Sidestream Drawof Column 179 FIGURE 7.-scheme .7.8 I.

N. 3. L. 37- 51 (May 1969).” CEP. Both of these schemes were shown to provide effective control for modest changes in feed composition. I?. 7.” ISA J.. Figure 7. Luyben. Luyben. .. A number of these are discussed briefly in a paper by L ~ y b e n . D o h . it is recommended for most systems.180 Control .” Ph. 4. 65(5):45trol Distillation Columns with Sidestream Drawoff. “Control of Sidestream Columns Separating Ternary Mixtures. “controlsfor Sidestream Drawoff Columns. 2. The intermediate product is then withdrawn fiom a tray or trays located between the two feed trays. ~ REFERENCES 1. Thus the heaviest and lightest components are detoured around the section of the column where the sidestream is withdrawn. and (2) to remove essentially all of the high boiler out of the bottom..” INE C H . The two product streams fiom the prefractionator are fed to two different trays of the sidestream drawoff column. D. Thesis. Since the D scheme is probably easier to implement.7 OTHER SCHEMES It is impractical to present here all possible schemes for controlling sidestream drawoff columns. Lehigh University. Buckley. 1978. and “pinches” are avoided. “Composition Control of DistiJlation Columns with Sidestream Drawoffs Separating Ternary Mixtures. Doukas.8 provides manipulation of the distillate product flow from the prefractionator to control sidestream composition. N.fShstream Drawof Columns intermediate boiler. L. 4 3 4 8 (June 1978). S.. together with the remainder of the intermediate boiler. W. Sidestream drawoff location is used to control the concentration of the lightest component in the sidestream product. and W..7 has been designated by Doukas and Luyben as the “L” scheme. “10 Schemes to Con- 42 (July 1966). The “D” scheme of Figure 7.

2. 8. Energy recovq-recovering and reusing the heat in the column product streams. conservation means designing and operating a column so that it makes the specified separation with the least amount of energy per pound of feed. Automatic control of composition of product streams. Geyer and Kline' give. whether they be liquid or vapor. as an example. an increase of 8 percent in energy consumption results. a 70-tray column separating a mixture with a relative volatility of 1.8 t Minimizing Energy Requirements 8. Operators commonly overreflux conventional columns with a single top product and a single bottom product. It can be shown that this results in a lower energy requirement per pound of feed than would feedmg on any 181 . Consequently we will discuss conservation first. but it is pointless to uy to recover energy unless we also try to conserve it. Feed provided at the proper feed tray. Conseraatian-designing and operating a column so that it makes the specified separation with the least amount of energy per pound of feed.2 CONSERVATION For distillation.6 percent high boilers in the base. If the operator adds enough boilup and reflux to increase overhead purity to 99 percent and base purity to 99. We have a number of techniques to accomplish this: 1. The main emphasis of this chapter is on the latter approach.4 and with specifications of 98 percent low boilers overhead and 99. 2.1 INTRODUCTION here are two basic approaches to minimizing &stillation column energy requirements: 1. Extra heat is used to ensure the meeting or exceeding of specified product purities.7 percent.

pressure and temperature compensation of flow measurements is highly desirable if steam is throttled instead of condensate. practically spealung. For vacuum columns there is some opinion' that mechanical vacuum pumps offer energy savings over steam jets. however. 8 3 DESIGN CONSIDERATIONS IN HEAT-RECOVERY SCHEMES . and the column may flood at lower boilup rates than it would when operating at higher pressures. But. Since low-pressure steam is seldom available at constant pressure or steam quality. Reserve capacities that may be required: -Extra heating capacity -Extra cooling capacity -Extra distillation capacity These are important for startups and shutdowns. Possible use of mechanical vacuum pumps. The two chief ones involve (1)"multiple effect" distillation. A number of schemes have appeared in the literature. Condenser capacity may be limited. analogous to multiple effect evaporation. Column operation at minimu pressure. and (2) vapor recompression.2Lower pressure usually means higher relative voIa&ty. the o p t i m u feed-tray location also changes.' In many plants excess low. changes in production rate. regardless of the scheme. The use of steam condensate receivers instead of traps reduces maintenance and steam losses. Where reboiler AT might be too small if the steam were throttled. the necessary separations can be accomplished with lower boilup/feed and reflux/feed ratios. one may use a partially flooded reboiler (see Chapters 4 and 15) and throttle condensate. Energy recovery in a distillation column means. Generally speaking. designed before the energy crunch. Dry distillation. 3. 4 Use of lowest pressure steam available. Older columns. it is sometimes economical to switch to steam-heated reboilers. can often benefit from new.Requirements other tray. and changes in product specifications. auxiliary" reboilers and condensers should be avoided. recovering or reusing heat contained in the column product streams. In many plants steam traps require considerablemaintenance and have sigmficant leakage. 5.182 Minimtzzw E M . This steam is usually cheaper than high-pressure steam. whether they are liquid or vapor. however. pressure steam is available that otherwise would be vented to the atmosphere. increased insulation. however. Therefore. Use of steam condensate receivers. As feed composition or enthalpy deviates &om design values. changes in feed composition. Insulation. there are five design factors that must be considered: 1. 6. if * "Auxiliary"condensers and reboilers are those installed in parallel with "normal" condensers and reboilers for startup or peak load purposes. The difference. 8. is usually small. For columns now using live steam. 7. .

say 8-10°C. Instead they always maintain at least a small load on these heat exchangers. considerable care must be exerted to obtain suitable value of controller gain and also proper valve sizing. this is sometimes a problem. Two methods have been used to allocate the energy to be recovered (1) throttling the vapor-heating medium to each condenser. Interactions. the gain 6 relay is calibrated to have an output span of 3-15 psig for an input span of 3-5 psig. load 6.4 MULTIPLE LOADS SUPPLIED BY A SINGLE SOURCE Sometimes. Accordingly. With elaborate heat-recovery schemes. but the concepts readily may be implemented with some digital or analog electronic controls. Process-to-process heat exchangers are commonly designed for very small temperature differences. 3-15 psig output. Inerts (low boiler) balance. Load 2 has the next highest priority. If recovered energy is t be distributed to several loads.8. Too high a concentration of inerts or low boilers will blanket process-to-process heat exchangers. Let us now look at three types of mdtiple-effect distillation. how is this to be dealt with? 4 Overall heat balance. we employ six gain 6 relays. as well as instrumentation and control complexity. This continues u t l the load with lowest priority. as shown in Figure 8.2. as shown in Figure 8. as shown in Figure 8.4 Multiple Loads Supplied /y a Sin5h Source 183 at all possible. too low a concentration will result in product losses through the vents. it should be recognized that the vapor-supply valves will tend to have small pressure drops.2. what o is the order of priorities? 3. each of which acts as a condenser. This obviously wastes energy. a column that is a very large energy user becomes the energy source for a number of loads. At the design stage of such a system.1. Some priority scheme must be established for startups and for any other occasion when vapor supply is temporarily short. and they prefer not to do it. so its gain 6 relay is calibrated for 5-7 psig input. One method of handling the priority problems is to use overrides and to split-range the various valves involved. Their use increases investment. and (2) operating each condenser partly flooded by throttling the condensate. which has the highest priority. it is advisable to have vapor flow control to each load. In view of the above. 2. it is apparent that control of columns with heat-recovery ifcl schemes is more d f i u t than control of conventional columns. How is this maintained? . 5 . Priorities.1. Elaborate heat-recovery schemes are often highly interactive. 8. Some users have had problems with turning auxiliary condensers and reboilers on and off. For load 1. with a set . 3-15 psig output. For the six loads in Figure 8. If vapor throttling is used. has a gain 6 relay calibrated for ni 13-15 psig input. The scheme shown illustrates the use of pneumatic devices.

184 Minimizing Energy Requirements 3 t -a 0 a c . 1 s i 3 a m d a C m 2 5 m -:8 i? a I *s 3 2 s5 E l .

8.4 Multiple Lo& Supphd ly a SinJle Source 185 c n + s Q & -0 m - 0 + m 8 c s U I E 3 5 NL oe d 2g 3 s gg .

the one supplying heat. that is. base level in the second column may experience serious upsets. For columns we have studied to date. maximum column pressure occurs at maximum feed rate and boilup rate. The A P control. but the AI' control will provide a rapid method of ensuring aonstant boilup. . 8. one can see that valve pressure drop will be very high a t low flow.3 and Figure 8. The variation in valve pressure drop with flow will be much greater than that normally encountered in a pumped system. SINGLE LOAD When there is only one source and one load (see Figure 8. The column that is the source does not need to be operated at a constant pressure-in the scheme shown. Assuming that the feed comes from a centrifugal pump. in turn. for the cases studied. the base composition of the load column is controlled by boilup in the supply column. Another problem associated with this scheme is the selection and sizing of the feed valve to the first column. If there is a serious discrepancy. it finds its own pressure.3 has an interesting dynamics problem. the first column gets only part of its heat from the second column. In another version a following column in the train supplies heat to a preceding column as shown in Figure 8. The controls must be so designed that changes in vapor flow from the supply column must reach the condenser-reboiler at about the same time as feed flow changes from the supply column. we have found it advantageous to let pressure find its own level in the second column. Vapor flow to the first column from the condenser-reboiler will not be constant. Again.3). the remainder comes from an auxiliary reboiler.4. After examining some complex heat-balance schemes. may have its set point adjusted by a composition controller for the lower section of the first column. For the illustrative example. and low at high flow. In the absence of vapor flow control.4. In this particular case. An interesting practical problem here is how to adjust the auxiliary reboiler on the first column.5 SINGLE SOURCE. the overhead composition of the supply column is controlled via reflux. control may be both simpler and more flexible. we decided that the simplest approach was to use column AI'. Interactions between the two columns may be severe. and very close control of supply and load pressures may be required. This column will run at a low pressure at low feed rates and at a higher pressure at high feed rates. It should be noted that for the schemes of both Figure 8. there has been no problem with flooding at lower rates and pressures. interactions may be severe. The scheme of Figure 8. particularly if the secondcolumn bottom-product flow is smd.186 Minimiziv Energy Requiremoats point fiom a primary controller.

Si@ i k 187 Load - 2 tn 0 a3 C vl ai 8 z 5 u a3 .5 Sw Swrm.i eg "Os i? 22 ZX ss .8.

188 Minimizing Ewm Requirements N E 5 W z m - ‘i 0 W C VI 2 -l m c 8 i s i “os 22 25 e l -?g F .

5). involves splitting ~. Another factor favoring vapor recompression is a small temperature difference between the top and bottom of the column. The feed enthalpy or temperature. The supply column. runs at a higher pressure than the load column.8 Energy R e m 7 iy Vapor Recompresswn 189 8. steam-heated reboiler and/or auxiliary ~. water-cooled condenser may be necessary for startup (see Figure 8. but rather at an intermediate site. N d 9 presents investment equations and data. This may make column-composition control difficult unless one employs either feedforward compensation or a t i heater with control for constant rm temperature or enthalpy. Today the main interest is in getting the column vapor compressed to the point where its temperature is high enough to permit using the vapor as a heat source for the r e b ~ i l e r . Such schemes have been used in the chemical and petroleum industries for years. The feed split is controlled to maintain a heat balance. Since feed flow is typically set by level controllers or flow-ratio controllers.An~auxiliary.8.) 8. Other papers on energy integration for distillation columns include those by OBrien” and by Rathore et al. which is used in some s y ~ t e m s .~ feed between two columns that make the same separation (Figure 8. The incentive in many instances was to be able to use water-cooled condensers. however.6).12 .~ Mosler7 discusses the control of a number of vapor recompression schemes. its flow rate will not be constant. therefore.7 COMBINED SENSIBLE AND LATENT HEAT RECOVERY In addition to the recovery of the latent heat of vapor streams. 8.6 SPLIT FEED COLUMNS A third arrangement. Fahmi and Mostafa” indicate that the optimum location at which to use the compressed vapor may not be in a reboiler at the column base. is apt to be variable. (See Chapters 5 and 11. in many cases it is practical to recover part of the sensible heat in the column bottom product and steam condensate by exchange with column feed. A review of compression equipment and methods of estimating operating costs has been presented by Beesley and Rhine~mith. thus avoiding the expense of refrigeration.8 ENERGY RECOVERY BY VAPOR RECOMPRESSION In the past vapor recompression (“heat pumps”) has often been considered for distillation of materials boiling at low temperatures.

190 Minimizing Enevgy Requirenents U m m c - 0 I w m % e + - 2g ZE s5 7 E dl .

8 EWJJ~ Recovery Czy Vapor Rempession 191 rnuuuLi FIGURE 8.6 Heat recovery via vapor recompression .8.

(July 1976). K. E. L. T. tions. Rhinamith. 3. H. J. 59-66 (Sept. e.. F.”AICbE J. Proc. F. Bmley... D s Dev. A. 8. Chiang. A. A. R. 1980). CEP. 37-41.. 11. “Reducing Column Steam Consumption. CEP. and R. E. Mostafa..” CEP. N. 65-67 CEP.. Dtitillation ControL. R. Recompression.” paper submitted to IOEC 49-51 (May 1976). 1978).” CEP. G. Shaner. F.” Proceedangs. 1977. 5. CEP. Powers. lation. “Distillation with Optimum Vapor 4.12. (Aug.. Fl. Kline. M. 1974. L. H.44-49 (July 1980). R. Shinskey.. S. Tyreus. AIChE Distillation Systems with Energy Workshop on Industrial Process Integration. Luyben. 58-64 (July 1976).. 47-52 (May Des. 20(5): 490control. “Heat . 1983). and H. Rush. 1976). (1983). L.. and P.” Cbem. O’Brien. D. and W. Fahmi. “Coneol of Sidestream and Energy Conservation Distdla. “Synthesis of tion Towers. G. Res. 10. Vanwormer. 2. Null. Luyben. 391-392 (Nov. Rathore. and W. N.. Geyer. Mosler.. 1974). and G. G.192 Minimking Energy Requirements REFERENCES Integrated Distillation Configura1.... A. R. 7. Tampa. New York.. 950 (Sept. 6. B. “Heat Pumps in DistilMcGraw-Hill. Eng. CEP. H. D. 9.

are approached or reached. To make a given control valve respond to more than one controller. Automatic controls. most existing plant instrument systems consist of a large number of single-loop controls: one transmitter. such as high column Al’.1 INTRODUCTION s mentioned earlier. although it may be deduced and written down in advance. To put it in 193 . In real life. One must superimpose upon them a kind of logic that spells out various courses of action to take as certain constraints. In the case of highly integrated (“bootstrap”) plants. the required reaction times in certain circumstances may be too short for typical human physiology. In addition. provide continuous (or almost so for digital controls). For conventionally instrumented plants. we must have a means of telling the valve which controller to obey. They are also much more reliable than human beings. one controller (usually proportional-reset). or high temperature. however. simultaneous surveillance and recognition of operating conditions and continuous. With existing systems the operator usually must put the control loop into the “manual” mode and adjust the control valve position as all but one of the pertinent variables change. on the other hand. many a valve must be adjusted in accordance with changes in more than one process variable. For example. low tank level. the steam valve to a distillation column reboiler may have to be adjusted in response to as many as half a dozen variables. this logic is a major part of the formal instructions for operating personnel.9 a Application of Protective Controls to Distillation Columns 9. may be too detailed and complex to be absorbed readily by human operators. the required logic. conventional “automatic” controls are automatic only in a limited sense over a limited range of conditions. quantitative response to them. one manual/automatic station. To put it another way. and one valve.

we greatly improve our ability to achieve certain other control objectives: -Automatic. rather than abrupt.~ 9. take over selectors. Usually interlocks must be reset manually by the operator. overrides are most useful in protecting the process and keeping it running as certain maximum or minimumpermissible operating conditions are approached. startup and shutdown -Automatic total reflux operation under specified circumstances -Maximum-capacity operation -Ability to make desired transitions in column terminal product compositions with minimum production of off-specification products -Minimum interlock shutdowns. and perhaps also to shut off steam. High column base temperature and high column differentialpressure are sometimes interlocked to shut off steam to the reboiler. we want a multivariable control system instead of multiple. . interlocks are most u s e l l in cases of equipment malfunction or failure. for example.194 Applicahn of Protective Controls t o Dtitillation Columns process control language. single-variable control loops. that is. we illustrate a number of techniques for accomplishing ti with specific applications hs to distillation columns. As a generalization. In the remainder of this section. We use the term 0 mh the most part to refer to the use of two or p . For example. the most common types of protective controls are interlocks and overrides.for more controllers connected to a control valve through high or low signal (Le. ability to stay on the line a higher percentage of the time -Minimum turndown requirements for both process streams and utilities The fourth item is particularly important when feed-stock composition varies widely and it is desired to optimize column or train operation as. that is.2 OVERRIDES AND INTERLOCKS In the chemical and petroleum industries. In so doing. they do not have to be reset manually. if a column feed pump fails. or at least easy. By contrast. a low feed flow interlock may be used to shut the column drawoff valves. with a computer.. overrides can be designed to provide gradual. Logic is built in to enable one controller to “overrideyy from) the other controller or contr~Uers~. Interlocks normally function in an abrupt manner to shut down a piece of equipment or one or more steps in a process. The usual objective is to provide protective controls that permit the column to operate close to constraints without exceeding them. corrective action and function in both directions.

2. or down to a predetermined minimum value (LL) as shown by Figure 9. Their accuracy is poor. One may also construct a median selector with two high selectors and two low selectors as shown in Figure 9. or to hold it a safe distance from a constraint. If either the high or low measurement deviates too far from the median value. So far we have found no applications to distillation. a low limiter is a signal source plus a high selector (Figure 9 .3 Implementatirm . Correspondingly.it also may be assembled with two devices or purchased as a combined device. but it is a technique that is worth keeping in mind.3.1.9. usually a valve. In practice one or more selectors are inserted between the output of a “normal” controller and its final control element.3 IMPLEMENTATION OF OVERRIDES Many ways of implementing overrides are possible. . one of the override controllers will “take over” or “override” the normal controller and drive the final control element in the proper directioneither to force the process away from a constraint. The major application has been for auditing multiple flow measurements to a chemical reactor. High Limiters (HL) or Low Limiters (LL) High and low limiters are devices that reproduce the input signal 1:1 up to a predetermined maximum value (HL) as shown by Figure 9. As many input signals as desired may be accommodated by arranging selectors in series. but a particular one that is inexpensive and works well involves the use of combinations of devices such as described in the following.4). As process constraints are approached. If pneumatic versions with adjustable gain are required. The outputs of the lo “override” controllers are a s connected to these selectors. an alarm or interlock is activated. High Selectors (HS) or Low Selectors (LS) These devices select either the higher or the lower of two input signals. In either case the cutoff or limiting value is readily adjustable. one should avoid those with pressure-dividing networks. Functionally. 5 ) . It thus can be constructed with two separate devices or assembled into one housing. Summers Summers are devices that add or subtract.f Overrides 195 9. a high limiter is a low selector plus a signal source such as a supply regulator (Figure 9. Multiple-input low-signal selectors are now available from vendors.


ApPlicatMn o Protective Controh to Dirtillation Columns f


Median selector (J. P. Shunta design)

9.3 ImplementatMn o O v d s f


FIGURE 9 3 . Low limiter



of Protective Cmmh to D&d!&on


FIGURE 9.5 L w iimiter schematic o

Three types of controllers have been used in override circuits; two of them widely, and the third less frequently.

Proportional-Reset(PI) or Proportional-Reset-Derivative (PlD) Controllers
Both of these are commonly used as normal controllers, although the PI type is occasionally used as an override controller. PID controllers, if used, should be so designed that the derivative acts only on the measurement signal, not on the controller output. Both PI and PID controllers must contend with the problem of reset windup, discussed below.

Proportional-Only Controllers
The fixed-gain proportional-only relay, either direct acting or reverse acting, with adjustable bias has become one of the most widely used devices in override circuits. Common values of gain are 2, 3, 4, 6, and 25. Use of the minus sign ( - ) with these figures implies reverse action, or negative gains. A proportional-only controller or relay follows a straight-line equation.
0, = KO,



0, = output signal



input signal bias proportionalgain


= =


We have found it convenient to calibrate pneumatic, fixed-gain relays in terms of the input signal, [O,],, required to produce 9.0-psig output. Then


= K[Oj]9









On substituting equation (9.3) into equation (9.1) we obtain:

K (ej - [o,],) + 9 (9.4) The instrument may now be calibrated by putting [Oil9 into the input and adjusting the bias for 9.0-psig output. For override purposes we use proportional-only conmollers if possible, largely to avoid some problems associated with PI controllers. If the latter are used for override purposes, no fixed relationship will exist between controller output




Application .fProtective Controls to DirtiUatMn Columns

and process variable. A sudden disturbance can cause an overshoot above the set point, with the amount depending on the magnitude and rapidity of the upset and the reset time and proportional band of the controller. Further, for a given upset, the output of a proportional-reset controller usually swings iis through wider lmt than the output of a proportional-only controller. These features make this type of override highly undesirable in those applications iis where, for safety‘s sake, certain lmt must not be exceeded. Maximum temperature in some chemical reactors is a good example of this. We thus use proportionalreset override controllers only in those cases where good control could not be obtained with gain 2 or higher proportional-only overrides because of closedloop stability problems. Examples include column AI‘ or base pressure. iis we For predetermined maximum and minimum lmt, select a proportionalonly controller with a gain that will drive the valve fiom 3 to 15 psig (or vice versa). Thus: 15 - 3 Proportionalgain = max - min where “max” and “min” are in terms of the process transmitter output. Consider a system where a low base IeveI override is to be used to close the steam valve. Its output is compared with that of the normal controller through a low selector. Let us say that we want full override action to OCCLZT between zero level (3 psig) and the 25% level (6 psig). Then the required proportional gain is: - 15 - 3 12 Proportionalgain = - --4 6 - 3 3

Since the gain 4 relay is to put out 15 psig at the 25 percent level, and its output goes to a low signal selector, it clearly will exercise no control action at a level of 25 percent. On the other hand, if level is dropping, the output of the gain 4 relay will drop below that of other controllers at some value of level above zero. We are never sure, therefore, a t what point the low-level override will take over, but we know positively that it will be between the zero and 25 percent levels.

Floating or Integral Controller
This type sometimes is used as a normal controller instead of a PI controller when required proportional gain would be very low or where it would need to be changed fiequently as process conditions change.



When control of a valve is transferred from one controller to another through a selector, we would like this to happen without a bump. Since the

9.5 Anta Reset-Windup


selectors will switch on very small differential signals, they will in themselves introduce no sigdicant jolt. The major potential source of bump is reset windup in overridden controllers. In al commercial pneumatic controllers, reset action is obtained by a positivel feedback, unity-gaincircuit that feeds the controller output signal back through a needle valve into a reset chamber. This feedback is normally internal, but in some commercial controllers it may be taken from an external connection by proper orientation of a switchplate on the body of the controller. If, now, we use the valve-loading signal as the reset-feedback signal, we will get normal reset action when the controller is controlling. When, however, another controller takes over, reset action in the first controller ceases, but its output goes up and down with the valve-loading signal. This tracking action is delayed by the reset time constant, which for many controls (such as liquid flow control) is small. The output of the overridden controller will differ fiom the external reset feedback signal by the product of its gain, K,, and the error signal, E . The a t reset-windup technique discussed above is known as “external reset ni feedback.” For most applications either it, or the modification mentioned below, is our preferred scheme. It has the disadvantage that the controller output signal, commonly labeled ‘balve position,” is really different from the actual position. It differs by the product of the error signal times the proportional gain. Lag in the reset circuit may cause further error. A modification therefore is introduced by some vendors, particularly in the newer microprocessor controls. This consists of setting the reset time equal to zero when the controller is ~verridden.~ technique is sometimes called “integral tracking.” It should This not be used with auto overrides. Another technique is called “output tracking”; the overridden controller output is driven to almost the same value as that of the overriding signal. This does have the advantage that the controller output signal is nearly equal to the ,~ valve-loading signal. However, according to Giles and G a i n e ~ this technique as well, at least under some circumstances, as does integral does not work tra~king.~ Various other techniques have been described by Khanderia and Luyben6 As an example consider the simple system of Figure 9.6. It shows a &titlation column with a base temperature controller and a column A P controller, both connected through a low selector to the steam valve, which has air-toopen (AO) action. Let us assume that at startup time the base of the column has enough low boilers that it will boil at a temperature lower than the normal temperature controller set point. If we do not lower the set point, the temperature controller will open the steam valve wide, the column-base contents will boil very rapidly, and column pressure drop will shoot up. As it goes above the A P controller set point, this controller output starts to decrease, and when it becomes lower than the output of the temperature controller, it takes over the steam valve through the low selector. The steam valve is now held open just far enough to keep column Al’ at the set-point value (chosen to be the maximum acceptable). Eventually the low boilers are taken overhead and base temperature rises to the point where the temperature controller takes over. Since the operator


Appliuuion of Protective Controh to Dirtillation Columm

does not need to change the AP controller set point, it can be located away fiom the panel in an override cabinet. Several points shodd be noted about this system: -Since the override controller is of the proportional-reset type, there will not exist a fixed, known relationship between the controller output and process transmitter output. -The overriding controller-in this case differential pressure-must have a smaller reset time than the normal controller or it will sometimes take over at values of column AP different from the override controller set point. -Manual-automatic switching is not necessary since both controllers have anti reset-windup. -If the system of Figure 9.6 is started up in the automatic mode, the control valve will open and the controlled variable will rise toward the set point at a rate determined by the reset time constant. The first two points lead us to use proportional-only overrides where possible, while the third suggests that automatic startup and shutdown can be obtained with an air switch and several three-way pneumatic valves. It should be noted that only a few of the electronic analog controllers on the market have external reset feedback or some other satisfactory anti resetwindup scheme. But newer digital and microprocessor-based controls (as of late 1983) use a variety of techniques, most of which appear to be satisfactory.
Cascade Controls

The preceding technique works well for conventional single-loop controls and for secondary or slave loops in a cascade system. But for primary or master controllers, we do something different since the valve-loading signal is no longer meaningful for reset feedback. To eliminate reset windup, we break the master controller internal feedback as before, but now we use the secondary measurement for feedback as shown by Figure 9.7. If, for example, we have temperature cascaded to flow, we feed the output fiom the flow transmitter back into the master controller reset circuit. This means that during normal control the lags in the secondary control loop appear in the reset feedback circuit of the primary controller. If, as usual, the hs slave loop is much faster than the master loop, ti technique will not z,>preciably increase the master controller reset time.

A convenient method for providing feedforward compensation that does not interfere with either normal reset or antireset windup involves the use of an “impulse” relay and a summer as shown in Figure 9.8. In pneumatics these functions are sometimes combined into a single device. The impulse function

9.6 Fee&mard Compensation with Overrtdes


FIGURE 9.6 Column base temperature control with AP override


Appluutiun o Protective Controls to Ditihktim Columns f


8. 0

-2 0


0. S W C


% wv,
e Q 3.k


9.7 Ov&

f Column Overhead S s e w ytm


passes only the transient part of the feedforward signal. This eliminates or avoids s c a h g problems with the summer since the steady-state or dc component of the feedforward signal is blocked. It can be shown that best results are usually achieved if the impulse relay time constant is set equal to the PI controller‘s reset time.’ The theory is discussed briefly in Chapter 12.



Let us assume we have a conventional column with the following normal controls:
--Condensate receiver level controls top-product drawoff. -Base level controls bottom-product drawoff. -Reflux is ratioed to feed. -Steam is ratioed to feed.

As indicated by Figure 9.9, this column has a horizontal condenser and vertical, cylindrical condensate receiver or reflux drum. We will assume that the level controller is of the PI type with set point at midscale of the level transmitter span and that gain 2 auto overrides are employed. The level overrides then function as described in the following.

FIGURE 9.8 Impulse feed forward with PI controller and overrides

9 Overrides for column overhead system .fProtective Contvols to Dktilhion Columns FIGURE 9.206 AppliGatiOn .

This seldom works well. This would require an additional override on the reflux valve. if the reflux drum loses its level.5 psig. This permits the reflux valve to open and the reflux controller to take over somewhere in the lower 25 percent of level. however. S1 starts the pump. These two overrides are particularly helpful for a column with a high reflux-to-distillateratio. Without such protection moderate changes in reflux flow could either flood the receiver or run it dry. when the level reaches 25 percent (6 psig). We are using this scheme increasingly.9.9. A10 usually has a gain of -25 and is so calibrated that it has an output of 9 psig when its input is about 3. High Condensate Receiver Level Override on Reflux If liquid level rises above 75 percent. the output of reverseacting override A1 with gain 4 starts to decrease from 15 psig. to have overrides act directly on the valve as in Figure 9. This drives the reflux valve open. override A10 holds the cooling-water valve open as long as the condensate receiver level is low. If the level drops again.7 Overrides@ Column Overhead System 207 Low Condensate Receiver Level Override on Reflux As liquid level rises to the lowest position (3 psig). Overrides on Distillate Valve The distillate valve may be held closed by override A3 u t l overhead ni composition or temperature reaches a satisfactory value. This permits accurate calculation and prediction of override behavior. This design should be compared with that of Figure 3. For some columns it is desirable to maintain a minimum reflux flow rate. it may also be closed by high liquid level in the next process step. It has been a more common practice in the past.26 where a PI level controller was used with auto overrides.2. or in the case of vacuum or pressurized columns. Low-Low Condensate Receiver Level Override on Cooling Water To guarantee condenser cooling-water flow during startup. and high and low level overrides acted on the reflux flow controller set point. Similarly. the output of reverse-acting override A2 with gain 4 starts to decrease from 15 psig and reaches 3 psig at 100 percent level. some plants like to control condensate temperature by throttling cooling water as shown in Figure 3. As the condensate receiver level transmi&r output increases above 3 psig. The normal controller for this valve may be the cooling-water exit-temperature controller. . this override starts to close the reflux valve. the output of A1 is 3 psig. it may be the pressure controller. the pump shuts down. Condenser-Cooling-Water Overrides As mentioned in Chapter 3. Reflux pump operation is made automatic by switch S1. and overall is probably preferable.

10.208 Applicatiun o Protective Contmh to Dktilhttim Columns f if at all.8 OVERRIDES FOR COLUMN-BASE SYSTEM The column-base system shown by Figure 9. causing the condensate temperature controller to cut back on cooling water. the cooling-water exit temperature rises. The base liquid level transmitter is installed and calibrated FIGURE 9. it is usually desirable to limit exit cooling-water temperature to a maximum of 50-60°C. so it is increasingly common to provide an override from coolingwater exit temperature as shown by Figure 9. To minimize fouling and corrosion. vertical-tube reboiler. 9. particularly if 316 SS is used.10 High cooling water exit temperature override on condensate temperature control .11 features a thermosyphon. If the condensate temperature becomes too low.10 will accomplish this. The override scheme of Figure 9.

8 O v d s fm Column-Base System 209 FIGURE 9-11 Overrides for column base system .9.

and the set point is at midscale. Pm -Override set pressure. . Whichever is reached first will then initiate override action. The level overrides then function as in the following. High-Base-Level Override on Feed As liquid level in the base rises from 75 to 100 percent. This prevents “baking” the reboiler tubes (which fouls them) when there is too little liquid to keep the tubes wet. consider four pressures: -Relief valve set pressure. For determining maximum permissible pressures for overrides and interlocks. this closes the feed valve through a low selector. just short of flooding. override A9 output increases from 3 to 15 psig. -4 Low-Level Override on Steam As liquid level rises from zero to 25 percent. A convenient way of determining maximum column A P is from the equation: A r m = = 1. PN A convenient way of setting Pm and PoR is to let and . override A7A or A B . acting through a proportional-reset controller. the output of gain override A5 goes from 15 to 3 psig. l’Rv --Interlock set pressure. This permits the steam valve to open and the steam controller to take over somewhere in the lower 25 percent of level.2 gives a Al’. This is based on the common practice of designing columns such that the flowsheet value of boilup is about 80 percent of the boilup that would cause flooding.2 Al’Fs (9. The factor of 1. will limit maximum steam flow.. This is accomplished by setting a 9-psig set point on the controller and biasing gain 1 relays A7A and A7B such that their outputs are each 9 psig when the maximum permissible column Al’ and base pressure are reached.5) where Al’Fs is the flowsheet value of Al’. the override stam to close the steam valve.210 ApplicatMn of l’mtective Controk to Driti&‘utiun Columnr as recommended in Chapter 4 The PI controller has gain 2 auto overrides. P O R -Normal operating pressure. If the level drops below 25 percent. High-Base-Pressureor High-Column AP Override on Steam As increasing boilup increases column AP or column-base pressure (especially during startup).

Typically the liquid level rises well above the vapor inlet from the reboiler and up over some of the lower trays. With s t e d f e e d ratio control. Depending on circumstances. depending on whether the column base is full (“wet”) or empty (“dry”)when the column is shut down. Instead. In most cases we have found it advisable to adjust override biases so that a high level pinches feed before steam. (Some column designers prefer to provide enough space below the first tray that it can never be flooded. 9.6. Miscellaneous Column-Base Overrides The base system contains two more overrides. A5 has an output of 3 psig. the column comes on line at the maximum speed permitted by the constraints. wet column startups and shutdowns are much faster. air-operated valve in the air line between the steam valve and its overrides. an automatic startup may be accomplished with an air switch connected to a three-way.) Such a shutdown may be made to accommodate production scheduling or may have been forced by process interruptions elsewhere. Wet Column Startup If a column is shut down by closing the steam and drawoff valves. as soon as steam and feed are available. the entire column contents accumulate in the Iower section of the column. But it is desirable to maintain enough boilup that the trays (sieve or valve) do not dump or weep. override A5 closes the feed valve. It should also be noted that the bottom-product valve may be closed by an override from the next process step.9 Aut& Startup and Shutdorpn 211 Minimum Steam Flow Controller As liquid level rises from 75 to 100 percent. A6 closes the steam valve if base temperature becomes too high. During startup A4 holds the tails valve closed until base composition reaches a desired value.9 AUTOMATIC STARTUP AND SHUTDOWN Two kinds of startup and shutdown are commonly encountered. The “on” position corresponds to a through . For this we provide an override (not shown) that is a minimum steam flow controller. The same basic override system permits automatic startups regardless of whether they are wet or dry. this would also cut off steam. As will be seen.9. It should be noted that no formal program or sequence control is necessary. At a level transmitter output of 15 psig. a column may or may not follow the same sequence of constraints from startup to startup. The bottom-product pump is operated automatically by switch S2. which turns the pump on whenever the base level transmitter increases above 6 psig. As shown in Figure 9. The column may be started up again simply by turning on the steam.

This procedure has worked well for small columns. In doing this one should also allow for displacement of liquid from a thermosyphon reboiler. it is the fastest possible procedure for getting a column on line. For larger columns there have been some cases where the lower trays were damaged. Startup of the train is accomplished by turning the master switch to the “operate” position. In the meantime inventory is starting to build up in the reflux drum. once boilup has been established. In addition. Eventually decreasing low boiler concennation in the base permits the bottom-product valve to open and be controlled by base level. The adjustable lag box (commonly an inverse derivative u i . With all controls on automatic. It is probably safer. including condenser cooling-water controls not shown in Figure 9. the output of the overrides is dead-ended and the signal connection to the valve is vented to the atmosphere. in the ‘‘of€’’ position. Since. As before all controls are initially on “automatic.” The steam valve immediately begins to open slowly. Some plants have spare or utility tanks that can be used as temporary reservoirs for excess column-base contents. the second column cannot start u t l ni . most of its tube volume will be vapor.212 Applitltiun 6Protective Controh to Diitilhtwn Columns connection.” the steam valve opens slowly. the columns have little or no inventory.or. It is probably preferable for each column to have its own switch. Liquid then drains down into each column base. puts a train on stream much faster than does a dry column startup.6. It stays at a partly open position until dropping base level closes it further. therefore. since each column in the train starts with adequate initial inventory. this shuts off the steam. Dry Column Startup In this case. feed. When the switch is moved to “shutdown. This procedure. it is only necessary to switch the master switch to “shutdown”.” The steam valve starts to open as soon as base inventory builds up. but a master switch can be used to start a complete train. in electronics nt a first-order lag or integrator) is so hooked up that when the switch is positioned to “operate.” the steam valve is closed immediately. the operator pushes the switch to “on. All controls then operate as indicated previously. To shut down a column or train. it is not necessary for downstream columns to wait for upstream columns to get started. therefore. Since this procedure permits maximum possible steam flow with only one or two constraints. and drawoff valves. The column stays on total reflux until overhead composition is satisfactory and reflux drum level has risen high enough for the overhead composition controller to take over the reflux valve. and continues opening u t l the ni column AP override pinches it a little. The column feed valve stays closed u t l ni f h g base level permits it to open. 6 feet or less in diameter. Column-base composition is then controlled by throttling the steam valve. however. before startup to pump column-base contents down to a level well below that of the bottom tray. top or bottom. say.

Shutdown is accomplished by shutting off feed to the first column. the bottom.” To get everything out of the columns. High Column Inventory If. three-way valves in the air lines to the drawoff valves. Note that feed pump failure will put the column on total reflux by this method. For total reflux a separate switch may be provided that actuates air-operated. Each column then automatically works its overhead and base inventories down to the lower 25 percent zone. the feed valve to a column is closed manually or closed by some upstream override. during normal operation. manual drain valves must be opened and residual column contents taken off in drums or other containers. as might be required for maintenance work. The starmp/shutdown switch then controls only the feed and steam valves. When this has been done. This design permits the operator to put the column into total reflux. or to take it out of total reflux during startup upon ascertaining that it is proper and safe to do so. Normal operation is restored by putting the feed valve back on automatic. the overhead and base inventories build up until the feed valve is shut off. The column is then on total reflux with the steam valve controlled by AI’ and reflux flow set by condensate receiver level. it has been proposed to switch first fiom total reflux to recycling product streams . The column can be put back on normal control by switching the two level controls to automatic. Product Recycle to Feed Tank To minimize the transition from total reflux to normal operation.10 “IDLE” OR TOTAL REFLUX The controls discussed are capable of operating the columns in an “idling” or total reflux condition with either a high or a low column inventory. Low Column Inventory If. when desired. the column will work its overhead and base inventories down to the 25 percent levels and then put itself on total reflux.9. during normal operation. or by the total reflux switch. At this point the operator must exercise manual control of the bottom product and steam valves to work column inventories down to very low levels. it may take a long time to get the whole train going. 9. the columns may be shut down by switching the master switch to “shutdown.10 “Idle“ m Total R$ux 213 it begins to get feed fiom the first.and top-product valves of a column are closed manually or are held closed by a total reflux switch. or by coming out of the override condition.

however. One popular method is to provide a bypass line with a restricting orifice. Another approach. to "dead head" a centrihgal pump very long. This eliminates the upset that sometimes occurs when switching fi-omtotal reflux to normal operation. The controls are so arranged that as the flow approaches the specified FIGURE 9.12. uses a small valve in a bypass line. It is undesirable. Then when the column is lined out-that is. it is sometimes necessary to have a centrifugal pump running for a period of time while no product is being taken. We are not aware.214 Appliuaim o Protective Conh-ohto Dictillation Columns f back to the feed tank. Various expedients have been devised to take care of this. making the desired separation-one may close the lines back to the feed and open the normal product lines to storage or the next process step. of any industrial applications. 9. shown in Figure 9.1 1 MISCELLANEOUS OVERRIDES Centrifugal Pump Bypass At startup and shutdown. however.12 Scheme for protecting centrifugal pump against dead heading . This is a low-investment design but it wastes horsepower.

the high-gain relay can be set to open the bypass valve as the normal valve approaches the closed position. Overrides may be provided such that if either valve goes beyond this ni value. Generally.This is so because the increase in entrainment o&ts the effect of increased reflux. It should be noted. A. . By measuring XD and WK and computing h D / d W R . the feed valve is pinched slowly u t l the utility control valve’s position is back to a reasonable value. one can devise an override control system that maintains a positive value of h D / d w R by pinching the steam or feed valve. it is not desirable for valves to go more than 95 percent open or less than 10 percent open.11 M ~ ~ U u n e o w Overrides 215 minimum value. This is not an override in the normal sense since no signal selectors are needed. Limited Utility Override If the condenser or reboiler limits before the column proper does.versus reflux flow. If the bypass line cannot be connected back to a tank or receiver as indicated. but must be connected to pump suction. xD. In “bootstrapyy plants. If a flow measurement is not available. A possible hookup is shown on Figure 9. at some high feed rate. except on a momentary basis. This would replace the AP override. an inmeme in reflux flow will cause a decrease in overhead purity.15. Entrainment Override With the previously discussed system.13. This usually occurs at a column AP considerably higher than the nominal maximum permissible AP. certain additional overrides may be needed. however. To put it another way. shows a maximum value of xD at a particular reflux rate as shown by Figure 9. a high-gain relay opens the normally closed bypass valve. The high-gain relay should have hysteresis or detent to prevent chatter. Overrides for Maximum Capacity Where it is desired to be able to run a column or train at maximum capacity. and AP.14.16. This requires a slow-acting override incorporating a propomonal-reset or floating controller. for example. however. For highly integrated plants. One automatic approach is to connect header pressure to overrides on the reboiler steam valves to close them partially as header pressure drops. a cooler should be installed in it to prevent overheating. A particular arrangement is shown in Figure 9. it may be necessary during wet startups to avoid putting such a demand on the steam header as to reduce header pressure seriously. Implementation is shown in Figure 9. it may be necessary to protect the utility supply during startups. wR. fixed value. For some columns. W R . one may use either the cooling-water valve position or steam-valve position for override purposes. a plot of top composition. xD. the limitation on any one column is the maximum permissible column AI’ that was taken to be a known. 8. for good control.9. that the peak of the xD versus wR curve does not occur at fixed. as one increases feed rate slowly and before flooding starts. reproducible d u e s Of XD.

14 Entrainment override .216 Applicatwn of Protective Controk to Dtitillation Columns FIGURE 9.

a zone between hard and soft constraints. as shown in Figure 9. then becomes the pressure controller normal set point. 9. If. the overrides must be so designed that the process does not normally approach the hard constraints too closely. thereby permitting maximum column capacity. the summer or “balance” controller calls for a slightly higher column pressure to increase condenser heat transfer. C. which is limited by stability considerations. * It span. as.17.12 Destgn Considevatwns 217 C. The manipulated variable always reaches its maximum (or minimum) value before the process variable exceeds its hard constraint. The column and auxiliaries. In many cases they are also subject to truly hard constraints. Since any feedback control system must have some room within which to work. In the case of overrides with proportional-only action. the condenser cooling-water valve opens wider than the steam valve. we can visualize. provided the column itself does not limit first. and if its loading signal is less than 4. Note that a high limiter holds signal A constant for cooling-water-valve loading signals greater than 4.18. for example. The width* of this zone is determined by the override loop gain. The soft constraint will correspond to the minimum (or maximum) value of the override output. Automatic Balancing of Condenser and Reboiler Heat Loads If the condenser and reboiler do not have balanced heat-transfer capability. is to feed the steam valve and cooling-water valve positions into a summer with gain whose output becomes the “set point” of the column pressure controller as shown in Figure 9. The summer bias.2 psig. One way of balancing these two heat exchangers on pressurized or vacuum columns. The takeover point between normal and override controls will be at a variable position (depending on operating conditions) somewhere between the hard and soft constraints. for example. must have adequate static pressure ratings.9. and the required range in pressure must not be so large as to cause adverse changes in relative volatility.12 DESIGN CONSIDERATIONS It is probably apparent that most overrides are really feedback control loops.8 psig. one of them will limit before the other does as feed rate increases. is very helpful to think of this zone as a fraction of the measured variable transmitter . there is a unique relationship between the value of the manipulated variable and the distance between the process variable and its hard constraint. They therefore are subject to stability considerations. With a proportional control loop designed for dead-beat response. maximum column-base pressure. of course.2 psig and a low limiter holds B constant for steam-valve loading signals less than 13.

218 AppLicatiOn af Protective Controlr to l>rsrillatiOn Columns I! c 2 L C 0 8 22 B P ki gz gg 3% .

9.12 Des@ Consz.17 Control scheme for balancing condenser and reboiler heat loads .de~ationS 219 FIGURE 9.

High and low feed-flow limiters are set according to the required column turndown (ratio of maximum specified feed flow to minimum specified feed flow). This is a purification column in a solvent-recovery system.1 3 OVERRIDES FOR SIDE-DRAW COLUMNS We will conclude this chapter with a brief discussion of some of the overrides that may be encountered on side-draw columns. the column of Figure 9. see Figure 9. Usually a gain 6 override is adequate. Following are the most important override^.22. A proportional-reset controller is usually necessary here. M overrides for the steam valve are shown in Figure 9. only one controller is necessary if the two transmitters are scaled to have the same gain. We will choose.5 and reset time is usually 20-40 seconds.21).19.^ 1. and a small amount of high boilers must be removed as bottom product. On-stream analyzers would permit a better job here. as an example.20. The function here is to hold the column on total reflux until overhead and base compositions are nearly correct. Low base temperature or low top temperature holds all drawoff valves closed (Figure 9. Increasingly we are putting the limiters in the set-point channel instead of as shown. High column AI' or high base pressure closes the steam valve. A small amount of low boilers must be removed overhead.18 Hard and soft constraints . Typical gain is 0.220 Appluation of Protectzve Controls to Distillation Columns 9. Most of the feed is taken off as sidestream. Since column dynamics are about the same for either variable. A high selector chooses between them. 2. but this really depends on column composition dynamics. FIGURE 9. 3.

19 Flow rate controls for composition control .9.13 Overrules fm Si&-Draw Columm 22 1 FIGURE 9.

this is accomplished by calculating the total vapor flow from the reboiler from the stuam flow. As shown by Figure 9. it is necessary to maintain a minimum liquid flow down the column b e h the drawoff point. the override controller pinches back on the drawoff valve until the downflow becomes adequate. this is accomplished by subtracting the sidestream drawoff flow fiom the estimated internal reflux to calculate the net liquid flow down the column. The output from this controller goes to a low selector in the path between the base level controller and the sidestream drawoff valve. For columns with a sidestream uapw drawoff. As shown by Figure 9.222 Appliuztaba of Pmtectave Conmh to DirtiuatiOn Columnr 4. This signal becomes the measured variable to a proportional-reset controller whose set point is the minimum vapor flow called for by the column designer. 5.23. the override controller closes the sidestream drawoff valve just enough to force the required additional vapor up the column.24. FIGURE 9. If this flow is insufKcient.20 Feed flow system . If. and subtracting the sidestream vapor flow to get net vapor flow up the column. it is necessary to maintain a minimum vapor flow up the column d o p e the drawoff point. then. For columns with a sidestream liquid drawoff. calculated vapor flow up the column becomes less than required.

9.13 OPerriaRFfbr SA-Draw columns 223 FIGURE 9.21 L w temperature overrides for drawoff valves o .

224 Application of Protectzve Controls to Dkilhtwn Columns FIGURE 9.22 Steam valve overrides .

9.23 Override for minimum vapor fo up column lw .13 Overrides fm Side-Draw Columns 225 FIGURE 9.

226 ApplitatMn Df Protective Conh.olr to DirtiUation Colamns FIGURE 9.24 Override for minimum liquid f o down column lw .

“Designing Override and Feedforward Controls.’’ presented at American Institute of Chemical Engineers meeting. 1976). ”Experimental Evaluation of Digital Algorithms for Antireset Windup. K.. “Integral Tracking Override is Better than Output Tracking. La..^' I”TC€J. Buckley. 4. “Override Controls for Distillation column^. D. Khanderia.” Cont. and L. P. 1971). 2.. 48-51 (Aug. 3.” Cont.. 1971).. Gaina. M r 1969. and W.yy Trans.. R. S. 6. F. . lO(4): 386-394 (1971). 5. Do. Eng. and R. 82-85 (Oa.. Gila.. Buckley. 51-58 (Aug.Refews 227 REFERENCES 1. P. “New Developments in Overrides for ISA Distillation Columns. at Part 2. 1976). P r 1. Orleans. “Protective Controls for Sidestream Drawoff Columns. S. L. P. S. New a. 6365 (Feb. Cox. Buckley. S. Eng.. 1968). Des. Buckley..” I e E C Proc. J.. Luyben. P. 15: 278285 (Apr.

stability. Consequently it is increasingly feasible to measure the composition of column product streams directly and use these measurements for column control. selectivity. pressure.1 0 i Indirect Composition Measurements 10. analyzers are virtually mandatory. This chapter discusses a number of indirect composition-estimationtechniques. one in each section.1 INTRODUCTION n recent years on-line analyzers have vastly improved in sensitivity. particularly for less critical applications. or in the stripping section if bottom composition is more important. They vary from the measurement of a single temperature on a tray somewhere in the column to the measurement of a number of variables that are fed into an on-line computer that performs rigorous tray-to-tray calculations. 229 . This tray is normally located in the recntjrlng section if distillate composition is more important. the techniques to be discussed have been demonstrated only for binary and almost-binary systems. and flow. On some columns two temperatures are controlled.' But analyzers are expensive and are somewhat more trouble-prone than the simple devices used for measuring temperature. and reliability. speed of response. The selection of the location of the control tray has been the subject of numerous papers over the last 20 to 30 years.2 SINGLE-TRAY TEMPERATURE Probably the most commonly used composition-estimation technique is the measurement of a temperature on a single tray in the column. Rademaker e t d 'present a well. For close control of multicomponent systems. Consequently. 10. For the most part. considerable ingenuity has been expended on the use of these simpler instruments in various combinations to deduce compositions.

One was then subtracted from the other. not in downcomers. this works fairly well as long as boilup does not change much. However. The relationship between tray temperature and the manipulated variable can be quite nonlinear if a tray near the end of the column is used. Nor should it be so close that it is too sensitive to changes in pressure and to light or heavy nonkey components in the feed.to high-purity columns. and if the two temperature measurements are separated by a substantial number of trays. If. Thus a control tray should be selected that is as close to the end of the column as possible but not so close that it gives a highly nonlinear response. temperature probes should be installed in the active part of a tray. An increase in heat input will drive more light components up the column. if the pressure is fixed. The composite effect is that A T has a minimum at a particular boilup. However. it will vary at the other end as a function of boilup rate. For good speed of response. the temperature changes at the ends of the column are quite small in moderate. a single temperature measurement will provide a reliable p d e to composition. For binary or almostbinary distillations. the temperature at the end of the column should be controlled in a binary constant-pressure system to maintain constant product composition. As pointed out by Boyd3 an increase in boilup tends to decrease A T due to increased purity but to increase A T due to increased pressure drop. For example. 10. Theoretically. An early approach to compensating for pressure variations was to use two temperature measurements. This nonlinear response presents difficultcontroller tuning problems when conventional linear controllers are used. Even if one fixes pressure at one end of a column. boilup does change significantly. suppose we have a system in which reflux is on flow control and a temperature near the bottom of the stripping section is controlled by heat input to the reboiler. Therefore. But pressure is often not very constant. a decrease in heat input will drop light component down in the column and the control tray temperature can change very drastically.230 Indirect Cumposition Measwements organized summary of the various ideas and criteria generated over the years. one usually near one end of the column and the other at an intermediate tray. small changes in pressure or the presence of other lighter or heavier components can affect temperature much more than composition of the key component. Today it is sometimes deliberately allowed to float to minim& energy consumption (see Chapter 8). but the control tray temperature will increase only very slightly because it is essentially pure high boiler already. however. This can be very confusing to the operator . Another consideration is nonlinearity.3 DIFFERENTIAL TEMPERATURE For a binary distillation. one may encounter a nonmonotonic relationship between boilup and AT. The normal procedure is to plot the steady-state temperature profile and then select a tray that is somewhere in the region where the temperature is changing fairly rapidly from tray to tray.

For a 56-tray deisobutanizer.’ The version made by Foxboro is known as the “DVP Cell. If the profile is so high in the column that the section between the two temperature points is filled with high boiler. Efforts inevitably were made to stretch the AT technique to fit multicomponent distillations.^ In another vintage paper. Here the intent apparently was not to compensate for pressure variations.” As shown in Figure 10. Vermilion5presents a somewhat different application of dfferential temperature control. Response to pressure changes is much faster than the response to composition changes. Where applicable. reusing an instrument originally bought for another job may not be easy.4 Dfetta Vapm Pressure zmnril 231 and may cause instability in the AT control loop. a bulb filled with liquid whose composition is the same as that desired on a particular tray is installed on that tray. Some users. the AT signal will be small.10. pressure in the bulb will be the same as pressure on the tray. The AT signal will also be small if the profile is so low in the column that it is filled with low boilers. but to control the temperature profile in the middle trays of the column. Bulb filling is a job for the factory. Deviations in the tray composition from that in the bulb are then reflected by AP transmitter output signals above or below midscale. 2. however. These two conditions will give similar 4T signals but vastly different product compositions. Successful application to a CJC. some practical problems: 1.1. When the liquid on the tray has the same composition as the liquid in the bulb. It is connected to one side of a differentialpressure transmitter. whch causes erroneous readings and sometimes control problems. For that condition the AP transmitter is normally set to read midscale. This nonmonotonic relationship between AT and product composition occurs even when boilup changes do not affect Al’. Some of the d16culties encountered and some guidelines for successll application are presented by web be^. thus leading to measurement errors. Liquid of the desired composition for 6 h g the bulb may not be chemically stable over a long period of time. There are. and thereby control terminal composition. temperature was measured on trays 22 and 42 (measuring from top down). and may either flood the column or shut it down. Use of the AT permitted successful control of isobutane in the bottom product. splitter column was reported by Bonilla. this instrument is capable of great sensitivity. therefore.6 10. .4 DIFFERENTIAL VAPOR PRESSURE Another early effort to compensate for pressure variations in binary distillations resulted in the development of the differential vapor pressure cell. The other side of the AI’ transmitter is connected directly to the same tray. have installed snubbers in the pressure impulse line. 3. This is easily visualized if one considers two extreme situations.

If the reference bulb has been properly filled with a liquid of known composition. atm.232 Indirect Composition Measur-B Although this instrument is primarily intended for binary systems. in bulb xR = mol fraction low boiler in bulb PL = vapor pressure. Tivy7 has shown that with a bit of patience and cleverness. it is necessary to write and solve a few equations. then at a reference temperature T R pressure in the bulb is: PR = XRpLyLR + (1 . To determine the required span of the DVP C l in inches of water corel responding to a desired composition span. of low boiler at T R FIGURE 10.1) where PR = total pressure. atm abs. it can be used on some multicomponent systems.XR) PHYHR (10.1 DVP cell schematic .

10.2).X L ) + PHYH pHYHR] (10.3b) Since PR “washes out?’ of equation (10. a m . atm. one might well ask.~ = where L ~ L Y L + (1 .xR) PT = ~ L ~ L Y L (1 . and the need for very accurately calibrated temperature and pressure measurements is eliminated. of high boiler at TR yL = low boiler activity coefficient yH = high boiler activity coefficient Now the DVP Cell actually measures PT . a m . ‘Why do we have it?”The answer is that there is only one instrument to calibrate rather than two.3).2): PT .X L ) PHYH (10.PR and is normally calibrated in inches of water for such spans as 20-0-20 or 50-0-50.1) and (10. atm xL = mol fraction low boiler in process PL = vapor pressure.X L ) + PHYH (10. From equations (10.4 Dtfferential Vapor Pressure 233 PH = vapor pressure.2) PT = total pressure in process. We can now solve for xL: (10. the process pressure is: PI.4) (10. TR.PR = ~ which reduces to: L ~ L Y L(1 . of high boiler at TR ym = low boiler activity coefficient in bulb ym = high boiler activity coefficient in bulb At the same temperature. of low boiler at TR PH = vapor pressure.3) .[xRpLYLR + ( l . The activity coefficients may be found from: (10.3a) which is the same as equation (10.5) .

For other ways of finding activity coefficients. 10. The vapor pressures of the pure components may be found by the Antoine equation: pL -PH . then.2) we have used assumes that the P-T curve at constant composition is a straight line for small deviations. the actual temperature is compensated back to what it would be at flowsheet pressure. PT . an HP-41C program has been written that is valid if the process and bulb contain the same chemical species. pressure deviations from PFsare large. . For systems that are not strictly binary but where the nonkey components are present in either relatively constant or very small concentrations.CL = Antoine coefficients for the low boiler CH = Antoine coefficients for the high boiler A H . BE are Van Laar coefficients. A particular hardware concept (see Figure 10..5 PRESSURE-COMPENSATED TEMPERATURE To get around or away fi-om the shortcomings of the previous methods. and subtract these temperature changes fi-om the actual temperature.16 If. for xL .xR constant. .PR will vary somewhat with pressure (and. T = temperature. therefore. as we have postulated. E. the DVP Cell may ofien be used by calibrating its output against laboratory analyses. "C C H = CL = For the two-parameter Antoine equation: 273.+AL+BL/(T+CL)I (10. see the various texts on distillation and vapor-liquid equilibrium. In effect. PT . One technique is to measure pressure deviations from flowsheet pressure. PT . That is to say. this event the equations are slightly changed. When the desired composition is such that the components would react with one another over a period of lime.PR will always be zero when the process has the same composition as the bulb.7) T = eFH +EN/( + CH)] where A L BL . one may substitute another fill material (if In one can be found) that has the same PR as desired at TR.6) (10. temperature). the bulb contains the same chemical species as the process. BH. composition can be computed fi-om temperature and pressure measurements. To facilitate calculations.234 Indirect Gmpositiun Measuremcna where An. Ifthe chemical species are the same but the compositions are different. PFs. calculate temperature changes at constant composition due to these pressure deviations.PR will change slightly as PT changes. there will be some error.

c E E 3 u l m H E 2 c . 5 z1 sk 2.10. 1c e 9) 9) + E 0 0 3 r p 8 P. .5 Presure-Compmated Temperature 23 5 c .

236 Indirect CmposittiOn Measuremenn The compensator has the equation: epc = Om. ... = temperature transmitter gain (10. + Ob (10. Theory for Binary Distillation With the above equation in mind. we can see that if there is a pressure change AP from PFs that causes a temperature change AT.T- = = temperature transmitter input span transmitter output span (whether pneumatic or electronic is immaterial) (AO. = pressure transmitter output signal . ..11) (10.13) T.AP K.Kpc 0. Kmp= pressure transmitter gain )ma P- - P- P . .8) where 6 = compensator output signal ... = 0 (10.).). . AO.. where = constant composition (10.KpcAO. In other words: AOpc = AO. we want no change in Opc ifcomposition has not changed. AT K.9) or Kpc= '8.. = temperature transmitter output signal 8 Kpc= compensator gain Ob = compensator bias T h s h c t i o n is readily performed by commercial proportioning relays or some proportional-only controllers.10) (10.12) AO. .. - P - = = transmitter input span transmitter output span (AO. 8.= K.

- ar/aT KP AP m cornposiuon constam -- K~~ ar/aT 1 “i 1 FIGURE 10.3 Composition vs. t and p .

It is common practice to adjust the compensator bias so that it reads midscale when T = TFsand P = PFs.e is a measure of the pressure-compensated temperature. The operator's control station can then be calibrated directly in terms of composition.ICpc. The process operator must know what this relationship is in order to adjust the set point for the controller properly.16)2i = l n E YiXiBiPi and aT 1 aPT aP. If a computer is available. Some operators have found it confusing to have both temperature and pressure-compensated temperature. ..3 and plotted C versus T (both read at constant PFs)on a new plot. The output of the compensator. we get: - (T + -1 273.238 Indirect Cornposh M e a s u r m a fiaction. and activity coefficient for each component and is defined: n PT = i= 1 yixi pi where n is the number of components and Pi is the vapor pressure at T for the ith component.8 = 0 [see equation ( l O . In some cases the transmitted pressure signal is so noisy that a snubber may be required.16)] so n pT = i= 1 yixi c k + B i / ( T + 2 7 3 . pi = e[Az+BJ(T+273.8 . We therefore have taken Figure 10. 3 .or more generally when: ./aT Some Practical Considerations 1 . 1 6 ) ] Differentiating. l ) ] Therefore: 2. . it is possible to take a considerably more accurate approach. For electronics an RC filter may be used.

Y3P3 +zP2 . To show what is possible. high boiler Equation (10.20) P = -&yipi i= 1 (10. By knowing T we can readily calculate PL and PH. but fiom there on we must find xL by trial and error.22) y3P3 Piand y i . let us consider as an example a three-component system where the ratio of the concentration of two components is essentially constant: x1 x2 _.6) and (10.R 3 (10. it is possible to use a more sophisticated computing scheme than otherwise would be practical.6 MULTICOMPONENT COMPOSITIONS COMPUTED FROM TEMPERATURE AND PRESSURE MEASUREMENTS If an on-line digital computer is available.20): x1 P. The activity coefficients .22).10. one may find Knowing Ai.21). 10. In looking at equations (10. This is easily done on any computer.= exp A. To treat this subject in a very general way is not feasible.18). however.(1 + +) (10.16 It is desired to find x1 fiom measurements of PT and T.6 Multimnpiment Comparitimc f. one may find Pi from equation (10. We may rewrite equation (10.19) (10.5). B j .The activity coefficients may be found by equations (10.17) yL = activity coefficient of low boiler yH = activity coefficient.18) The individual vapor pressures may be found by the Antoine equations. Then knowing x1 from equation (10. see equations ( 10. the basic equation is: PT = ~ where L ~ L Y L + (1 -~ L ) ~ H Y H (10.4) and (10. taken at the same tray in the column.4).17) may be solved for xL: (10. and (lO.7).21) T + 273. (10. m Temperature and Pressure Measurements io 239 For a binary system.S). we can see that xL is really an implicit function. + ( = Y2 YlP1 p .

23) Plots of A2T versus X. the y’s would be functions of composition.03 . Luyben’s method partially fixed the profile in either end of the column. The thought has occurred to a number of people that perhaps additional temperature measurements could be used to calculate composition in multicomponent systems. they both may be regarded as improvements on simple AT measurements. but they did a show a maximum at about xB = 0. .TlO) (10. 10. He then defined: A2T = ATR .24) ATR = temperature dfference in rectification section AT. and also was nonmonotonic.) showed that it was affected very little by disturbances. The computed and measured data checked well.ATs where (10.240 Indirect Composition Measurements (yls) are sometimes fairly constant (if‘ the range of composition changes is small).02. Two papers that illustrate two Werent approaches have appeared in the literature. Therefore. Both approaches effectively cancel out the influence of boilup rate and pressure on temperature measurements. Boyd‘s technique partially fixed the composition-versus-tray profile in the vicinity of the feed tray and on both sides of it. In each case A2T was used to control top product. and are primarily usell for binary or almost-binary systems.0. He defined: A2T = (TlO - 7-15) - (7-5 . Bonilla6 reported that Boyd’s double-differential-temperature technique did not work well for a CJC. each system was calibrated by gas chromatographs. The signal was found to be dependent on feed composition. (mol fraction E. 10.7 DOUBLE-DIFFERENTIAL TEMPERATURE Strictly speaking one temperature and one pressure at a point in a column can be used to predict composition accurately only if the system is binary. Boyd’ took a somewhat different approach. and 15 (numbering up from the bottom). = temperature difference in stripping section Plant tests on two columns showed great sensitivity. Luyben’ made a computer study of a 25-tray deisobutanizer with temperature measurements on trays 5. trial-and-error solution would be required. so an iterative.10. splitter. More generally. He located two temperature measurements in the rectifying section of the column and two more in the stripping section. this might well be useful for control. Since most deisobutanizers operate in a range of xB = 0.

Control problems are frequently encountered in these columns because the process gain is very high (a small change in heat input makes a drastic change in the temperature on any single control tray)." This technique is useful for distillation columns in which a large temperature change occurs over a few trays (a sharp temperature profile). and is linear.8 AVERAGE TEMPERATURE The use of multiple temperature measurements was proposed by Grote' and independently (but much later) by Luyben. 10. Both simulation studies and experience on an industrial column have been reported. distillate composition is estimated using an equation of the form: xQ = ajlTl + aj. As the profile moves up or down. This is found in columns where the components have widely differing boiling points (peanut butter and hydrogen would be an extreme example).9 COMPOSITION ESTIMATORS Brosilow and co-worker~"-'~studied the use of several temperature and flow-rate measurements to estimate product composition. ti average temperature changes gradually. the system saturates easily: the temperature drops to the boiling point of the low boiling component as the profile drops below the control point but shows no change thereafter as the light component works down the column.10. The Brosilow estimator employs a linear combination of selected tray temperatures and steam and reflux flow rates to estimate product compositions.25) where Fs = steam flow rate Lo = external reflux flow rate The a coefficients are constants that are determined experimentally or by calculations. In addition. hs The average temperature signal is fed into a conventional feedback controller. (10.9 Composition Estimm 241 10. However. when nonlinear ihn . This technique has been successfully applied to many industrial columns. An average temperature signal is computed from these multiple temperatures (this can be inexpensively implemented in conventional pneumatic or electronic analog hardware). Shah14 had good success with the Brosilow estimator as long as the column was operated w t i a linear region. The multiple-temperature-control scheme uses a number (four or five) of temperature sensors located around the normal temperature break. The estimator uses steady-state relationships. They used the term "inferential control" for this type of composition estimation and control using secondary process measurements. handles multicomponent systems.T2 + aj3T3+ a j z s + aj5L. For example.

AIChE 1. 29. M. 1975. U. 14. L.. h 127 (Mar. M.. Rt$38(5): 187191 (May 1959).. 13. 10. R. “The Use of Process Analyzers for Composition Control of Fractionators. Brosilow.. ..351 (Nov. En. I m . et al. DynamM and contrd OfContinuOw Di&llathn Units” Elsevier. Grote. A. L. 9. B. the difference between two temperatures. Joseph. the linear estimator performs poorly. 1948). . is made and the procedure is repeated. that is. a new guess of X. Hyahfux Pnx. K. Hyd.733. and H. Boyd. E.. 3. Luyben. 1969). distillate. W. 24: 485(1978). 1976). vapor-liquid equilibrium and physical property data. 18: 238 (1972). The work of Weber and Mo~ler’~ should also be noted. W. Rijnsdorp.. Symp. Patent 2. Maarkveld. P t ”e . 18: 614 (19721.. 11. Smith. 15.S. B. eis Chem. 8: 739- (1979). and steam flow rates and one or more tray temperatures). L. R. Luyben.. R. 7. 98a 100 (Aug.6/1(1979). 744 (Nov.S. W.. 2. 1955). L.. and W. H. 24-0 (Nov. Mosler. E. U. V e d o n . and C . known process parameters (tray efficiencies and heat losses). B. 1961). REFERENCES 1. Shah14mended Bmilow‘s work. Patents 3. 12. New York.. Bondla. and the distillate-to-feed ratio to adjust manipulated variables. Joseph. Luyben. and D. J. 6.855.725. O. proposing a nonlinear composition estimator for binary systems.. J. Sre. Parsons.. CEP 71(6):55-60 (June 1975). 13(1) 8.. Suppose. AIChE J. and measured variables (feed. Pet.and one wants to calculate distillate composition x. They proposed several control schemes using the sum of two or more temperatures. Webber.” ISA Trans. poor estimates of product compositions are made. D.242 Indirect compmitimr Measwemm phenomena occur. that rdux and distrllate flow rates and a tray temperature in the rectlfjrlngsection are known. and A. 5. D. Griffin. J. Weber. Rademaker.627 (1973) and 3. A fundamental nonlinear mathematical madel of the column is used.: et al. The rigorous estimator makes a guess of xD and calculates tray-by-tray down the column u t l the tray is reached on which the temperature is measured. Webber. AIChE J. 56. 2. Shah. 4. W. W. “Automatic Control of Fractionating ~ h m n s . . Shah verified the effectiveness of the nonlinear estimator by both simulation studies and experimental tests on a pilot-scale column.. 21. (Nov. for example. E. Oil and G sJ.074 (1974). V.. no. Tivy. A. 1976). IEC Fund... involving tray-to-tray material and energy balances.. O. If ni the calculated temperature on this tray does not agree with the measured temperature..

For mixtures of components. a number of items we think are important. however. We are very suspicious of the accuracy of most inplant flow meters because only rarely do plants have facilities for their calibration. The most common method of determining internal reflux makes use of external reflux flow measurement and the number of degrees of condensate subcooling* (see Figure 11. 243 . Important inches for column separation ability are internal reflux/distillate and boilup/bottom-product ratios. has been placed on temperature-measurement dynamics since this subject seems to be widely misunderstood or inadequately understood. Considerable emphasis. IncreasL?gly we are controlling columns on the basis of internal reflux rather than external reflux.Z CALCULATION OF DISTILLATION-COLUMN INTERNAL REFLUX In the design of distillation columns. on the other hand. We have selected. A glaring omission is that of flow metering. And we have devoted some space to control-valve sizing and selection since the new (1974) ISA equations for calculating the operating C.l INTRODUCTION e make no effort to cover all measurements and hardware here.1 and reference 15): * This method works fairly well for relatively pure material. or have not been covered adequately elsewhere. or are worth emphasizing. but we have nothing new or unique to offer for improvement. 11. and for predicting the flow regime are not yet widely used. engineers use internal reflux in their calculations rather than external reflux.I1 W Miscellaneous Measurements and Controls 1 1 . it is sometimes advisable to take into account the difference between bubble point and dew point.

psig KR = summer gain. electronic. but those that use pressuredividing networks for gain should be avoided.244 M~celhmm Measurmnk and Controh (11.la) internal reflux rate. that the two temperature transmitters have the same span. Summer The Foxboro 136-1 summer has the equation: p=KR(A-C) + B where (1 1. . and a multiplier. pcu/lbm "C h = latent heat of vaporization of vapor in column. psig C = signal from external reflux (condensate) temperature transmitter.2) p A = = output pressure. Ibm/hr wR = external reflux rate. The discussion that follows. incidentally.3) t There are several satisfactory pneumatic summers on the market. could be readily extended to other pneumatic.we need a summer with adjustable gain and bias.l). however. psig signal from vapor temperature transmitter. "C at point of entry to the column To implement equation (ll. pcu/lbm To = vapor temperature. psi/psi It is assumed. Ibm/hr cp = liquid specific heat. or digital devices. and therefore the same gain: (11. A commonly used hardware arrangement is shown in Figure 11.1) where WRl = (1l.2. psig B = bias. They are not accurate enough. For the former the Foxboro 136-1 is a suitable devicet while for the latter we would use the Foxboro 556-8 multiplier. "C TR = condensate (external reflux) temperature.

2 Calculution o Dirtillation-ColumnInternal R$w f 245 FIGURE 11.1 Measurements needed for internal reflux computation FIGURE 11.11.2 Pneumatic hardware configuration for internal reflux computation .

has the same span as that of the external reflux flow transmitter (assumed linear). so equation (11.TR is some specified maximum value.Next we must have a procedure for determining B and KR.9) Let us put the signal from the external reflux flow transmitter into the B connection and the signal from the summer into the C connection. Let us a s say that we want the output span of the summer to lo be 12 psig for 0 to [KscImax.3)] (11.3)[(1 12 - S)12 + S(C .6) and ( 11. we want p = 15 psig.[KsC]. and B corresponds to [Ksc]1. .2) becomes: p=B (11.TR)maxKmr Thus by combining equations (11.6) AlS0: (11.I-. We now need to find the f and S factors.7) (A .7) and solving for KR.. Let us assume that when To .). When the two temperatures are equal-no subcooling-Ksc will equal 1. proportional to the internal reflux. we get: (11.corresponds to 12 psi. (To .00 = [KSc]-.4) = Since 0 .8) Multiplier The Foxboro 556-8 multiplier has the equation: A -3 = I ( B .T is maximum. then: B = [KxImk x 12 + 3 = x12+3 [KSC lmax [KscIm When To . A will equal C.5) (11. A . Let us hrther assume that the output of the multiplier.C>m = ( T O .C is maximum and R (11.

11) ( 11.S) x 12 + l2 S ] CKscImax (11.13) to find S: 6.1 a) 1 6.12) and ( 11.11._ x 6. we want: A .00 S + 1 = [KscImax [Kscl.17) (1 .S) x 12 + 1’ 2 [Ksclrnax ] (11.00 [ ( l .00 (11.00f Ksc = [KsclmaX Note that C .13) C = 0: Equations (11.2 Ca&uLa&n o DljtiUatMn-ColumnIntwnal Refw f 247 When the external reflux is: (WR)max 2 B .12) -_ Note that forA - (1 .00 = [K.S) 6.5 (1 .15) (11.10) and ( 1 l .la)] (11. = x 0.14) This value off may be substituted into equations ( 11.10) . x) a 2 and if the temperature difference is zero: A -3 = 6.3 = 6.00 [KscImax 6.00 x [Ksclmax [fromequation (ll.].3 = 12 since: Next if external reflux is: ( W R.S whence S = l +s (11.18) .S) x 12 12 = + 12S] (1 1.00 psi If at the same time the temperature difference is maximum.3 = 6. l l a ) may now be solved to findf (11.16) ( 11.00 [Ksc]12 = 6.

TR = 40°C p = = 0.TR).000 pph To .22 psig From equation ( 1 1Ab) : 12 x KR = 1- 0.22 = 13.348 From equation (11.248 MirceUaneow Measurements and Controls Exampple.302. S = 1.302 = 9.00 (10.18).000 pph Temperature transmitter spans = 0-150°C tP = 0. and fiom equation (11.33psig =c Then: A .00 1.302 150 = 0.302 = f From equation ( 1 1.725 28.3 = .000 12 = . (To .= 100°C A = 235 pcu/lbm From these: [Ksc]max 1 + .1'302 x 6.x 100 = Oo71 235 = 1.14).5): B = .f As a check let: WR = 1.l2 + 3.348 x 40 l2 + 12.71 235 15 4 1.1136 + 12.22 150 [fiomequation (11.33) 12 = 6.022 pph Therefore: WRl = -x 50.71 pcu/lbm "C.2)l 1. Let us consider a case where the following conditions apply: External reflux flow-meter span = 0-50. = 25.22 + 3.00 = 12.

20) = k2 dAbo (11. used in designing orifice orifice reference temperature.11. psig. of flowing stream where p is measured p.21) . psi = ps = T.000(1 = + 235 x 40) 28.3 Temperature and Pressure CmpensattiOn o Gas F h Meten f 249 Check via equation ( 11. inches W. = temperature transmitter span. "C For a particular value of Ab. It should be measured upstream if orifice reference pressure pc is taken upstream.C. = orifice reference pressure. psia. "C. p = = orifice pressure drop. and pc: (11. used in designing orifice pressure transmitter span. 60°C orifice flow constant flow.7 psia. and downstream if pc is taken downstream.19) where Ab.3 TEMPERATURE AND PRESSURE COMPENSATlON OF GAS FLOW METERS The basic gas-flow metering equation is: (11. measured at T./pll = ratio of gas density to that of air at 14. "C.1): wR1 = 25. scfm or pph SC = gas supercompressibility kl q Let us also define: = = pc T. pressure.021 pph 11. T = temperature.

"C = . + 273 (11. and T # T. A common method of mechanizing equation (11.. TI 45-3a of November 1969. "C (Tm)-= minimum measured temperature.3) f and Foxboro defines for this application: (Tm)max B .22).. we can divide equation (11.26) (11. so the following will apply to Figure 11.21) into equation (11.3) 122 + S(C . 1966).For the same Aho but p to get: # pc and T # T.23) with pneumatic instruments is given in Figure 11.20) (1 1.3.24) = + 273 T.7) As can be seen.22) Substituting equation (11. "C (Tm)- - (~m)Inirl = ATS temperature transmitter span. We have been unable to find in them any real advantage over the arrangement of Figure 11.+ 14. Temperature Compensation The usual divider form of the 556 is: (A .3.27) where (Tm)m= = maximum measured temperature.23) (T+ 273)(p. we obtain: = k2 Jm(11.23) provides a convenient way of determining true flow whenp # p . A more recent Foxboro bulletin.12 x (11.3 = . features two alternate arrangements.3 (see Foxboro TI45-3a of October 14.25) ( 11. equation (11.

3 Temperrnre and Pressure Cornpenration o G s Flow Metm f a 251 FIGURE 11.3 'Qpical compensated gas flow metering scheme .11.

X (B ..q-.3) f 12 X [122 +S(C . receiving the temperature transmitter signal. Some engineers omit temperature compensation of steam flow on the grounds that it changes more slowly than pressure. Pressure Compensation The multiplier form of the 556 is: A -3 = . . (11. takes care of span suppression and converts both T and T. and (11.8 On substituting equations (11. The wisdom of this is questionable. e. we get: B . ( ) (Pm)= = maximum value of transmitter span. transmitter signal.32) (11. This calculation.33) where p .. psig .26).25). [(Tm)where + 2731 + ATs .27) into equation (11.29) is equal to T + 273.3) (11.8 . and C is the port . + 273) x 12 (Tm)max + 273 - (T.24).29) 1 2 3.8 = = signal from Abotransmitter signal from temperature transmitter It can be seen that the denominator of equation (11.252 MhceUaneaa Measurements and Cont~oi3 A is the port receiving the Ah.. therefore. + 273)(8. to absolute units as required.3)] (11. psig minimum value of transmitter span.30) and for this application Foxboro defines: (11.31) (11.3 = (T.

one must be careful to stay within available ranges f o r f a n d span S .27) into equation (11. therefore. Errors Incurred by Uncompensated Gas Flows For estimates of flow-metering errors that may occur when compensation is not made for temperature and pressure deviations fiom reference values.7 + 14.32). + 273)(p + 14. we follow the multiplier with a square root exrractor to get a signal proporrional to flow that is corrected for pressure deviations fiom pc and for temperature deviations from T.34) + (pm)maxAps 14. (11. + 14.1I .3 = (Pm)max 12 [Ip. + 14. and C is the port where 8 the pressure transmitter signal . Substituting for B .7) (11.7)(B .3 from equation (11.7) (T + 273)(p. + 14.3 is now a measure of q2. takes care of span suppression and converts both p andp. Substituting equation (11. (11.29). + where .36) 1 signal from pressure transmitter Clearing.31)..3. we obtain: But: so: p A - = (p.30): A . to absolute Units. we get: A - 3 = (e.71 + 14-7 (Pm)max + 14-7 ] (11.+.7(8. This sometimes requires juggling transmitter spans. we . 8 = - 3) 1 (11.37) This procedure.3) (T. and (11. enters.3 Tempmature and Pressure CunzpmattiOn o Gas F h Meters f (pm)max 253 - @m)min = = 4 5 pressure transmitter span. psi B is the port where the signal from the divider enters.38) Since A . In speclfjrlng Foxboro 556-8's and 556-93.

45) 100 Percent error = 2@ + 14. Exmpk 1 Operating temperature = 397°F Operating pressure = 215 psia 1500 = -78 pph (-25) 2 x 240 = error in steam flow-meter reading -78 x 100 = -5.7) x 4 (11.46) Consider.42) (11. (Aq) = .aT - -AT a9 (11.40) (11. J .43) - 2@ + 14.may take the partial derivative of equation ( 11.7) -1 (11. for example.19) with respect to temperature and pressure.2 percent error 1500 .39) Aho@ + 14.' x (PJlPU ISc aT - (-i) 1 x (T + 273)3/2 (11.41) 2(T + 273) 2(T + 273) -1 Percent error = Next: -loo x AT ( 11. the following: Steam flow = 1500 pph Design temperature = 397°F Design pressure = 240 psia Ahois assumed to be constant.7) * = k .

4 HEAT-FLOW COMPUTATIONS As mentioned i Chapters 3 and 4. for example. We wish to know the increase or decrease of enthalpy FlGURE 11.W-Fh Computationr 255 &a#tple 2 Operating temperature = 437°F Operating pressure = 240 psia -1500 2 x (397 + 460) x 4o = -35 pph = error i steam n flow-meter readmg -35 x 100 = -2.4. reflux rate.3 percent error 1500 11. it is sometimes desirable to calculate n heat flow.4 Heat flow computer for heat transfer . Let us consider. or boilup rate.11.4 Ht. the heat exchanger of Figure 11.

we may calculate the rate of heat transfer: 4 Let: = J W p (To - Ti) (11.To)is the temperature drop. In this case tpB is the coolant flow and (To . wBU: (11. For the above applications.T.= maximum flow-meter span. that is. is Similarly. A common method of measuring To . for reboilers heated with hot oil. . If we h o w the flow rate wB and the two temperatures Ti and To. Then: (1 1. liquid level. we will want to know and control the rate of boilup. I . it is immaterial whether one puts the flow signal into the B port and the AT signal into the C port of the Foxboro 556-8. For a condenser we would probably like to know wc. Ibm/hr (To . "C For the Foxboro 556-8. Of these column-base level is probably the one where difficulties are .49) where (fp.. or vice versa. lbm/hr condensed.. column AI'.Ti)span = ATspan.256 M&4!aneow Measurements and Cont~oh of stream B. = span for estimated lbm/hr condensed pcu/lbm h = latent heat of condensing vapor.50) In this case wB is the hot-oil flow and (T.48) In many cases we are not as interested in heat transferred per hour.Ti) its temperature rise. and specific gravity. p / h r = (wB). is to connect two resistance thermometers differentially.47) qrneas heat-transfer span. it readily may be shown that: (11.).S COLUMN-BASE LEVEL MEASUREMENT' I Introduction Many problems have been encountered in making head measurements in and around distillation columns. The output of the converter or relay is then a measure of To .T.

density. Heat must be transferred. liquid must be partially vaporized.11.2 ft lb mass/sec2 lb force liquid phase density. lb force/fi?. Ib mass/ft3 H AHN = = QL AP" = B ( p L ~ J H AH^ - 9c 9c ' I FIGURE 11.)H + 84 pvA H N B C (11. and liquid and vapor must be properly separated.5 Column-Base Lepel Measurement 257 most often encountered.51) APH = head. between two nozzles local gravity acceleration. we will assume that process design is adequate. Ib mass/ft3 liquid-vapor interface elevation above bottom nozzle. These difficulties are at least partially process oriented in the sense that a column base with associated reboiler is often very complex. It is the subject of this section. We will be concerned only with problems associated with determining level by means of head measurements. feet nozzle-to-nozzle spacing.5 Head-level relationship in a v e s s e l . ft/sec2 flL = gc = mass-force conversion factor = 32. Regardless of the type of transmitter used-AP or displacer-the basic equation relating liquid level. feet pL = pp = vapor phase density.5): APH where =B -(pL L 5 c - p. and head is (see Figure 11. In the discussion that follows.

52) B C With the basic mathematics of level measurement in hand. Then: f = 0. but this is not necesarily valid at pressures much above atmospheric.50 cps.’ This method says that the velocity of surface waves in a shallow environment is : (11. however.54) where f = cycles per second (cps) L = tower base radius. consider a tower 17 feet in diameter with an average liquid depth of 5 feet.. however. 1. . then standing waves with the center as origin will exist at frequencies of: ( 11. If.75 cps. feet As an example. If this is done properly. = surface wave velocity.53) where V. however. nonlinear transmitter operation.. let us next turn our attention to the problem of noise. This is mostly attributable to turbulence. fi/sec gravitational constant. feet flL = R = If we consider the column base to be a circular pool. p. such filtering or damping greatly reduces the probability of saturation in the transmitter and minimizes output signal errors due to transient. fi/sec2 average liquid depth. equation (11.. Regardless of its source. it is best to filter noise out at the transmitter input rather than at its output.51) reduces to: -pLH (1 1. is commonly assumed to be negligible. etc.25 cps. is very small. but may also be due to the type of transmitter used. 2. In the case of pneumatics.258 Mimllanem Measuremen& and Controls The vapor density. at predicting the lowest frequency of standing waves by a method suggested in part by Binder. Distillation-column base-level measuremen= tend to be extremely noisy.” = APH liL Wave Noise in Large Columns Predicting the amplitude and frequency of wave noise generally is very difficult. p. such filtering m i n i m i z e s demands on the air supply and therefore m i n i m i z e s the probability of “gulping. We can make a stab.

75 cps.6).11.55) L = length of piping from base to reboiler.5 Column-Base Lmel Measurement 259 Noise amplitude is usually greater at low frequencies.\/Lcps where 1. in this case. Roughly this is: -fL = .3 (11. It has the natural frequency of the “manometer” between the column base and the reboiler (Figure 11. feet FIGURE 11. Therefore. any filtering or damping that we design should provide sigdicant attenuation for frequencies as low as 0.6 Column base-reboiler manometer . Column-Base-Reboiler “Manometer” Another source of noise is sometimes encountered in columns with thermosyphon reboilers.

The incomplete information available indicates that these have a higher natural frequency and more damping than do the nonforce balance type.3 CPS The amplitude of this cycle may be arbitrarily attenuated by a properly sized restricting orifice in the line.7 Schematic diagram of displacement-type level transmitter . The use of adjustable hand valves is highly undesirable.7) constitute an open-loop. The basic mathematics are given elsewhere (pages 187-191 of reference 2).3 = 0. The displacer and torque tube (Figure 11. displacer-type transmitters on the market. in many cases "fiddling" with such valves has seriously delayed column startup or has caused premature shutdowns. for 20 feet of piping: f = M z 1.260 Mhceh~w Mearurements and Controh As an example. which A I R SUPPLY FIGURE 11. mass-spring system that is usually highly underdamped. Displacer-type Transmitter Resonance A third source of noise-ofien the most serious-is attributable to the basic design of most displacer-type transmitters. or by snugly sized piping. let us consider an internal damping chamber for use with either AI' or displacer-type transmitters. There are now several force-balance. for our immediate purposes it is sufficient to note that most commercial transmitters of this type have a natural frequency in the range of 1-3 cps. Internal Damping Chamber As a first approach to noise filtering or damping. As a consequence any transient disturbance causes the displacer to bounce up and down at a fairly well-determined frequency.

8.8 Internal damping chamber . this chamber has two key features: (1) a hood or cap to keep out liquid from the last downcomer or the reboiler return nozzle.5 Column-Base Level Measurement 261 functions as a first-order lag. Here a restriction and volume pot are installed in each impulse line to ensure that the inputs to the transmitter are dynamically equalized. The combination of orifice and chamber gives about 1 O : l damping (attenuation) at about 0. Useful for level measurements.0-1. FIGURE 11. this design is almost mandatory for specific gravity measurements via AP for slurries. a second l-inch hole may be cut in the lower section of the chamber. and (2) a l-inch-diameter hole in line with the lower level nozzle to facilitate rod out.11. a pot with 12 in3 and a Taylor snubber (58S104) will give 1 O : l attenuation at about 2 cps. one may provide damping as shown in Figure 11.6 cps. It is probably apparent that almost any desired damping may be obtained by the proper choice of cross-sectional area in the chamber and of orifice-flow cross-sectional area. If there is a concern about solids collecting in the chamber.9.5-0. Many combinations of volume pots and restrictors are possible. External Damping for A P Transmitters When a AP transmitter is used for level measurement and is installed at an elevation above the upper level tap and equipped with purges. The damping will then be 1 O : l at 1. as an example. It may also be used with gas or steam flow-meter installations with self-draining impulse lines.2 cps. As shown by Figure 11.

FIGURE 11. particularly in column bases.. if too large a time constant is chosen.7 second.9 External damping for AP level measurement . it may interfere with control. Its dynamics approximate those of a first-order lag with T = 0.e. band rejection filters) for the dominant frequencies. filter approach may not always be adequate. The damping affects both input signals and the feedback. first-order. Apparently many level applications.262 Miccelhneow Measuremena and Controls The simple.” Experience to date (10-15 years and several hundred applications) has been very favorable on flow-control applications. A P Transmitters with High-Viscosity Fill At least one commercial AI’ transmitter is available with a high-viscosity liquid fiU. and a bit less favorable on level. This instrument has the desirable feature of nonadjustable damping and so is “fiddleproof. In fact. and thus minimizes pilot saturation and output errors. A more sophisticated approach is to determine experimentally the noise-frequency spectrum and to design notch filters (i. need additional damping.

vapor condensations may occur in the upper line. in the range 1-3 cps.10. most &placer-type transmitters have a sharp resonance “ot”. This is discussed in a very interesting paper by sander^. changing the damping while the process is in operation may have a serious effect on control-loop stability. This feature therefore is not recommended. the purge rate must be fixed and the liquid level elevation in the chamber known. This provides a fiddleproof design. The chamber and upper line should be insulated and heated. level is sometimes measured with the nozzles located on vessels or pipelines of different diameters. if it is desired to have a liquid purge and damping. send a fixed amount of condensate via a metering orifice or capillary to the measurement chamber.^ 2. One could deliberately use the upper connecting line as a condenser. the guard valves should be left wide open in normal operation and need be closed only for maintenance work. and overflow the remainder back to the column. fixed damping provided where required. Where such instruments are bought on a project. both internal and external damping may be used. downflow is occasionally encountered on distillation column bases or vaporizers with the upper level tap on the column base and the lower tap on the bottom-product line. As discussed later. Two cases are of interest: downflow and upflow. For typical displacer housings ( b o s ) we have calculated that 1O:l attenuation at 2 cps will be obtained by installing a 1-inch-diameter orifice in the lower line connecting the displacer housing to the process vessel. Downflow By far the most common case. But an excessive amount of damping can have an adverse effect on control. the variable damping feature should be left wide open and external properly designed. This may take place fast enough to cause a sipficantly higher level in the chamber than in the column base. It is pertinent at this point to mention certain potential problems with displacer-type transmitters: 1.5 Column-Base Level Measurwnt 263 Many AI’ transmitters on the market have adjustable dampmg in the feedback. For extremely noisy installations. This leaves adjustment up to the instrument mechanic. in difficult cases a small gas purge to the upper line is helpll. Damping for Externally Mounted Displacer Transmitters As mentioned earlier.11. If the liquid velocity V in . Level Measurement Errors Due to Velocity Effect As shown in Figure 11. provided the time constants are selected with some care. 3. the natural fiequency of the “manometer” may be low enough to interfere with level control. If excessively long connecting lines are provided between the vessel and boot. who seldom has a rational basis for setting it. If a restricting orifice is installed as suggested in the lower line connecting the level measuring chamber to the column base (or if an internal damping chamber is used).

= .9 fi/sec. then the error he in inches of process liquid is: he = -0. a given column may have a different bottom specific gravity as a result of changes in bottom-product specifications.264 Mkcellaneous Measuremma and Controk the bottom-product h e is in feet per second. Specific Gravity Compensation for Level Measurements Occasionally we find enough difference between column-feed specific gravity and normal bottom-level specific gravity that there is a serious base-level measurement error at startup.1 inch. upflow In this case there is no velocity error. FIGURE 11. h. at times. In other words. the transmitter output signal wiU indicate a level that is 1 inch lower than the true level.10 Velocity error in head measurement . Also.28 Vz When V = 1.

52) can be rearranged: H= APH (BLIBC 1PL ( 11. a variable density froth exists above the tube bundle. but the one shown in Figure 11.0 feet maximum specific gravity . the level transmitter signal is divided by the specific gravity transmitter signal to give a corrected level signal.3 0.11 Specific gravity compensation o head measurement of liquid level f .3 There are many possible ways of compensating for variable specific gravity.11. if the column has an internal reboiler. consider the following: AHT = level transmitter span.05 = value at which level transmitter is FIGURE 11. As an example.5 Column-Base Level Measurement 265 Further. feet of process fluid = 4. note that equation (11.11was developed for a specific application. a = = U& 1.56) In accordance with this equation.8 calibrated = 5 = 1. For an understanding of this scheme.

) is obtained for sipficantly less than full nozzle spacing. Alternatively the output of a standard transmitter may be compensated by a microprocessor or computer._ 1.0. it is desirable to calibrate the level transmitter so that a full-scale output (3-15 psig. Sealing. Low-velocity laminar flow is usually adequate for sealing.8 1.0. but as we shall see later. 2.615 0.9 . a Foxboro 556 relay is used as a divider. cubic feet. . which may be required where solids in the process liquid tend to plug the impulse line. Characterized Displacers Sometimes it is necessary to operate a distillation column with the liquid level down in the bottom dish to minimize total liquid holdup. 4-10 mA. For these applications gas purges are usually used. S. This usually requires a vastly higher purge flow rate than is common practice. and Z are terms used by Foxboro for adjustable relay parameters. Now let us look at some sources of level measurement error due to purge systems. gallons. Flushing or sweeping. purges are intended to accomplish either or both of two functions: 1. it is common to locate it above the upper level nozzle and to purge at least the lower impulse line. But for liquid or gas. Now for level control purposes we want a transmitter whose output signal is proportional not to liquid height. we obtain: 1. nonplugging) process liquid. etc. This permits f l ul transmitter output if the process specific gravity is somewhat less than design. which is simply the isolation of the transmitter f?om a clean (no solids. A common practice is to zero the transmitter at or slightly above the bottom level nozzle and to calibrate for full output at 90 percent of nozzle spacing.3 1.If. See also recommendations in Chapter 4. but to inventory (pounds. Purge System Errors When a AT' transmitter is used for measuring base level.). for example.= 0. Generally speaking. turbulent flow through the exit diptube is recommended. and often both impulse lines. In the absence of specific gravity compensation.3 1.3 .3 Z = .385 .5 = 0. This situation usually arises because of thermal degradation. To accomplish this some vendors can furnish displacers with an eltiptical top-to-bottom shape.3 wheref. may contribute to dynamic measurement errors. Gas purges ofien are not adequate.24 f =-z S = 1. etc.

Inadequate Purge Flow If purge rate is low enough. one must be carell to ensure that both purge systems are equalized-that is. This has been particularly troublesome with vacuum columns where the change in volumetric purge flow rate from startup conditions to normal operating conditions may be 100:1 or greater. Again the error disappears as the purge flow displaces the liquid. that they have the same flow rate and same length of same diameter impulse or purge line. a rapid increase in liquid level will back liquid up into the lower impulse line.5 Column-Base Level Measurement 267 Unequal Purge Flows When the level transmitter span is small.13). Failure to accomplish this equalization may result in an offset in transmitter reading that will not be detected by conventional calibration procedures. This consideration is particularly critical for specific gravity measurement by AI'.11. transmitter lag in response to rapid rise in level .12 Insufficient purge. A rapid increase in vessel pressure can cause a different kind of error if purge rate is low enough. Static Pressure Variations Static pressure variations on the process side cause changes in the true volumetric flow rate of gas purges.12). It is probably apparent that either of these phenomena can play havoc with a level control system. The result FIGURE 11. In this case liquid is backed up into the lower impulse line and the transmitter indicates a false low level (Figure 11. The transmitter will only gradually reflect the level change as the purge flow slowly dsplaces the liquid (Figure 11.

The rotameter pressure. Upstream gas pressure is high enough that critical drop exists across the needle valve. This design has a number of sources of error. the rotameter does not indicate true flow since it reads correctly at only one temperature and pressure.14. FIGURE 11. The purge is then connected to the transmitter impulse line close to the transmitter.7 psia. Consequently with this design true flow is rarely known since it requires determining correct rotameter pressure and temperature and substituting both into a correction formula along with indicated flow. which is followed by a rotameter. however.13 Insufficient purge. will ride up and down with column-basepressure. level transmitter erroneous response to rapid rise in pressure . usually 70°F and 14. This is rarely measured or calculated. Consequently it is recommended that gas purges not be used where process pressure is 6 psia or less unless the system is carefully designed and is calibrated after startup. the fact that the purge is connected close to the transmitter rather than to the process means that there may be an error due to impulse-line pressure drop. and most common calibration procedures do not allow for it. Second. and may even load the vacuum jets. First.268 Midlaneow Measurements and Controls may be excessive pressure drop in the impulse lines. Installation Errors Typical gas-flow purges are arranged as shown in Figure 11.

the required C. An even better system. A far better gas-flow purge system is shown in Figure 11. capillary restrictor that has been designed to provide a specific.16. This is probably a good point at which to comment on horizontal versus angled nozzles. the restrictor usually chosen is a needle valve whose annular clearance changes with ambient temperature. supply at 50 psig.11. known flow for a particular application. is shown in Figure 11. this system does not require critical flow. the purge joins the impulse line close to the process rather than close to the transmitter.15.0-scfh air flow. particularly for specific gravity applications. Consider. is about 0. for example. Here the rotameter is upstream of a fixed.14 Typical gas flow purge system . a needle valve with an orifice 1/8 inch in diameter. particularly when plug and seat are made of different materials. For most applications the horizontal design is adequate and is cheaper.004. Further. For 1. The angled design is believed by some engineers to offer advantages FIGURE 11.0002 inch.5 Column-Base Level Measurement 269 Third. In addition to being more reproducible. The plug-seat clearance is only about 0. and critical flow. Here the fixed capillary restrictor is used in conjunction with a purge-flow regulator.

1s Improved gas flow purge system FIGURE 1 I .I6 Best gas flow purge system .270 Miscellunwus Memurements and Gmmk FIGURE 11.

17. some years ago we had a problem where our fluid mechanics experts recommended a purge velocity of 4 fi/sec through the damping orifice at the bottom of the displacer housing to defeat plugging. FIGURE 11. particularly if a bottom connection to the measurement chamber is used. if any damping is used.0 inches higher than it would have been with no purge. a purge-flow regulator is mandatory. the calibration should allow for the extra head due to the purge flow.55 inch. When used with displacer-type instruments. This is usually omitted with horizontal nozzles. The level in the housing was then 9. Liquid Purges Liquid purges are sometimes used with AP transmitters for difKdt applications where there is otherwise a tendency to plug the impulse lines. In addition. and the calibration procedure should allow for impulse-line pressure drop. With angled nozzles it is common practice to use a diptube as shown in Figure 11.5 Column-Base Level Measurement 271 when solids are present. Such installations should be designed to avoid any pockets in which solids can accumulate.8.11. The orifice was calculated to have a diameter of 0. We chose 3. As an example.0 gpm of a liquid with a specific gravity of 0. constant flow. Liquid purges are also sometimes used with displacer-typetransmitters. For accurate. either an external orifice or an internal chamber. it reduces the available level transmitter span.17 Angled nozzle with dip tube .

18 Level measurement with AP transmitter with double remote seals . AP Transmitter with Double Remote Seals (Figure 1 1. some plants have experienced sufKcient leakage with closed guard valves that removal of a level transmitter during operation is not permitted.1 9) Some vendors make AI' transmitters that provide sealing only at the highpressure connection with a so-called flush diaphragm. In addition. of course. For base level measurements. one must use so-called extended diaphragms that are flush with the vessel interior wall. inappropriate. Sometimes such an instrument with the low-pressure connection connected to the upper level nozzle FIGURE 11. Many processes are sufficiently hazardous that mechanics are not allowed in the area except during shutdown. For severe plugging applications such chambers are. internal damping chambers should be used if sealing only is required. Two Flush Diaphragm Transmitters (Figure 1 1. In such applications one cannot use guard or block valves.1 8) Several transmitters of this type are now on the market.272 M&Aluneow Measurements and Conwok Various designs for systems with unusually difficult plugging problems are discussed in a paper by Schnelle and S~hmoyer. This is not necessarily a serious objection. If solids are present. They have the advantage of providing sealing without the use of purges.~ Other Head-Measurement Techniques Other head-measurement techniques include the following.

Their low-pressure connections are vented to atmosphere and the outputs subtracted in a summing relay.20) A variation of the preceding uses a flush diaphragm on the lower level nozzle and a 1:1 repeater on the upper level nozzle. For severe applications. however. flange or screwed connections.11.). or other? -Body size and trim size. such as linear or equal percentage. The output of the repeater goes to the low-pressure connection of the flush diaphragm repeater. FIGURE 11. highgain positioner provides better performance in both respects than does a spring and diaphragm actuator with or without a positioner. The second is usually specified in terms of the flow coefficient./(X. (C. A valve must have adequate threshold sensitivitv and speed of response. The first refers to the pipe size to which the valve end fittings. Flush Diaphragm Transmitter and 1 :1 Repeater (Figure 1 1. two-stage. to be made about each valve: -Valve type. fractional Ilft. -Actuator and positioner. andX.. -Inherent flow characteristic. as well as a host of minor ones. What type should it be-globe.)N. where the vapor tends to condense and crystallize or polymerize at the upper tap. A piston actuator with a double-acting.6 Control Valves 273 is adequate.for the wide-open valve. two flush diaphragm transmitters are sometimes used./( C. ball. or stroke. will mate. This is the relationship between C.6 CONTROL VALVES Control valves play a very important role in continuous processes. 11. travel..). butterfly.19 Level measurement with two flush diaphragm transmitters and a summing relay . This is especially true for disullation columns. Typically there are four major decisions.

In pumped systems there is a trend toward minimizing valve pressure drop to reduce pumping ~ 0 s t s . including those with anticavitation or low-noise features. the engineer must recogme three s i p f i a n t rig factors that bear on control-valve selection: 1.&' 3.14 In t y n to make these decisions.274 M i s c e U u m Measurementr and Controt5 Other important but non-control-related decisions include those of selecting mechanical and metallurgical features relative to corrosion. pressure. ti size. and fire safety.. that are much more accurate than the old FCI (Fluid Controls Institute) equations. have a greater tendency to cavitate in liquid service. as compared with globe valves. lower price. 2. l ~ General Approach A general approach to choosing body size. There are newer methods of calculating the normal or operating C. There are today many new types of valves. and valve type has rm been developed that is based primarily on equations for C. There are many disk and other rotary valves with much larger C.. or are noisier in gas service. temperature.(flow coefficients). however. developed by an FIGURE 11. and greatly reduced weight.20 Level measurement with flush diaphragm AP transmmer and 1:l repeater . Some of the newer designs.

of wide-open valve) for the chosen valve. They also take into account three factors that the older equations do not: 1. incipient cavitation 3. To the ISA equations we have added for our own use an equation for incipient cavitation. If the valve service at original startup is to be water or some fluid other than normal process fluid.. required for the flow rate and other conditions specified 2. This even permits calculating the performance of valves larger than line size. no cavitation 2. a noise check almost certainly would be run to find out whether the proposed installation will conform with OSHA specifications. The influence of associated reducers and expanders. Meeting this criterion results in a valve whose travel is 60 percent or less at flowsheet conditions. . choose (C. For gases we have subcritical and critical flow regimes. Neither are there any generally accepted guidelines for specifjmg operating C. Subcritical flow. With the calculated operating C.Some engineers. If the program shows that the first valve choice for a liquid service will be subject to cavitation. full cavitation 4 Critical flow. 3. all of the previously mentioned procedures must be repeated. calculation purposes. type of valve. At this point valve sizing. for example. some rather exotic designs (with corresponding prices) are available... the engineer compares it with the (C. There are no equations for calculating which size valve to purchase or install. Many vendors now offer anticavitation trim. flashing . various low-noise valve designs are available. All of this information has been combined into calculator and computer programs. The engineer first makes a trial choice of vendor. ceases to be a science and becomes an art. If it will not.11. For gas service. as typically practiced..6 These equations calculate the normal or operating C.)N(maximum C. Expected flow regimes For liquids we have four possible flow regimes: 1. Valve-body geometry via coefficient FL. the engineer may choose one and rerun the program. and for extreme cases. Subcritical flow. for 50 :1 equal-percentage valves.)N. in hand.)NI5 x C. not only for C. 2. if the program predicts critical flow for the engineer's first choice. The computer or calculator program then reads out two pieces of information: 1. and body and trim size. the engineer very likely will want to look at another valve type and rerun the program.6 Control Valves 275 ISA Standards Committee. for the specified flow conditions. but also for specification of valve operator type and size. Critical flow. as a fraction of (C. Operating C. Upstream liquid-vapor pressure.

These curves may also be generated by the "calculated variable" function of computers and of some distributed control systems. such as flow.7 [C. At this point it is appropriate to note that the increased complexity of the new C. A suggested procedure is to ensure that the operating C. for control purposes one must be able to throttle more than flowsheet maximum flow and less than flowsheet minimum flow. How does one determine maximum required flow?Under transient conditions a valve must be able to pass more than flowsheet maximum.. In other cases we have installed computing relays in the control loops to generate desired curves. But how much more? There is no simple answer to these questions. and pressure-control systems.]. Maximum Flow and Turndown A control valve must be able to pass a specified maximum flow and should preferably. piping and equipment pressure drops. calculated for flowsheet maximum flow is: 1. The slope of this curve is the valve gain that must be known for quantitative design of control loops. But whether one valve is used or two. This is 91 percent lift for a 50: 1 equal-percentage valve. for valves with full-size trim for which reduced trim is not available." But the older FCI procedure offers no flags to indicate when the new procedure is required. We have calculator and computer programs to facilitate these calculations. liquid-level. If this cannot be accomplished with one valve. Next one must look at a particular system-pump curve. the mathematics is simple and the results have been tabulated. flow-ratio. or nearly so (splitrange arrangement).5 [C. for globe valves with one-size-reduced trim. calculation procedure has led some engineers to refuse to use it except "when needed. pass the minimum required flow without getting too close to the closed position. And soon dedicated microprocessors probably will be available. is known. For some commonly encountered systems. This is 82 percent lift for a 50: 1 equal-percentage valve or 65-70' open for a typical 90' butterfly valve. but not necessarily.]. No greater than 0. The criterion we prefer is that of choosing the installed characteristic that minimizes or eliminates the need for retuning the controller as throughput changes. Here 6. No greater than 0.276 Miccellaneow Measurements and Controls The best inherent flow characteristic for a given application cannot really be chosen until the desired installed flow characteristic. Q or w versus 6. Valves sized according to either of these criteria will be less oversized than if more common sizing procedures are used (see Chapters 3 and 4). In some cases we can help matters with a cam-operatedpositioner. a small valve that opens first may be used in parallel with a larger valve that does not start to open until the small one is wide open. is the signal to the valve positioner. and so on-and see what inherent flow characteristic comes closest to providing the desired installed characteristic. . 2.

The installed flow characteristic also becomes verv Merent fiom the lnherent flow characteristic. all curves. and around distillation columns. This pushes us toward line-sized valves. System with Pumped Liquid For pumped systems rapidly increasing energy costs have aroused considerable interest in designing for very low valve-pressure drops. valve turndown suffers (becomes smaller). it is desirable that valve lift be no more than 95 percent for and no less than 10 percent for (&). valve AI’ is almost never a factor in quality of control. a liquid-level control svstem manipulates an outlet valve whose maximum capacity is 100 gpm. while an equal-percentage valve tends toward an installed linear characteristic. If. If. low-holdup chemical reactors constitute the only applications we have found to date that require high valve pressure drops. But regardless of what flow criterion is used. A suggested typical process turndown figure for valve sizing and selection is therefore five to one. however. If piping and equipment pressure drops are large in comparison with valve-pressure drops. . A few high-speed.6 Control Valves 277 Determining minimum required flow is even more difficult. valves should be so sized that maximum manipulated flows will be greater than maximum disturbance flows. from. As an alternative to the above. note that in a chain of process equipment. level off at a maximum flow.25 and (&) 5 0.2 but this effect is so small for most applications that it may be neglected. often require very low flows for awhile.and downstream pressures. it is suggested that for valve-sizing purposes we let P 1. A linear valve tends toward an installed square-root flow characteristic. for a given application. Fortunately most wet chemical processes seldom have a turndown (maximum required flow/minimum required flow) greater than two or three to one for nmmal operation. It has been shown mathematicallv that higher valve AP’s do provide better control.)Fs (aax) Turndown and Inherent Flow Characteristics as Related to Process Applications A. two or more split-ranged valves may be required. small valve AP’s mean less tendency to flashing or cavitation.11. One occasionallv finds rules of thumb to the effect that valve AP should be at least one third or one half of system pressure drop for good control. The use of small valve pressure drops makes it more dfficult to get a desired installed flow characteristic than if larger valve pressure drops were used. Valve-flow turndown is defined here as the ratio of flow (not C. regardless of valve inherent characteristic. Finally. it is probably desirable to provide at least 5 psid for the valve at flowsheet maximum flow rate. the tak will overflow if inlet flow goes over 100 gpm. (aax)(a. Startups and shutdowns. In view of the difficulties of accurately predicting up.) at 95 percent lift to that at 10 percent lift. for example. For flow to.75 (&)Fs. however. high-lift end of the plot. This is so because the valve has run out of pressure drop. On the positive side. calculated valve-flow turndown is less than process turndown. however. At the high-flow.

A valve may be noncavitating at flowsheet maximum but cavitating severely at 50-60 percent of flowsheet maximum. in many cases. in most cases. Note that these valves f'requently must contend with cavitation or flashing. at high flows and lowest AP. C. B. In many applications. Studies have shown that cooling-water turndown requirements are usually much higher than process turndown. split-ranged valves should be used. Line pressure drop is usually negligible. D. we have found that the distortion is moderate. at low flows. valve pressure drop increases with flow since the heat exchanger has a higher condensing temperature-and therefore a higher condensing pressure-at higher loads. pcu/hr. so anticavitation trim may be required. columns with floating pressure have a higher pressure at high throughput and lower pressure at lower throughput. E. This requires calculating shell-side pressure of the heat exchanger as a function of heat load. For reboilers we have calculator programs that calculate downstream pressure. Cooling W t r ae Water-cooled condensers and coolers are widely used in the chemical and petroleum industries.278 MhceUumw Measurements and Controh For pumped systems where valve AP's decrease with increases in flow rate.. care must be taken to check valve maximum capacity at minimum AP. Since. Heating-medium required turndown is usually much greater than process required turndown. however.. Heating-Medium Condensate Condensate &om heat exchangers heated by steam. This tends to make a square-root valve have nearly a linear installed characteristic and to make a linear valve have an equal-percentage installed characteristic. letdown valves will have highest AP. and valve turndown at minimum flow at maximum APv. dual. which means that a linear installed flow characteristic is desired. and so forth is often collected in a pot and let down to a condensate header (or pumped away). coolingwater turndown may be proportional to the square or cube of process turndown. In most cases.. These flows. . Cascadmg fiom level control to flow control is recommended. will be manipulated to control liquid level. Contrary to the case of pumped liquids dscussed earlier. The practical problem is to determine valve downstream pressure. The product will usually be hot. Heating-Medium Supply (Condensing Service) The heating load here is usually a reboiler or heater. Upstream pressures and temperatures are seldom constant so orifice flow meters should be temperature and pressure compensated. Many cooling-water valves are nearly closed in the winter and wide open in summer because of the difference between minimum winter load and maximum summer load. Process Letdown from a Higher Pressure to a Lower One Since we will be looking increasingly at column control systems where column pressure is allowed to float. Dowtherm. In fact. it is desirable to check the flow regime at various flow rates.

4. which are usually near ground level. The AP transmitter may then indicate a sizable. and probably should not be used for AP’s much under 50-75 inches of water. Where this is not done. and usually enters the column at an angle to facilitate draining. . Some engineers prefer at least a 2-inch-diameter line. The instrument or control engineer is interested in the study of the dynamic behavior of thermowell/primary-elementcombinations for these reasons: 1. however. while others will settle for 1 inch.11. This technique is not highly accurate. which provides rigidity and mechanical protection. or gas-filledexpansion bulb. To predict the dynamic behavior quantitatively in order to design a temperature control system quantitatively. This high-side impulse line is ofien the source of errors. if possible.7 Column AP is most commonly measured by installing a AI’ transmitter above the top of the column with self-draining impulse lines. Particularly if it is not sufficiently vertical. 1 I COLUMN AP MEASUREMENT . however. it is most apt to be in the form of a “pencil” or sheathed assembly. Vapor line AP. It should be insulated and. The primary element is usually a thermocouple.4 on base-level measurement. In view of this problem and the need to avoid tight shutoff in the winter. in some cases. negative AP. the impulse line may act as a condenser. may introduce considerable error. the line may have slugs of liquid that can cause a partial vacuum to develop.8 TEMPERATURE-MEASUREMENT DYNAMICS I Introduction Probably the most common method of measuring moderate temperatures of liquids and gases in the chemical and petroleum industries involves a primary element (detector) and a thermowell. If a thermocouple is used. it is located above the condenser or condensers. (See discussion of purges in Section 11. The high-side impulse line is usually connected to the vapor space just under the bottom tray. except perhaps for new facilities being built at a new site. heated with steam or electricity to minimize condensation.) It is also possible to use repeaters as discussed for level measurement in Section 11. Opinions vary as to the optimum high-side impulse-line size. to facilitate maintenance. We prefer to compromise at 1 1/2 inches and. resistance thermometer. splitranged butterfly valves are recommended. dual. Sometimes. to purge the line.8 Temperature-MeasurementDynamics 279 Another vexing problem with cooling-water valves is that upstream and downstream pressures are rarely known accurately (another reason to use dual valves). I .

8 5. In view of the large annular clearance.22. Consider next the effectof changing clearance.6 8.6 93.0 €t/sec r.280 Mhcellaneow Measuremnts and Controls 2.005 inch r.21.. for the same thermocouple.. For the base case of 152 fi/sec.34 Step responses are given in Figure 11. Illustrative Examples-Forced Convection To illustrate the application of the model mentioned above.3 7.405-inch OD by 0. To decide on the design of optimum thermowell/primary-element combinations.’ The model is an improved version of one discussed in Chapters 21 and 22 of reference 2. and either the step response to ambient temperature changes or frequency response. 3.2 7. Calculator programs calculate the two time constants. there may be some error here in assuming a purely conductive heat-transfer mechanism for the annular fill.1 . To support these interests. a single time constant of 1.4 8.020 inch 0. Neglecting the thermal capacitance of oil and mercury as we did may introduce some error here. and for air in the annular space we obtain: Annular Clearance 0.7 0. sec 91. sec 7 6 .4 107 49 Step-response curves are presented on Figure 11. Example I : Gm Flow. The service is steam and the base case velocity is 152 filsec. let us look at several cases. To select optimum installation practices. If the thermocouple had been used bare.2 7.2 7. Let us consider a l/S-inch OD pencil-type thermocouple in a well 0. The details are not reproduced here but are presented elsewhere. For the same outside thennowell diameter. 7.2 8.4 14. let us examine the effect of using different annular fills: AIR OIL MERCURY 92.and 7 6 .205-inch ID.9 seconds would have been obtained for V = 152 fi/sec. sec 92. sec 7 6 . we have devised a mathematical second-order model that is simple enough to be used as a practical working tool for design engineers or plant engineers. for V = 152 fi/sec.2 92.040 inch 0.6 16.6 52. The effect of velocity on the two time constants is shown in the following table V = 300 ft/sec V = 152 ft/sec V = 50 fi/sec V = 5.6 22.

8 Temperature-Measurement DynamuZr 281 FIGURE 11.11.21 Effect of velocity on step response .

22 E M on step response of various annular fills .282 Miwe- Memuremena a d Conmi? FIGURE 11.

which are determined mostly by the service.525-inch OD by 0.1 fi/xc v = 0.51 2. reducing the annular clearance from 0. a single time constant of 0.7 6. the major time constant changed very little u t l velocity ni became very small.250-inch OD. Comparative step responses are shown in Figure 11.52 second would have been obtained.3 4.6 3.055 inch 0. oil. sec 25. step responses @"C/ft) 0.005 inch reduces the major time constant by almost a factor of 6.e s r n n Dynamk eprzeMaueut 283 As demonstrated.19 For ol the value of thermal conductivity used was 0.3 1.260-inch ID. Applications For large-scale processes.5 16. reboilers are very fast.0005 inch are plotted in Figure 11.6 fi/sec 26.3 1. and gas-gas exchangers slowest of all-but we frequently find cases in which temperature measurement provides the major lag or lags in a control loop.. ..040 to 0.3 Step responses are given in Figure 11. In reality the size of such equipment has little to do with its dynamics. all for 10-ft/sec velocity. one often hears the arguments that fast temperature measurement is not important because large didlation columns. The pencil-type thermocouple has a 0. Annular Clearance 0.26.1 31. condensers a little less so.0 25. sec 7 6 . Consider next the effect of varying thermowell internal diameter while holding the external diameter constant.3 61. Comparative step responses are shown in Figure 11.6 10. even though the mass of the thermowell is increased.11. Example 2: Liquid F h . Thermowell has a 0.25. Next let us consider flow of an organic liquid.24. and so on are slow.079 i e. Again. Let us look first at the effect of different liquid-flow velocities.005 inch r. and mercury. sec 228 1.25 0.8 1. in this example velocity was maintained at 10 ft/sec and annular fill was air. v r. 00 v= 1.01 k/xc ft/= 25. heat exchangers.23. If the thermocouple had been used bare in the first case (10 ft/sec).8 T m e a w . AIR OIL MERCURY r. liquid-liquid heat exchangers slower yet. sec = 1. For instance.3 1. Next let us consider the effect of different annular fills: air. sec 7 6 .5 In this particular case.53 1. sec 7 6 .0 v = 0..

mr.nuudUS FIGURE 11.284 Mhcehneow Memurmnts and Controls ..23 Effect on step response of annular clearance ..

11.24 Effect of fluid velocity on step response .8 Temperature-MeasurementDynarnirs 285 FIGURE 11.

25 Effect o annular flll on step response.286 M? k- Memuremenk and Controb FIGURE 11. v = f WseC .

11.8 Tempermre-Measurmt Qymunkv 287 FIGURE 112 6 Effect of annular clearance on step response .

as seen by the primary controller. If exploratory calculations show that fl resistance is too high. this problem does not exist. whereby a manipulated flow is divided by a wild flow. There are three reasons for this convention: 1. the optimum installed valve characteristic is linear. Install the thermowell in a forced-convection environment. 2. Linear flow measurements are a must since square-law flow measurements would cause the primary controllers to become unstable at low flows. Provide an annular fill of ol mercury. If the foregoing is not feasible. 6. Flow-Ratio Conventions Two more important conventions relate to flow-ratio control: 1. With linear flow measurements. they should be followed by square root extractors. and expansion bulbs may be used without thermowells. i. Flow-control loops are often secondary loops in cascade systems. approaches infinity as flow approaches zero. This is so because the flow loop gain. it can be shown’’ that the most desirable installed valve characteristic is also square law. 1I FLOW AND FLOW-RATIO CONVENTIONS . 3. or pleated aluminum foil.005 inch or less. If possible. The divider technique. If orifice flow meters are used. 2. Most instrument enpeers choose equal-percentage valves whose installed characteristics most often are somewhere between equal percentage and linear. This is to maintain constant stability as flow rate changes. A linear flow measurement is therefore a better choice. Use a bare detector if possible-many thermocouples. install im the thermowell in a section of pipe with reduced diameter to increase fluid velocity. avoid natural convection environments. If a linear flow meter is used.288 M u u W Measurmnts a d Cmtti-oh To obtain a good temperature measurement. If flows are to be added or subtracted. resistance thermometers. If a square-law flow meter is used.9 Linear Flow Measurement Convention A simple but very important instrument convention that should be followed routinely is that of using linear flow measurements. flow measurements must be linear. the following guidelines are suggested: 1. 5. choose a minimum-sized thermowelldetector design in which annual clearance is 0. 4. With the latter a wild flow is . install the thermowell parallel to the direction of flow and pointing upstream. is preferable to the multiplier technique. 3.

0-psig input signal causes a 9. For the small valve. Such an arrangement is used to get a large flow turndown.0-15. This means that if . The former lends itself well to our preferred antireset windup scheme for cascade control where the latter does not. In past years split ranging was accomplished by installing special springs in valve positioners. The problem is that the gain of the ratio loop as seen by the primary loop is zero at low flow and high at high flows. This technique. The gain 2 relay for the large valve is biased such that a 12. however. Normal level control of the reflux drum is via the distillate valve.12 2. See item 2.12 We have not found a convenient way to compensate for this with analog equipment without interfering with antireset windup. the relays may be so biased that the large valve starts to open slightly before the small valve is wide open.: both techniques are characterized by nonlinear gain. as mentioned earlier.0-psig output. Here we have a large valve and a small one in parallel. This applies whether one uses the divider (closed-loop) or multiplier (open-loop) flow-ratio scheme. A more flexible approach that has gained wide acceptance is that of using two or more amplifiers between the controller or override outputs and the valva. consider Figure 11.28. This technique is not very flexible. it has been shown81. For example.0 psig. it may mean controlling two or more valves fiom two or more controllers.27. has great flexibility. I 1. and plant maintenance people do not like to have to keep track of nonstandard valves. The h7el controller is a gain 2 relay biased such that a 9. I O CONTROL-VALVE SPLIT RANGING “Split ranging” is a term applied to techmques for controlling two or more valves from one controller. With linear flow measurements. the divider causes less nonlinearity than if square-law measurements are used.0-psig signal produces an output of 9. In effect this bias makes the small valve signal range 3. Or the relays may be biased so that the large valve ni does not start to open u t l well after the small valve is wide open. however. The small one opens first.0 psig.10 Control-Vdve Split Ranging 289 multiplied by a gain factor (sometimes remotely set) to calculate the set point for a manipulated flow controller. the gain 2 relay is biased such that a 6.0 psig. If overrides are involved. As another example.0 psig.0-9. In effect this makes the large valve signal range 9. An alternative that works well is to cascade to flow control and to use our preferred impulse feedforward scheme (see Chapter 12) for the wild flow. Although the divider technique is sometimes objected to on the 2$ that the flow-ratio loop has nonlinear loop gain.11. Cascading temperature or other variables to flow-ratio control should be avoided unless ratio turndown and flow turndown are less than 2 :1. the large one remains closed until the small one is fully open. A simple dual-valve installation is shown in Figure 11. In pneumatics the amplifiers are commonly fixed-gain relays with adjustable bias.0-psig input signal causes the output to be 9.

27 Split range of large and small valves .290 Mhce- Measurcnaentr and Conmh FIGURE 11.

11.28 Split ranging reflux and distillate valves .10 contro1-valvc Split RRngi??ly 291 FIGURE 11.

“Selection of Optimum boilers-Liquid Level MeasureFinal Element Characteristics. 8. ‘What is a ‘Fire-Safe’ Valve?.. 10. McGraw-Hill.11. Houston. Buckley. Cliffs.. Dynamics Improves Level Control 12.J. Driskell. W h g t o n .. N. P. S. Sanders. New York.01. R.0 psig in terms of liquid level. the distillate valve is closed. Buckley. 1959). Binder. the distillate valve is wide open. P. 7.5 psig.. R. “Control Valve Sizing with 16. N.. 1982.’’ ISA Paper NO. Buckley. Techniques . Proc..” presented at Texas A&M Symposium. K. J. Jan.’Handbook. 1955.. 6. P. A&M Symposium. New York. S. “Design of Pneumatic Flow Controls. S.. W. 5. “Control Valve 15. if the level rises to 75 percent.0 psig. S.Eng.” 1984.f Process 4th ed. D. Schnelle.. B. S.. 55-6-3.292 Mticelluneow Measurements and Cmttrok the level drops to 25 percent. Lupfer. S.”Proment.” presented at National ISA Trans. 1983.0-6. “Internal Column Re.5 psig from the level transmitter. 33-48 Valves for Pumped Systems.. “Analysis of System 1964. the relay output is 9. 9. Driskell. Symposium. 13. Schmoyer.0 psig in terms of the level transmitter.” Chem.” INTECH. Control... L.0-15. 59(2):66ceedinds. P. 1963). the relay output is 9. 6(6):34-39 (June and Se-. P. Research Triangle Park. E. and R. ANSI/ISA S75. and D. 269-281 1951. P. Buckley. PeWys Chemical Enginem. 14. Arant. Tex. we provide a gain -4 relay so biased that at 13.. we provide another -4 relay so biased that for a level transmitter signal of 4. Advanced Fluid Meat Texas A&M University. 3. Fifth National Chempid 69 (Feb. 1958. ISA Conference. Suppose now that the level rises above 75 percent or drops below 25 percent.” presented . Del.. 2. chaniq Prentice-Hall. perature Measurement. 12( 1):45-55 (1973).. (July 1974). Berger. C. S. ISA. Wiley.” mentation. Jan. Englewood 1979.C. 1976. 1963. Flow Equations. “Liquid Level Mea“Problems in Slurry Service Instrusurement in Distillation Columns. D..”Hydrocarb. E.” presented at Texas of Chemical Operation. Buckley. To protect against high level. Hepp. Control Valve Sizing “Computer Control of Distillation Reflux. C. L.. Prod. “Dynamics of TemJan. P. 4.. This gives the air-to-close reflux valve an effective signal range of 12. (May 1981). This relay makes the effective signal range of the reflux valve 3. 1964.0 psig.” ISA J. To protect against low level. REFERENCES 1. “Optimum Control ISA Formulas. P. Buckley.

signal flow diagrams. * It is assumed from here on that the reader is familiar with differential equations. Transfer Function Before the age of computers. singleoutput) control theory. it is our intention to provide. electrical engineers adapted perturbation techniques originally worked out by physicists to develop a body of control theory based on linear differential equations. salted with a modest understanding of control theory. and Laplace transforms are the primary working tools. Laplace transforms. q&titative procedures for &e design of multivariable &tillation controfsystems: How do we go about this? Today there are four basic approaches to control system design. where possible.1 WAYS OF DESIGNING CONTROL SYSTEMS As discussed in Chapter 1. single-output (SISO) loops. multivariable systems. This approach. frequency response. 295 . z-transforms. this is a qualitative approach based primarily on past practice and intuition. Today it is the cornerstone of useful process control theory.I2 a Instrumentation Approaches to Quantitative s mentioned earlier. The end is a multiplicity of single-loop controls. often provides a good starting point for quantitative design. Although originally intended for single-input." three of which are quantitative. 12. It is almost the only available analpcal (noncomputational) approach to control system design. and other elements of SISO (single-input. Frequency response. this approach has been extended to moderately complex noninteracting.

Distillation composition is nonlinear. Adaptive control is geared to nonlinear processes in which one set of tuning parameters may not suffice. usually programmed in BASIC. ranging fi-om simple PID controllers that adjust the tuning according to a predetermined relationship between a measured variable and controller tuning parameters (“gain scheduling”) to statistical types where an internal model is statistically fitted to the measured response and the tuning * One author can recall this term being applied to the Ziegler-Nichols papers of the early 1940s. . Rapid evolution has taken place. These permit the engineer to design many commonly encountered control systems while seated at a desk. and (2) the ability to study larger systems. In addition. is proving to be a very usell portable machine. the HP-85. Study of control systems by these methods requires a fairly capable computer. and partly because it is not really appropriate for many applications. while the latter permits including nonlinearities as well as prediction of overall system performance. “modern control theory”. we have a large collection of control system programs. Newer Methods Since about 1960 a variety of newer mathematical and computational techniques have been introduced. the HP-65. In the last few years. Small but powerful interactive computers with CRTs and printers. It can be used as a “smart” terminal to access mainframe computers. although not very severe compared with waste neutralization. The former can be employed to establish overall control concepts and structure. Two of the best known approaches are the (1)multivariable fiequency-responsemethod. These are sometimes called. are also available. the process control theoreticians studying multivariable control have been focusing on (1) adaptive control and (2) predictive control. for example. tray-to-tray dynamic simulations coded in FORTRAN or Pascal. pioneered by Rosenbrock and associates in England. The first card-programmable pocket calculator. we have terminals for minicomputers that permit large.’ and (2) the “state space” or “state variable” approach. For the HP-97 and HP41C. This is partly so because few practitioners are familiar with the technology. The HP-9845B is a good example. collectively.296 Approaches to Quuntitative Design Simulation The availability of computers beginrung in about 1950 permitted two advances: (1) the ability to work with nonlinear differential equations (usually ordinary). and today there are many small but powerfd machines that can be purchased for about the price of a color television set. on-line implementation usually also requires an on-line computer.* Their application to industrial process control so far has been limited. Analog computers were used at first but have been displaced mostly by digital computers. Use of both the transfer-function and simulation approach has been very h i d i d . appeared on the market in 1974. and again about 1950 to transfer-function techniques. which has been popular for aerospace and military applications. There are many adaptive techniques. A smaller BASIC computer.

12. great pressure to make the necessary design decisions as quickly and inexpensively as possible. money. .27and M ~ r a r for ~ i ~ details. therefore. -Systems with time-varying parameters requiring adaptive tuning. we progress from the most basic kind of infbrmation. The reader is referred to Bristol. liquid-level. -Interacting systems with two inputs and two outputs where straight decoupling (noninteracting design) is not satisfactory. references to work of Brosilow et al. There is. ~ et al.2 K i d OfInfmmatiOn Available 297 parameters are determined to achieve some prescribed criterion of performance such as minimum variance. to Predictive control describes control algorithms that use an internal discrete convolution to predict the process output... there is not enough time. Some usell insights are provided in a recent book by McAvof’ and a paper by Ray.2 KINDS OF INFORMATION AVAILABLE In almost all design projects. and pressure controls.29 An interesting industrial perspective is presented in a paper by Doss et There seems to be little incentive to apply this newer technology to most flow. In the following categories. The reader is referred to et al. which permits the quickest decisions. but one can visualize a number of other process control situations where its use would be beneficial: -Interacting systems with more than two inputs and two outputs. Since there are real economic incentives for more extensive use of multivariable control systems. This logically leads to consideration of self-tuning regulators such as hscussed by h r o m . and since it is becoming increasingly common to have on-line computers. pH control systems with variable b d e n n g are an example. Consequently it is common practice to start with the simplest possible methods. Batch reactors and batch distillation are examples.u~”7’9 12.2~ name a few. and have H o o ~ ~ s . This approach is suited to processes that are &cult to model parametrically. for further explanations.).20 and Sa~try. -Systems where controlled variables cannot be measured directly but must be estimated fi-om other measurements (see Chapter 10. it is likely that these techniques will be used more in process control in the future. 2 0 -Systems where time-optimal control is economically attractive. to more sophisticated knowledge. Those rare engineers who have extensive training in process control and who are also experienced have a wide selection of tools with which to work. The latter requires increasing understanding of control theory and more calculations. are multivariable.2’ &zrom. or personnel (especially quahfied personnel) to do as good a job as we would like.24Touchstone Mehra ~ constraints. and to depart from them only when necessary. and then adjust the controller parameters so that the process follows a specified trajectory.

hardware installation. Catalog of Open-Loop Process Transfer Functions At this level simplicity begins to fade. and in the heads of experienced engineers. We find. the catalog is usually a substantial time saver. Process Control Catalog Here we begin to use quantitative procedures. as well as instruments. heat exchangers. We have a sizable collection of calculator (HP41-C) and computer (HP-85)programs. there are no simple relationships for conaroller tuning. This is usually in the form of frequency response or step response. or simulation methods. Since two or more of these transfer functions may have to be combined. resonance peak magnification for the function: KGOw) 1 + KGcjtp) Instrument Dynamics Catalog Although not assembled in one place. there exists considerable information today on the dynamic performance of transmitters. valves. deriving process transfer functions is often tedious or difficult. Here we have transfer functions for pipelines. and closedloop resonant frequency. or both.298 Approa&i?s to Quantit&ve Des&n Practices There exists in the literature. a body of knowledge about good practices: process configurations. Control Loop Catalog Here we have a collection of open-loop transfer h c t i o n s with simple equations. however.e-” The e n p e e r who knows that this transfer function fits a given situation can quickly find controller parameters with a minimum amount of calculation and a minimum knowledge of theory. Since. control system structure. We also have tuning charts for flow and flow-ratio controllers. Instead the e n p e e r must use analpal methods. including one that takes an open-loop transfer function and automatically calculates loop gain for a specified Mp. simple algebraic formulas for selecting level and pressure controller settings. reset time. or tables for finding controller gain. and since there are many possible control systems. An example is: (75 + 1)(p7s + 1) K. charts. relays. for example. and so forth. for example. controllers. Part I1 of this book tries to assemble this kind of information for distillation columns. . and distillation columns. and so on. including frequency response.

In a recent book. Hougen5 presented an excellent discussion of techniques and equipment. Many commercial controls. and for manualautomatic switching -Anti reset-windup for loops with overrides (also simplifying manualautomatic switching) -Feedforward and interaction compensation without interfering with either normal reset or anti reset-windup -Use of Smith predictor and other advanced controls without interfering with either normal reset or anti reset-windup Satisfjmg these requirements is not basically a matter of either hardware or s o h a r e . Note that the controller gain. one must resort to the textbook approach: Start from the beginning. The design must accommodate such diverse requirements as: -Minimum operator attention for startup and shutdown. but there were indications that this might be changing. or both. If all else fails. are seriously deficient in one or more of the requirements listed previously. and the reset time. experimentally.3 Functional Layout ofControl Loops 299 For some applications. but the point is that if a PI control algorithm with external reset feedback is used. write the equations for all elements in the loop. feedback from the valve loading signal. and use frequency response or simulation or one of the newer methods to design the control loop. Few plant loops have all of these features. is shown in Figure 12. and feedforward compensation (or decouplers). with over 35 years’ experience in d e s i p g control systems by quantitative methods. straightforward fashion. and with optional overrides. they all fit together in a simple. K. 12. Rather it is a function of algorithms (mathematical functions) and their topological relationship to one another in a control-loop layout. It also easily accommodates M y automatic startup and shutdown-only a switch and a lag unit need be added. predictor. we resort only occasionally to this method. it may be necessary to determine process or instrument dynamics. whether analog or digital. TR. As of 1981 most vendor-designed digital systems were less flexible and less versatile than the best analog systems. such as temperature measurement. Today.6 A single loop with external reset feedback (ix. “Book’ Design of Control Loops When the above-mentioned sources are not adequate. not from the controller output).may be nonlinear or may be adaptively .1. This structure satisfies all of the design requirements stated earlier..12. we have HP-41C programs that calculate either step or frequency response..3 FUNCTIONAL LAYOUT OF CONTROL LOOPS The overall functional layout of a control loop is extremely important.

g z ai > m > z i c . c .300 Approaches to Quantitative Desbn 5 e U E 8 2 m a . 2 5 z s E s T: E z L 5 T 2. W' !sg E= LLm . e E s U 0 c . U L vi ti i a .

The basic PI controller equations are: eC(s) = Kc[e&). additional functions may be required for more complex control systems.em(s)] + OR($) (12. the . but is sometimes combined with PI in analog devices. it is nearly so when high sampling rates are used. Derivative is shown here as a separate unit.12. controllers (PI and PID). and either difference equations or z-transforms are better suited for some digital applications. = e&) and we can combine equations (12.3 Functional Layout of Cuntrol Loops 301 set fiom external signals.is usually desirable. At this point we wish to point out that (1) the feedfonvard function. The theory and mathematics involved in combmng feedforward compensation. The configuration shown is for “increase-decrease” controller action. as will be indicated later. rD is the derivative time constant. This feedforward function is also shown ahead of overrides and limiters. It may be noted that Laplace transform symbolism has been used in this and the next two illustrations to represent reset.1) and (12.1) and (12. and (2) making the feedforward time constant. derivative. A single feedforward function is shown but. and feedforward functions.commonly termed “impulse feedforward. and anti reset-windup have been covered in detail elsewhere7 and are reviewed briefly below. K$--s/(r-s + l). r-.1. and a is a constant.2) If there are no overrides and no feedforward signal. for “increase-increase” action. Note that rR is the reset time constant.1 box shown in the measured variable channel would be moved to the set-point channel.” is a convenient way of feeding forward without interfering with reset when external reset feedback is used. usually in the range 6-20. Following is a brief review of the mathematics of a PI controller with external reset feedback and impulse feedforward compensation. Although not totally valid for digital computers or microprocessors.3) Considering next the system of Figure 12. which might not always be the case. Care should be taken to avoid putting derivative inside the reset loop. however. rR. overrides.2) to get the conventional PI controller equation: (12. equal to the reset time constant. This approach to feedforward also provides a convenient way to connect interaction compensators (decouplers) into control loops. It is not valid for nonlinear functions. is being increasingly accepted by control engineers. The basic structure of Figure 12. let us write the equation for the summer output: .2 with feedforward compensation.

4) If we assume that there are no overrides: &(s) = el44 Letting: (12.5) to get: [e& @ .8) . @!E & + 1 (12.5) .6) which reduces to: e.2).1).4). this reduces further to: Om(s)= K c - +1 E 7R.(s> +1 (12. (12.e&>l = E (5) we may combine equations (12. ( s ) = Kc E eJ-u(s) Kf735 (4+ 3 8f(s)735 + 1 + (12. (12.7) If we make 7f 7R.- 7Rs +1 7Rs 9 7 R s + 7Rs E (’) Kff 7j5 7R + 1 e.2 Single-loop with feedforward compensation M S ) = ec(s> + ef(z.302 Appodes to Qmntithve Des@ FIGURE 12..F (4+ Kff!f(S) (12.(s) = = K. and (12.

If we look at these loops from the standpoint of classical.(s) also shows a step change. Further work needs to be done to compare their performance with that achieved by PI or PID controllers with feedforward compensation. whose set point is normally set by the primary controller. for example: Of(S) = S which is a step function. and (3) constraints. For example. then O. that is. But we increasingly cascade to flow controls that are usually so much faster than the primary loops that there are no stability problems. Perhaps. 12.4 ADJUSTMENT OF CONTROLLER PARAMETERS (CONTROLLER TUNING) In Chapter 1we suggested a control system structure in which most control loops are divided into three classes: (1) material balance. we see that the material-balance loops function as low-pass filters (slow dynamics) while the product-quality loops function as bandpass or high-pass filters (faster dynamics). A true cascade system (see Figure 12. If. the set point is adjusted fiom another device. This greatly minimizes the need for manual-automatic switching-none is needed in the primary control station.3) has at least two controllers: a primary or master one and a secondary or slave one. Cascade control probably has been used many times when feedforward compensation would provide superior control. Use of the Smith and other predictors recently has been shown by Seborg et aL3 to provide substantially better feedback control of distillation-column terminal compositions than do PI and PID controllers. A ‘kild” flow is multiplied by a constant to calculate the set point of the controller of the manipulated flow. the most common example is open-loop flow-ratio control. Our preferred scheme uses the secondary variable for external reset feedback to the primary controller. This is not cascade control since only one controller is involved. single-loop theory.12. Occasionally one finds a single loop with external set. For stability reasons the secondary loop in a cascade system should be five to ten times faster than the primary loop and should have dead-beat (nonresonant) tuning. (2) product quality.4 Adjustment o Controller Parameten (Controller Tuning) f 303 With this design steady-state signals are not fed forward. An example would be column-top temperature control cascaded to reflux flow control. . one might want to maintain a constant holdup time in a chemical reactor and use throughput flow to calculate desired liquid level. This usually occurs when one wishes to control a variable whose set point must be calculated. but changes are. however.

e 2 m 5 c . s 2 E c 8 ul m 7 in c .304 Approaches to Quantitative Des@ E % 8 3 I + . 0 3 0 c k 4 3 d 0 28 #E 0- =E bk . c 8 U ! ci c 8 I 8 m E 3 .

Since the on-line models to be employed require calibration. we design the product-quality loops to have high. The materialbalance loops not only filter out flow disturbances. Feedforward compensation often may be used instead of feedback to get improved control without problems with either stability or resonance Peaks.5 Enhanced Contvol OfDistdatim Columns via On-Line Models 305 To eliminate interaction between the two. to keep the plant operating safely and smoothly when it might otherwise be shut down by interlocks or operator decision. the tests are difficult to run outside of a laboratory or computer. in most cases. it provides great attenuation for disturbances that might otherwise upset product quality. usually implemented by a digital computer. some degree of on-line identification is highly desirable. The last of these is greatly aided by constraint controls. An Mp of 2db is the normal criterion. it is also true. For averaging level control we try for a damping ratio of at least unity. the three best opportunities for profitable operation via process control are usually: (1)maximum-capacity operation if the plant is production limited. commonly called overrides. and (2) that for all other applications.5 ENHANCED CONTROL OF DISTILLATION COLUMNS VIA ON-LINE MODELS Once a distillation column or train for a continuous chemical process has been designed and built. With these two limited exceptions. Functionally these will be.12. An interesting implication of studies to date is that for product quality . A modern perspective on controller tuning has been presented elsewhere. or constraint controls. closed-loop natural frequencies compared with those of the material-balance loops. but also filter out composition disturbances if the holdups are well agitated. second. First. (2) minimum-cost operation if the plant is market limited. This is discussed in more detail in Chapter 16. and the end result usually is a highly resonant system that is easily upset. Although not usually pointed out in the literature. The two Ziegler-Nichols procedures are empirical. sophisticated feedforward compensators. Maximum-capacity or minimum-cost operation is facilitated by the use of on-line models. that the engineer should be careful not to design a feedback system whose closed-loop natural fiequency is within the frequency range of disturbances. as Gould' has indicated. For controller tuning we therefore follow two basic philosophies: (1) that for averaging controls such as most liquid-level and some pressure controls. it eliminates interactions between the two classes of loops. optimization is not considered in this book. This design has two advantages. which is considerably more damped (less resonant) than the response obtained by Ziegler-Nichols and similar "tight" tuning procedures.' 12. For other applications we usually strive for tuning that will make the loop slightly underdamped. and (3) increased annual availability (sometimes called utility).

" the static gainschanges in top and bottom compositions in response to changes in distillate or boilup or other variables-can be extremely nonlinear. This means that much of the time the loop will be overdamped and sluggish. This can play havoc with controller tuning. In the absence of adaptive tuning. and there is no implication that the techniques presented are universally desirable. temperature and flow measurements may require much increased accuracy-less than one part in 10. For many columns it will be adequate to approximate the separation by the model for a binary system where relative volatility may be a nonlinear function of composition. and column pressure. for one example &/a0 (change in top composition per unit change in distillate) decreased 30 percent for less than 1 percent change in distillate flow. Adaptive Tuning of Feedback Controllers The feedback controllers we have in mind here are those for control of column terminal compositions. as top or bottom composition becomes purer.306 Approaches to Quantitatipe Des@ control we may need. Since effective use of the model requires a knowledge of the number of theoretical trays. Reference 10 shows that many kinds of column gains may be calculated by a modified tray-to-tray method. in addition. As shown in a recent paper. greatly increased sensitivity or resolution in control valves. will be more variable. the static gains become smaller. Thus we have reflux-to-feed. feed enthalpy or 4. This means that reset and derivative (if used) need not be changed. But there are at least three other variables that can be important: feed composition. particularly controller gain.000. steamto-feed. The economic value of the enhanced controls discussed in this book varies widely from column to column. Generally speaking. In most cases it is probably satisfactory to run the model no more frequently than every 5 or 10 minutes. some sort of physical calibration or identification is required. . after the computer has calculated the appropriate column gain for a particular loop. and sometimes other ratio-control systems. therefore. it can easily calculate required controller gain to hold loop gain constant. production personnel will usually adjust the composition set point for higher purity. Therefore. feedback controller gain must be set low enough that the loop never becomes unstable. at least more frequently than in the past. This means an increase in average utility consumption and lower maximum column capacity. Feedforward Compensation Our use of feedforward compensation for distillation columns currently is based almost entirely on feed-rate changes. As shown in reference 10. Composition. The work of Wahl and Harriott" suggests that column dynamics are relatively insensitive to small changes in reflux and boilup. At present it appears that only controller gain need be adaptively tuned. To ensure getting a good product.

More Accurate Determination of Constraints The most important constraints here are those for flooding. A possible way to accomplish this is to design the downcomer bottom reservoir so that its outlet weir is a Sutro weir. we can determine whether base level should be increasing or decreasing and at what rate. Ordinarily our approach to protection against flooding is to use a high column AP override to pinch steam. The outflow fiom such . are really functions of a number of variables. we can calculate internal reflux. one may calculate the required changes in disullate or reflux at column top. There are no problems with stability such as sometimes occur with tightly tuned. Floodmg and dumping. Similarly. Usually we test for flooding by fixing as many column operating variables as possible and gradually increasing ni boilup u t l a rapid increase in column AI' or a rapid decrease in column base level occurs. sophisticated feedback controllers.12. Rippin and LambI2 showed that most of the benefit of feedforward is from the static gain term. the accuracy of these models is not sufficient. or with cascade controls. A more positive approach is to measure internal reflux. and therefore flooding. not only to the column base. From material-balance and energy-balance equations. for a given column. A more rapid drop in level than predicted would indicate loss of internal reflux. we will have to obtain the data experimentally.5 Enhanced Control o f D i d h M n Columns On-Line M o h h 307 By use of the tray-to-tray model mentioned previously. and fiom each tray where there is a side draw. 3. The three great advantages accruing &om the use of feedforward compensation are: 1. In work carried out at the University of Delaware a few years ago. entrainment. and boilup or bottomproduct flow at column base required to hold terminal composition constant in the face of changes in the variables listed. Fairly simple feedforward dynamic functions such as first-order leadlag are usually adequate and may be calculated off line. however. Maximum column capacity is increased since manipulated variablesreflux or distillate. Compositions are held closer to set points than is possible with unaided feedback controllers. We could also check for flooding on h e . To protect against dumping or weeping. the column is less likely to weep or dump. Upon knowing boilup and bottom-product flow. dumping. but also to each feed tray. 2. boilup or bottom product-are not changed as much as if unaided feedback controllers were used. The set point is usually determined in advance by consultation with the column designer. More column capacity could be achieved by using on-line models to calculate maximum permissible AP or minimum required steam flow. we usually provide a minimum steam-flow override whose set point again is determined by discussion with the column designer.l3 If. and maximum and minimum column-feed rates.

the greater is the column turndown.15 Determining tray efficiency is very desirable in order to calibrate the column model used. the column feed rate should not exceed that which the column can handle (ix. Increasingly. the greater is the column capacity. Some of the considerations. Similarly. If reflux flow thus calculated becomes significantly lower than predicted by the column material and energy-balance model. it use is urged because otherwise column temperatures might not be good measures of composition. therefore. there is seldom a good technical or economic reason for its existence. to find its own level-or is manipulated either to minimize steam consumption per pound of product or to maximize column capacity. That is.l6 With the advent of heat-conservation schemes. this relationship may be tracked and noted for control purposes. This problem is sometimes approached in another way by determining column tray efficiency on line. can separate properly without flooding). We can also use such flow measurementsto determine column inverse response. would give an output signal linearly proportioned to flow. fixed value is well established in the chemical and petroleum industries. For best results column feed rate must be limited. Both maximum and minimum feed rate may be calculated by the column model. The closer to flooding the column can operate. entrainment becomes dominant before flooding occurs. Entrainment is related to floodmg and there are models for its prediction. as the column operation approaches dumping. pressure either is allowed to float-that is. feed rate should not fall below a minimum to avoid using excessive steam per pound of feed.l4 Dumping is indicated if base level rises much faster than predicted by the column model. are as follows: 1. therefore.308 Approaches to Quantitative Desgn a weir is linearly proportional to liquid height above the weir reference line.. Balancing Energy. flooding is indicated. In some of the older literature. .” Although the practice of using automatic pressure control at a predetermined.and Material-Handling Capacities This section could well have another title: “How To Use Column Pressure as a Manipulated Variable. If the condenser capacity is biting (compared with those of the column and reboiler). ordinarily one would expect each increase in reflux to result in an increase in overhead purity. But temperature by itself is not a good measure of composition. raising column pressure will increase condenser capacity. But sometimes before flooding occurs a point is reached at which a hrther increase in reflux results in a decrease in purity. however. Sometimes. With an overhead composition measurement. it is particularly difficult to generalize about how to manipulate pressure most successfully. The closer to dumping the column can run. But this will reduce reboiler capacity. A differential-pressure transmitter. however. Control at the maximum reflux rate may be achieved by a peak-seeking optimizer (see Chapter 9). As maximum permissible column A P is approached.

12. If reboiler capacity is limiting. Operation at lower than design pressure may cause entrainment or flooding. together with a pressurerelative volatility-separation model.5 Enhanced Control OfDljtilLatiun Colzrmns via On-Line Models 309 2. let alone inlet and exit cooling-water temperatures. 5. 95 percent open. On-Line Identification Design equations for process equipment rarely have an accuracy of better than k 10 percent. if maximum capacity is not needed. say. and column material balance.C o l u m n tray efficiency --Column inverse response -Column-material and energy balances Some of the preceding information can be used by the computer directly for enhanced control and some can be used to predict the need for maintenance. may be required. we might increase column pressure to prohibit the valve from opening further. For example. From the preceding one can see the fairly elaborate considerations regarding the role of pressure in heat transfer. entrainment. rarely does one fmd a measurement of cooling-water flow rate. For example. and columns occasionally plug. A wide-open valve indicates that the associated heat exchanger is doing its utmost. 3. 6. it is desirable to run tests periodically. For many columns. Raising column pressure may interfere with or limit the capability of the feed system to feed the column. pumps and valves wear. This means that accurate determination of capacities and operating characteristics must rely on tests of the completed plant. Lowering column pressure may interfere with pumping away or letting down column product flows. The use of material. 4. But tests on a water-cooled condenser are useless without this information. if the condenser cooling-water valve is. Fortunately we most commonly operate with low pressures at low feed rate and higher pressures at higher feed rates.and energy-balancemodels. it may be increased by lowering column pressure. operation at the lowest possible pressures is desirable to take advantage of higher relative volatility to achieve minimum steam consumption per pound of feed. and often are much worse. flooding. . With sufficient measurements an on-line computer periodically can determine: -Heat-transfer coefficients --Column approach to flooding and entrainment -Column approach to dumping . We previously suggested (see Chapter 9) that the positions of the steam and cooling-water valves give clues as to whether the reboiler or condenser is limiting. Typical plant instrumentation is seldom adequate for such testing. Since heat exchangers sometimes foul.

G. “Inverse Response in Distillation Columns. F.. 5. (Jan. 1976. Proc. Mass. and P.. Budey.. Academic Press. P. T. and Oct. “Experimental Evaluation of Analytical and Smith Predictors for Distillation Column Control. Luyben. Prqg.. Computer A d d Control Systems Des@.” IEC Proc. E.” E C P m .” presented at American Institute of Chemical Engineers Meeting. L. F. 1977). E.” Hydrocurb.. 2. Buckley..” Chem. F. Buckley.310 A p a d e s to Quantitative Design Procedures for running tests on line either by statistical analysis of operating data or by injecting test signals are commonly labeled “identification” or “parameter esti~nation. S. Vol. 8. Harriott. “A Modem Perspective on Controller Timing. V. C h e m d Process Control. Rippin. D. K. Seborg.-L.. Years ago this was called “black box” testing. “Status of Design Methods for Multivariable Control. B.” Chemical P r o m Control. Hougen. P r m t r . 18(1):177-182 (Jan. K. . and C. and D.. S. 1971). 1960. Beck. 1977.... H. Arnold. H. Addison-Wesley. J. (Aug. 16.. “Designing and Programming Control Algorithms for DDC Systems. 1977. C. P. J.. 25(1):24-32 (Jan.” presented at Texas A M Symposium.” Cont. R. Eng. Cox. W.” Cont. N. Tex. 281-285 (Nov..” AIChE J.”’~ Basically a computer processes input data to do one or more of the following: 1. “A Theoretical Study of the Dynamics and Control of Binary Distillation. K. 9(3):396-407 (1970). ISA... A. D s Do. Distillation Column. Rosenbrock. R. Reading... S. Tyreus. Edgar. J.. E. aaee 17. e. J. Buckley. University of Colorado. B. Ryskamp. L. Thesis. Find a suitable Merential equation or transfer function for a process operation where a mathematical model is lacking. S. 15. 1979. “How to Use a Small Calculator in Distillation Column Design. Gould. AIChE Symposium Series. 6. and D. P. L. Lamb.. New York. “Multivariable Control System Design for an Industrial e. Wood. 49-55 (June 1978). 1969.. J. McGraw-Hill. ‘Tray Flooding Sets Grade Thruput. DirtiuatiOn Control. “Identitication of Distillation Column Tray Efficiency. “Understanding and Prediction of the Dynamic Behavior of Distillation Columns. Estunanon in Engineering and S k . Determine static coefficients such as tray efficiency or heat-transfer coefficients. H. Eng. 3. Measurements and Control A p p l ~ 2nd . 2. 13. . G. 14. L.C. 72..ed. D s Dm. 4.. Wahl. 1977. Cox. Meyer. T. “Designing Overrride and Feedfomard Controls. 159.. REFERENCES 1. 1979). Determine numerical values of coefficients of preselected differential equations or transfer functions. Jan. Wiley. 1975. 3. New York. 1973. New York. and D. 9. 1974. Department of Chemical Engineering.S. University of Delaware.” M. 11.. ~ Research Triangle Park. 7. 10. 1979). Mar. 1977). 12. O. Rollins. Eng. E. no.” Buletin. and K. H. Renaud. D. Bristol. P. and W. R. Wade. Shinskey. Houston.

21 (1982). Sastry. 23. 25. J.Rejimms 18.’’ ISA Trans.. W. et al. 1981. Nov. San Francisco. “The Design of Industrially U s e l l Adaptive Controllers”... K. Cutler. Advanced Process Con~ol. T. 2. Jr.. S. et al.’’ Proceedinas Intematiod Symposium on Process Control. E. 1982. V. M. E. Interaction Adysis. Engineering Foundation Conference.. P. R. L. “Tntemal Model Control. “Topics in Multiple Input-Multiple Output Adaptive Control.. 19. 30.. McAvoy. Do. T. Ga. 26.. Ramaker. 28. and C. Fla. Garcia. 1983. 1983. New York.C. Research Triangle Park. Calif.. H. 21. DOSS. Compio. 1981. Mehra. A. D. for Process Control in the Eighties”. A. 1977.. Aug.” Proceedings 1977 ISA Spring Industry Oriented Conference. lators-Design Principles and Applications. et al.. 29. 1980. 22(3) (1983). Orlando. R. Touchstone. presented at American Institute of Chemical Engineers Meeting. H. 24. W.. “Adaptive Control. Morari.” Report LUTFD2/ (TRFT-7177) 1-068( 1979) from / Lund Institute of Technology.. E.. J. et al. B. H o o p . Schnelle.” presented at Joint Automatic Control Conference. McGraw-Hill... 22(3) (1983). Bristol. H. . “Dynamic Matrix Control.. Washington. 311 Sea Island. “Self-Tuning Regu20.C.. “New Directions J. “Multivariable Process Control-A Survey. 22.” Proceedings. C. and B.. Kyoto. Anaheim.” LEC P m .. Japan. 1981. N. Des.” AIChE Meeting. K. ‘‘Model Algorithmic Control. Sweden.. Ray.. ISA. Ray. 27. AIChE.. ISA Trans. H. “A Self-Tuning Controller.. D. and A. Automutica 13 (1977)..

The first of these is commonly termed “inverse response” because it causes a momentary increase in low boilers in the column base. however. depending on tray design and operating conditions. although for many columns the first-order approximation is adequate. or (3) not to change at all. that. * This chapter is based on reference 5. and then derive an approximate model for a combination of trays. In this chapter we will derive the dynamic relationships between internal reflux and both vapor rate and external reflux (or feed) as function of tray and column design. A discussion of reboiler dynamics will be deferred to Chapter 15.1 3 i Tray DynamicsMaterial Balance* 13. It has also been assumed that internal reflux is not affected by vapor flow changes. In reality.’ although Harriot? observed that in an AIChE report3 a bubble cap tray at low liquid rates had less holdup at higher F factors than at low F factors. first-order lags. and heat-storage effects will be assumed to be negligible. (2) to decrease temporarily. We now know. an increase in vapor flow may cause internal reflux: (1)to increase temporarily. Vapor flow will be assumed to occur without lags. This apparently was first noted by Rijnsdorp. The basic tray hydraulic equations are based on the treatment by Van Winkle4 First we discuss what happens on an individual tray. as we shall see.1 INTRODUCTION n the past it has been common practice to treat the response of internal reflux to either external reflux or feed changes as the result of a chain of noninteracting. 3 13 . it is a chain of noninteracting second-order lags. which is followed by a long-term decrease in low boiler concentration.

between the flow from tray n + 1. Ibm/ft3 H h = height.314 Tray Dynamia-iUawiul Balance 13. + ~and the downcomer outflow.2 TRAY HYDRAULlCS (See Figure 13. P. In Laplace transform notation: (13. liquid density in downcomer. from tray n into downcomer + 1.1) Downcomer Liquid Level The height of liquid level in the downcomer is the integral of the difference .2) where A POc = downcomer pressure drop. assumed to be uniform fi?. above liquid on tray n Note that we assume that pOc does not vary. is pressure on tray R . feet. lbf/fi?..1) flow. lbf/ft' prn = density of aerated liquid on tray n. feet. fromtrayn + 1 overflow. lbm/ft3. in downcomer Downcomer Pressure Drop and Flow The downcomer pressure drop is determined by the differences between the liquid heads and static heads: (13. from downcomer. liquid usually slightly aerated downcomer cross-sectional area. lbm/sec. lbm/sec. woe. . + 1. w . liquid height. of aerated liquid on tray just downstream of inlet weir Pn+= pressure.

2 Tray Hya'radh 315 FIGURE 13.13.1 Distillation tray schematic for flows and liquid evaluations .

(s) .4) where hd = liquid head in inches of liquid-flow pressure drop through the downcomer downcomer flow. 0 5 0 3 6 ~ &X BL PDC (13.11) .0 .03 Q& hd = (100A h ) 2 (13.BL . gpm minimum downcomer flow area.2) may be Laplace transformed to: AP.6) B C (13.-[PDcKx(~) .c (13.5) . we may write: (13.--i 2 hP.9) 1& . ftz &= Ah = This equation may be rewritten: (13. from page 508 of Van Winkle:* 0.7) or and (13.KRPTR(J)I 8 6 (13.3) - [Pn(s) .P n + l ( J > l Next.10) Now since wx is determined by APx.3 16 Tray Dynumia-Mated Balance Equation (13.jSmHk(s) .

we assume a linear gradient across the tray: where V. - 1 Pm [From equation (13.. By Laplace transforming we obtain: ( 13. + H.1) A. feet.13) or in Laplace transform notation: . f (area where ? bubbling occurs) height. .HW(4 + H%f) Vm(4 Am 2 Inlet Liquid Height Over Weir Let us assume that the change in inlet height over weir is the same as the change in outlet height over weir. ( ~ ) = H25) ( 13. over outlet weir height.. Ibm. aw.15)] . = volume. H. of outlet weir (constant) inlet (see Figure 13. Then: ~ . is the active tray holdup.2 Tray Hydraulics 317 Aerated liquid Holdup and Gradient on Tray With reference to Figure 13.. = = HA = aerated liquid height. feet. fi3.13.16) where av. of aerated liquid on active area of tray (downcomer volume excluded) active tray area.15) where W . = H. (13. at tray = H.1.14) Volume of Liquid on Tray The volume of liquid on a tray is related to the inventory and the density: (13.. feet.


Tray Dynamia-M&&



avm -aPm


-wm P2


-vm Pm

[From equation ( 13.15)]

Tray Liquid Material Balance
On the active part of the tray, the liquid material balance in Laplace transform notation is:
~ D C ( S ) - fP,(s)




where w, = overflow, Ibm/sec, from tray n.

Aerated Liquid Density as a Function of Vapor Velocity
In Figure 13.16, page 516, Van Winkle4 presents a p3t that shows that froth density decreases with an increase in vapor rate. From thls we can calculate a slope, apm/aF. Then:

F = F factor
= U V A p y = -p .
vv -0.5


u~~ corresponding to A~ =
wv = vapor rate, Ibm/sec Ibm p v = vapor density, -

vapor velocity, ft/sec,


Liquid Overflow f o Tray rm
The Francis weir formula can be written in the form:

(13.19) (13.20)

13.2 Tray Hydratrlia

31 9


Q = overflow, ft3/sec, of aerated liquid



a constant for any given weir (see Van Winkle: pages 507-508)

In Laplace transform notation:
(13.21) where (13.22) and (13.23)

Tray Pressure Drop
In Laplace transform notation:

aArm P,(s) - P , , + ~ ( S ) APm(s) = = WV(4


For perforated trays Van W d e (page 519 of reference 4) gives the relationship: (13.25)

velocity, ft/sec, through the holes
P A h

total hole area per tray, ? f

PL =

dry-plate pressure drop, inches of clear liquid density, lbm/fi3, of clear liquid

P. = vapor density, lbm/fi3


Tray Dynanak-Material Balance

C, = discharge coefficient that is a function of both the tray-thickness/hole-diameter ratio, and the hole-area/active-area ratio. This is presented in Figure 13.18, page 5 19, of Van Winkle.4
Thistlethwaite6has developed a correlation for C, that facilitates computations:

where T, = tray thickness and Dh = hole diameter. From the above:




For valve trays, if the caps are not M y lifted:

Once the caps have been fully lifted, the equation for perforated trays applies.

As a first step in deriving an overall tray equation, let us construct a signal flowdiagramfiomequations (13.1), (13.3), (13.11), (13.13), (13.14), (13.16), (13.17), (13.18), (13.21), and (13.24). This is shown in Figure 13.2. By successive reductions this signal flow diagram can be reduced u t l the following ni equation is derived:


13.3 DeriPatiOn .foPerall Tray Eqrcation





For all columns examined so far, the denominator quadratic has factored into two terms with substantially unequal time constants. The smaller, in most cases, has been roughly equal to the numerator time constant in the multiplier for w&). Then equation (13.29) reduces to: (13.30) where





aF ap, +--- ap, 2 aF How(A, aw9 aF at9, a 3 3
lbm/sec x sec lbm/sec




Note that K ,

has the dimensions



Tray D y ~ ~ - - M a t e d Balance

z m




8 s
a m
c ,


c ,






N "

: 0

m C m

C W-


us Gk

13.4 Mathematical Moa!.elf Combined Trays w


We can now make three generalizations about KTRand w,: 1. If KTRis positive, an increase in w,, causes a temporary increase in internal reflux. This results in an “inverse” response, since low boiler concentration is increased temporarily in the base of the column. 2. If K m is zero, a change in w, causes no change in internal reflux. This we term “neutral” response. 3. If KTRis nedative, an increase in w, causes a temporary decrease in internal reflux. This we have termed “direct)) response; it temporarily augments the long-term decrease in base low boilers caused by increasing w,,.
The term that seems to affect the sign of KTRmost strongly is dAPm/&,,. This term is almost zero for valve trays, and for al valve-tray columns we have l checked so far, the calculations predict inverse response over the entire range of normal operation. For sieve trays, dAP,/dw, is small at low boilup rates and the calculations predict inverse response. This term increases rapidly with boilup, however [see equation (13.27)] and the calculations indicate that as w, increases, the column shows next neutral response, and finally, for large values of w,, direct response.


A column with a number of trays may be represented by a mathematical model such as shown in Figure 13.3. This may be simulated readily on a large digital computer. For a mathematical analysis, however, or for hand calculations, we need a simpler model. A system with a large number of identical first-order lags may be represented T by a dead time a, which is equal to n where n is the number of lags. Therefore, the response of reflux from the lowest tray, wl, a change in external reflux to n trays away is:

To = condensing temperature, “C

TR = external reflux temperature, “C

external reflux, Ibm/sec

A simplified model for response to w, may be arrived at by a somewhat intuitive method. Consider the overflow response of an individual tray to a


Tray Llymwaia-Mated


Material balance coupling with vapor and liquid flow



step change in w,:

w(t) = Aw, K m e -t/rm Ibm/sec Tm


The integrated outflow is:

W ( t )= Aw,Km(l
At t =


(13.35) (13.36)

W(w)= Atp,,Km
is :

For n trays the total amount of liquid that is either displaced or held up
C W ( m ) = nAwKm


Since we are assuming that Aw, is felt by all of the trays simultaneously, equations (13.29) and (13.30) apply to all trays at the same time. But the outflow fiom each tray must flow through all lower trays. The time required for the bulk of the liquid displaced or held up to flow down is approximately nTm. Therefore, the average outflow rate is:

Overall, then, we obtain the approximate transfer h c t i o n for an n-tray column:

It was previously indicated that if the sign of Km is positive, the columnbase composition will exhibit an inverse response to a change in boilup. Equation (13.39) indicates the possibility of another kind of inverse response, that of column-base level. If Km is positive and if K m / ~ m greater than unity, an is i w e a s e in boilup will result in a temporary i w e u s e in base level. As will be discussed in Chapter 16, this can cause great difficulties where base level is controlled by boilup. If K m / ~ m positive and close to unity, a change in boilup causes no is change in level for a period of time equal to nTm. The control system seems af€liaed with dead time, but in reality, as shown by equation (13.39), it is not. Thistlethwaite6has carried out a more extensive analysis of inverse response in distillation columns.

1. Rijnsdorp, J. E., Birmiqgbam Univ. 2. Harriott, P., Process Conml, McGrawChm. E M . 12:5-14 (1961). Hill, New York, 1964.


Tray DynamM -Mat&


3. Williams, B., J. W. Begley, and C . Wu, Rollins, “Inverse Response i n “Tray Efficiencies i Distillation n a Distillation Column,” CEP, Columns,” final report of AIChE 71(6):83-84 (July, 1975). Research Committee, 1960, page 6. Thistlethwaite, E. A., “Analysis of Inn verse Response Behavior i Distil13. 4. Van Winkle, M., Distifhtk, McGrawlation Columns,” M.S. Thesis, Department of Chemical Engineering, Hill, New York, 1967. 5. Buckley, P. S., R. K. Cox, and D. L. Louisiana State University, 1980.

1 4

DistillationColumn Material-Balance Control


n this chapter we will look at the relationship between level control in the overhead condensate receiver and that in the column base for several different column-control schemes. Since flows are commonly measured in pounds/unit time, we will use these units instead of molar ones. Later, in Chapter 16, we will look at individual level controls in more detail. To illustrate mathematical modeling for column material-balance control, let us first use the conventional column of Figure 14.1. The feed, tpF, is split by the column into two parts: top product, wD, and bottom product, W E . It is assumed that vent losses overhead are negligible. It is further assumed that the heat-transfer dynamics of both the condenser and the reboiler are negligible; this will be true for most columns. Let us start at the column base and work up. For convenience the equations are written in Laplace transform notation.

Column Base, Including Reboiler


w l = liquid flow fiom first tray, Ibm/min w p = vapor flow leaving column base, lbm/min

1 Distillation column material balance .328 Dirtillation-ColumnMaterial-Balam Control FIGURE 14.

Ibm/min Usually vapor flow changes are propagated up the column very rapidly.2) where pB = density. pcu/lbm latent heat of process fluid. f? i H B = liquid level in base. mol/& . within the level transmitter span.4) where VR = vapor flow leaving feed tray. (AH.(LS . mol/min Vs = vapor flow entering feed tray. Ibm/ft3 of liquid in column base AB = cross-sectional area of column base. at column base = ws = steam flow.Actual feed enthalpy) Molar latent heat of vaporization of feed . Ibm. Therefore. Ibm/min WB = liquid inventory in column base.1 Mathnnntiurl Moa2l-Open WB Loop 329 = bottom product flow. mol/min Ls = liquid flow from feed tray.(Enthalpy of feed as vapor at dew point .3) where An = latent heat of steam. pcu/lbm. Feed Tray The feed-tray material balance is usually written in terms of molar flows: VR = F(1 . mol/& q enthalpyfactor .4) + (14.14. feet If the reboiler is heated by steam: (14. 5 = Laplace transform variable Next: (14. no great error is introduced by assuming that they appear instantaneously at the column top.L R ) = F L R = liquid flow from top tray. mol/min F = feed flow.).

8) Since (wf+J++.). and k4 = w F ... Ibmlmin From equation (14. From the definition of q above we can write: Lf+l . (u+/q.9) or (14. q = 1 and V.Lf = -Fq In terms of weight u i s equation (14.7) Note that: wF = feed. Ibmlmin wf = liquid flow from the feed tray. although the last may not be true if the feed composition is radically different from feed-tray composition. q is greater than 1 and V is less than V.8) in Laplace transform notation: (14. (wdLf)are constants. = V.10) wf(s) = k2wf+1(2)+ k3WF(I) + kdq(f) Usually k2 = 1.7): (14. we can rewrite equation and (14. In going back to equation (14. = L f + V. k3 = q. + F + Lf+. we obtain: (14. If the feed is subcooled liquid.4) and expressing it in weight units.6) (14. ' It is also true that: V.Note that if the feed is a liquid at its boiling point.5) (14.6) becomes: nt (14.11) whence .

15) where Ns = number of trays from the column base to the feed tray Enriching-Section Liquid-Flow Dynamics (14. k7 = 1 .17) where NR = number of trays above the feed tray a1 = N R T ~ Note that: (14. although the last two may not hold if feed composition is radically different from feed-tray composition.14.18) . T~ (no inverse response assumed): (14.k d s ) (14. each with a hydraulic lag. and k8 = wF. Stripping-Section Liquid-Flow Dynamics The transfer function between w 1 and wfis: (14.1 Mathematrcal MoakL-Open Loop 331 or Wr-l(s) = ~ w A J )+ k7~14s).16) where wm = liquid flow (internal reflux) from top tray. Ibm/min (14.q.14) where G&) is the cumulative effect of the individual tray hydraulic lags.13) Usually b6 = 1.

ftz ..HAS) WAS) PAT (14. pcu/lbm"C . To = average vapor temperature.20) where wD = top-product flow rate. pcu/lb . feet. = 1 -k -((To APT CP - - TR) wR = external reflux flow. lbm/ft3 AT = cross-sectionalarea of receiver.21) where H = height of liquid.19) where Am = latent heat of vaporization of process fluid specific heat. Condensate Receiver Material Balance fp.332 where Dhill&m-Colirmn Mated-Baluna Control K. Ibm/min Overhead Material Balance The vapor flow to the condenser is: (14. cylindrical vessel: . in receiver T pT = density of top product. wc = wtP1. Ibm/min WT = condensate receiver inventory.(S) - wD(S) S - wR(S) = w~(~) (14. "C wc = vapor flow to condenser.. lbm If the condensate receiver is a vertical.c = process fluid specific heat. "C T R = average external reflux temperature. lbm/min Note that if there is no subcooling.

we do not believe this introduces a serious error. the level controller increases the steam flow. psi lbm/min (WD)- Note: linear flow meter .c)KmbrKdrrGdt(S) x 1 KD w (14. But at high boilup rates. The second limiting factor is entrainment. external reflux.Without proper design the level controller can become very confused. steam flow. The third factor is the simplified. and let column base level set bottom-product flow. Let feed rate be set by averaging level control of the feed tank. and top. entrainment may be severe enough to invalidate the simple materialbalance model we have developed. We will assume that each level controller is cascaded to the appropriate flow controller. Limitations of Preceding Analysis There are three factors that limit the accuracy of the precedmg analysis. As can be seen. we can design any desired type of material-balance control. But this causes a momentary increase in base level due to the extra liquid coming down the column (also due to thennosyphon reboiler “swell”). Overhead Level Control The necessary additional equation (no subcooling) is: = ~T(. It is characteristic of valve tray columns and some sieve tray columns operating at low boilup rates. 14. Feed tray dynamics will be dealt with more rigorously in Chapter 18. By providing the proper additional connections.22) where Ke = - distillate flow-meter gain. If the level becomes too high.and bottom-product flows are all inputs. the feed flow. Normally we assume that the only way we get liquid overhead is by condensing vapor.2 CONTROL IN THE DIRECTION OF FLOW Let us look at a material-balance control scheme that is in the direction of flow.2. steady-state treatment of the feed tray.2 Control in the Direction of FIap 333 The preceding equations now can be combined into the signal flow dagram of Figure 14. let condensate receiver level set top-product flow.14. The first of these relates to the phenomenon of inverse response discussed in Chapter 13. It exercises its most serious effect in those columns where base level is controlled via steam flow. This is discussed in detail in Chapter 16. For purposes of this chapter.

2 Signal flow diagram for column material balance .334 DirtiuatiOn-Column Mated-BalanU Control FIGURE 14.

23a) For pneumatic instruments: 12 psi (14.14. psi/psi = G.2 G m ~in the Direction o Flow l f ( I V ~ = top-product ) ~ ~ 335 flow-meter span.3. lbm/min Kdrr controller gain. in feet. For averaging level control. this introduces little error.T)Kdf (14.25) FIGURE 14. We can now define a characteristic time constant: K& = [THIT = PTAT(AHT)T &~(wD x) a . psi/fi KdT Note that we have ignored the dynamics of level measurement and of the flow control loop. From this we can see by inspection that: WD(4 - 1 1 + IVd5).23) ATPTK~ = &d%l7-(5) (14. (14.(s) = = controller dynamic gain receiver level transmitter gain. We can now prepare the partial signal flow diagram of Figure 14.IVR($) ~chrGchr(.24) (AHT T ) where ( A H T ) T is the level transmitter span.3 S g a flow diagram-condensate receiver inl . for a 3-15-psig output.

psi/psi KmhB base-level transmitter gain.30) where (AHT). If a proportional-only controller is used.4. Ibm/min = controller gain. Note that for best flow smoothing or to another process step.is the base-level transmitter span in feet for 3-15 psig output. then we must achieve the desired [ 7 H ] T by proper choice of A T or (AHT)= both.23) becomes: (14. say 2-5 minutes (120-300 seconds).29) 1 + KC~BGC~B (s)Km/l~ fPl(4 .fp&) PdBKngrB KHBGHB(s) By analogy with equation (14.28) where KnsB= bottom-product flow transmitter gain. = KdB psi Ibm/min 12/(~~)~=. This is large enough to ensure that instrument and pneumatic transmission-line dynamics i will not be sigdcant. psi/fi = We can now prepare the partial signal flow diagram of Figure 14. rR[TH]Ts2 +1 + rl?.27) Base Level Control The necessary additional equation is: (14.fpR(s) where rR is the level controller reset time in minutes.29a) (14. .. If we fx K. equation (14. as indicated in Chapter 16: tpg(s) pc(c(s) = TRs 1 .25) may define: we = (14. as Usually it is desirable to have [rHIT small as convenient. it is recommended that KAT = 2 be chosen.. even an additional or buffer or surge tank. one may need a larger (rH)*.336 DiniUation-Colurnn Matffial-Balanu Cmttvl If a proportional-only control system is used.s -k (14.fiom which we can see that: wB(s) 1 (14. (fpB)- = flow-meter span.26) whde for proportional-reset level control it becomes. for best control of the associated column.

3 Control in Dire& Opposite to Flow 337 For proportional-only control.31) WB(S) - TR$ +1 tPl(S) .tp. this statement would still be true. and if KdE is specified.4 Signal flow diagram-column base .fpp(S) TR[TH]BS’ + TRS + 1 (14. base level sets feed flow. As before. level controllers are cascaded to flow controls with linear flow meters.(S) while for proportional-reset control it is: WB(4 1 [THIBJ Wl(4 +1 (14. and both reflux and bottom-product flow are ratioed to top-product flow.5 where top-product flow is the demand flow. FIGURE 14. It is usually desirable for best control of the associated column to make ( T ~ = 10-15 minutes.29) becomes: . both. tank. one may need a large ( T ~ ) or even an additional surge or buffer ~ . If we were to add reflux/feed and steam/feed ratio controls. or KchB= 2 is recommended. For proportional-only control. equation ( 14. condensate receiver level sets steam flow. Note that the two level controls are independent and noninteracting. We can now prepare the overall closed-loop material-balance dagram of Figure 14.14.3 CONTROL IN DIRECTION OPPOSITE TO FLOW As an example let us choose the case of Figure 6.5.32) where T~ is the level controller reset time in minutes. then the proper time constant ) ~ is achieved by choice of ABor (AHT)B. 14. Note that for best flow smoothing to another process step.

338 DtitiUatMn-Column MaterialB&m Control FIGURE 14.5 Signal fo diagram-material balance control in direction of f o lw lw .

the steam flow transmitter should have a linear relationship between flow and transmitter output. it is redrawn into the form of Figure 14. Also.14. This is a much more complex diagram than that of .qj = steam-flow transmitter gain.36) Closed-Loop Signal-Flow Diagram The closed-loop signal-flow diagram of Figure 14. Base Level Adjusts Feed Flow (14.7.33) PSI K?fj where K.6 may now be prepared. since we have a cascade system.. Bottom-Product Flow Ratioed to Distillate Flow (14.34) where = feed flow-meter gain. If an orifice flow meter is used. Ibm/min 12/(w5). lbm/min Nore that we assume the flow control loop to be very fast compared with other dynamics. = psi lbm/min 12/(%)max lbm/min ( B J ~=) feed flow-meter span. where (w5)- = steam flow meter span. the AP transmitter should be followed by a square root extractor. ~ ~ Reflux Flow Ratioed to Distillate Flow Let: Q(5) = ~RDGRD(-T)fPD(5) (14.35) Physical techniques for accomplishing this are discussed in reference 2.3 Control in Direction Opposite to Flow 339 Condensate Receiver Level Cascaded to Steam Flow Control The necessary equation here is: 1 %(5) = -K c h T G & T ( 5 ) W 5 ) (14. To show the relationship between wD and wF more clearly.

340 Disdla&a-Column M a t d .6 Signal flow diagram-material balance control in direction opposite to flow .B a l a w Control FIGURE 14.

0 C a 0 e 2% -m g gs 3 k gs .14.3 Control in Direction Opposite to F h 341 9 r -- L 8 g Lr.

not composition control. Base level adjusts side draw and reflux drum level sets reflux flow. control functions hs must be correctly preselected. it is probably apparent that conventional “tuninjf‘ procedures are essentially useless for a system of t i complexity.39) and (14. and (2) liquid sidestream.5.)max psi Ibm/min -.Figure 14.w The steam flow is set by ratio to the feed flow: fP&) = fPv(4 (14. Also. 14. About the only new relationship we need is that which defines vapor flow up the column above the point of side draw: tp. and control-loop parameters calculated ahead of plant operation. a small bottom-product purge. Three flows are ratioed to feed: top product. B ( J ) K+ = sidestream flow. This point of view is at variance with that sometimes expressed elsewhere in the literature.4 MATERIAL-BALANCE CONTROL IN SIDESTREAM DRAWOFF COLUMNS Let us consider two cases: (1) vapor sidestream.12 - and . The functions Gm(s) and GBD(s) have to be chosen with care will because of potential difficulties with stability.(s) = KubB -K h B G c h B ( w . Vapor Sidestream As an example let us consider a column such as that illustrated in Figure 7.1. Feed is assumed to enter at its boiling point. bottom product. these two functions must be chosen with primary regard for material-balance control. Finally.This column has a small top-product purge. Ibm/min where (14. (tp. and a side product that is most of the feed.fP. and steam..38) (14.37) K+ = sidestream flow-meter gain.40) .

There are many columns operating today with condensate receiver level controlling reflux and base level controlling bottom-product flow. assuming that we cannot. we can prepare the signal flow 49 diagram of Figure 1 . usually steam. . Another expert recommends the second. WF(4 (14.. for some reason.41) K?@ 1 and %S () - K$F&&&)- rc. How do we choose between them. There are other columns in which condensate receiver level adjusts top-product flow while base level manipulates steam flow. It is probably apparent that we cannot follow both recommendations. Controversy also exists as to whether it is better to have column-base level control bottom-product flow or the reboiler heating medium.45) With the same control scheme in mind.42) The pertinent equations may now be assembled into the form of Figure 1 .14. Inventory in the receiver may be regulated a little more readily by manipulation of the large flow than of the small one. . One well-known author argues strongly for the former. In a superfiactionator. have both levels adjust drawoff flows? It seems to us that it is largely a matter of convenience.5 TOP AND BOTM)M LEVEL CONTROL COMBINATIONS Considerable controversy has existed on the question of whether to have the condensate receiver level adjust the reflux flow or the top-product flow.43) (14. It does . 14.5 Top and Bottom Level Control Com&u&ms 343 The two remaining ratio controls are defined as follows: 1 WB(S) = K#$FKR3GR3(5)- (14. 48 Liquid Sidestream In this case we will assume that the liquid side draw is taken from a point above the feed tray. for example. This does not mean that level control via the small flow is either impossible or impractical. at least one of the two levels must control a drawoff flow. the reflux flow may be ten or more times greater than the topproduct flow. Then we define: (14.

344 Dutdhthn-Column Ma&riul-Balana Control FIGURE 14.8 Material balance signal fo diagram-vapor sidestream drawoff lw .

14.5 Top and Bottom Level Control Combinatwns 345 FIGURE 14.9 Material balance signal flow diagram-liquid sidestream drawoff .

Since reflux is controlled by level. Controhg base level by steam has another disadvantage if a thermosyphon reboiler is used. there will be a certain change in reflux flow and another change in top-product flow required to restore the top composition. generally. that we cannot use averaging level control. It is sometimes argued that where reflux flow is much greater than topproduct flow.and low-level protection by means of overrides on reflux flow would be needed. This means. a change in feed rate or feed composition changes overhead composition. we must design for tight level control. one may control top composition more easily by adjusting topproduct flow than by adjusting reflux flow. Thls is discussed in Chapters 9 and 16. This is discussed in Chapters 4 15. and that the differenceis against the argument rather than in favor of it. Composition control via distillate (top product) has the disadvantage that no change in composition takes place u t l the reflux flow changes.for example. and 16. For this reason we normally prefer to control composition via reflux. Actually a little algebra will show that there is not much difference. the ni dynamics of the level control loop appear in the composition control loop. and leads to the conclusion that we would normally prefer to control base composition by manipulating boilup. .346 Dktdhtim-Column Materiul-Balance Control mean that high. These two required changes are the same in the steady state regardless of which variable is manipulated to control top composition. . A similar line of reasoning may be followed at the base of the column. If. interchange of inventory between column base and reboiler sometimes leads to severe dynamic problems.

we find in the chemical and petroleum industries two principal types of liquid-cooled condensers: (1) the horizontal type with vapor on the shell side and coolant in the tubes. and (2) the vertical design with vapor in the tubes and coolant on the shell side. An approximate analysis for a condenser that has a single pass on the coolant side is presented in Chapter 24 of reference 1.1 LIQUID-COOLED CONDENSERS WITH NO CONDENSATE HOLDUP s mentioned in Chapter 3. Although these reduce the complexity somewhat. 2. and will not be repeated here. Heat storage in the heat exchanger metal is negligible. We can also think of condensers in terms of whether the coolant goes through just once (no axial mixing) or is recirculated to achieve good axial mixing. latter type of cooling is said to have cctempered” Condenser with No Axial Mixing of Coolant Once-through coolant is by far the most common choice. where subcooling may be of concern.1 5 a Condenser and Reboiler Dynamics 15. much simpler model. For important applications. Subcooling is negligible. one should probably resort to simulation. A condenser with the cooling. It involves the following simplifjring assumptions: 1. we will not pursue their dynamic equations further. we are still lefi with the job of solving partial differential equations. Mean temperature difference is arithmetic.~ most liquid-cooled Since condensers are fairly fast with time constants in the range of 10-60 seconds. Perhaps the easiest to read descriptions . has been proposed by Thal-Lar~en. of their solutions are those by Hempe12 and by G ~ u l dA ~ largely empirical. 3. 347 .

pcu/lbm. and Ja~dEet. A mathematical analysis of its dynamics is given in Chapter 24 of reference 1. fcz Condenser with Well-Mixed Coolant Qualitatively the condenser with well-mixed (tempered) coolant is discussed in Chapter 3. Archambault. one paper has been published by Luyben. O K coolant specific heat. We will not repeat it here. but it leads to the following results (simphfjmg assumptions are the same as in the previous section): . Ibm/sec latent heat.2) (15. O coolant inlet temperature.348 Gmdenser and Reb& Qparnia The literature on subcooled condensers is very sparse. O K coolant exit temperature. pcu/lbm "C K condenser heat-transfer coefficient.4) process condensing temperature. the following static gains may be derived: (15.~ another by Tyreus. pcu/sec "C fcz rate of condensation.19).6 and If the sensible heat load of subcooling is not too large compared with that of the condensing heat load (and this is usually the case). Section 9 (see Figure 3. of process vapor condenser heat-transfer area. Ibm/sec coolant flow rate.

As the liquid level in the shell varies.12) The first-order dynamics of this type of condenser make it much easier to control than the condenser with once-through coolant.10) (15. Also: ( 15. and rate of removal of liquid fiom the condenser shell.8) (15.1). Ibm The subscript OL means open loop.9) where w. so do heat-transfer area and rate of condensation.15.2 FLOODED CONDENSERS-OPEN-LOOP DYNAMICS Most flooded condensers are of the horizontal type with vapor on the shell side and coolant in the tubes as shown in Figure 15. We would like to find out how condensing pressure and rate of condensation are affected by rate of vapor flow into the condenser.2 Fhded Co&men-Open-Loop Dynamics 349 (15. of coolant coolant holdup. O K .7) (15. 15.1 (see also Figure 3. WA average bulk temperature. We will make several simplifylng assumptions: . back pressure downstream of the vent valve.

2. is proportional to liquid level above the bottom of the lowest tube. A. Submerged heat-transfer area..” To make this assumption valid may require that in some cases the tube bundle be slightly rotated about its axis. we will assume that the vent valve position is fixed. 3. is then the negative of the change in A. FIGURE l S .350 Ce &r and Reboiler W m i a 1. 4. The sensible heat load is small enough in comparison with the latent heat load that it may be neglected. This assumption is not bad if there are many tubes and if they are not “layered. A. l Horizontal condenser with coolant in tubes and partially flooded on shell side . Heat storage in the heat exchanger metal may be neglected... For the time being. The change in condensing area.

downstream of vent valve vent valve resistance. pcu/lbm. Ibf sec/fi5 pounds of liquid in shell between bottom of lowest tube and top of highest tube total heat-transfer area. "C or O K average coolant temperature. lbm/sec vapor inflow. ft3/sec ~ u.20) (15. ft2.21) where v= qc = = system vapor volume. f? i condensing temperature. lbm/sec vapor outflow. of condensing process vapor liquid outflow.2 F h h d conrlensers-Op~-Loop -b 351 We may now write the following equations.14) (15.15) (15. pcu/sec condenser heat-transfer coefficient. lbf/ftz.17) (15. ft3 rate of heat transfer. of condenser flow.15. pcu/sec fi? "C condensing heat-transfer area. lbm/sec submerged heat-transfer area. = T = wc = wo = s= = = ww A.13) (15. A. some in the time domain and some in the s domain: (15. p "C or O K Ibm/sec process vapor condensed latent heat of vaporization. Rv = = = ATC Q = . "C or O K vapor space pressure. = Tc = TA = Pc = PR = w .19) (15.18) (15. lbf/fi? pressure.

to Various Inputs To find the response. one of which has a typical time constant of 1.25) The first term on the right-hand side of equation (15.+Kco Fc a - s+ 1 U A SIAp CC TA(s)+ &PR(s) s+a R V + w&) where 1 (15. and (15. while the other is only a few seconds.2. Pc(s).24) Response.(s). various inputs.3 and the final reduction in Figure 15.25) may be written: KC F Tis2 +2 ( 1 + a) 5 ~ ~ 1s + For all such condensers we have studied.5-3 minutes. we combine equations ( 15. (15.4. P.352 Chdenser and Reboiler Dynamia Upon Laplace transforming equations (15.23) (15.13) to through (15. the denominator has a large damping ratio so that the quadratic may be factored into two terms.16).22) qc(s) = UcZc AT(@ + UchTAc(s) (15. we obtain: (15. The first reduction of this is given in Figure 15.24) into the signal flow diagram of Figure 15.13). From this we may write by inspection: RvlPv ( s PC(S) = [: a + a) - -Rv a -2 + 1 V + d .14). The above then reduces to: KC F where (5 TFCS +1 + a) .

-c 3 0 e 2 LL $e .2 F h k d COnrtenret-Op~-Loop +h 353 i l U C 5 8 B TI e IC 0 0 0 4 " L 4 2 m - e 0 U E 2.15....

354 Gmhmcr and Rcbodc~Dynamia 2 r 2 w0 s E 3 U E b : I B F cn w0 G E -2 vs i -3 ZE! Q gg .

8 33 2s LLr.15.2 F W condnrr~--Optn-Loop w h 355 i U C 5 8 B U 0 0 G rc 0 4 ” e E P U L E rnm r e rn ZTi sG . .

+ w&)] uAd’TA(s) s+a The first term on the right may be written: Kkc (s + a) S(&S 1) (15. the last equation reduces to: PC(S) = W P V ) (s + 4 s + 1) uczc aTc avpc+ -& Pv a r c + . fiom which we can write by inspection: .(s).356 Gmdenscr and R e W LIpunia Case Where R.26) + where Kkc and = ucxc aTc avpc+ --r Ecp P v a c 1lPv Response. w.2 can be redrawn to show wC(s)as an output as shown in Figure 15.5. Approaches Infinity For the case where Rv becomes very large (very little vent gas flow). This can then be reduced to the form of Figure 15. Various Inputs to The signal flow diagram of Figure 15.6.

Heat-TransferDynamics The combination of thermosyphon reboiler (or any high circulation rate reboiler) and column base or separator may be represented as shown schematically in Figure 15. . becomes very large. Approaches Infinity Again. The analysis should take into account the temperature of the entering liquid and the sensible heat effect of the liquid mass. the preceding equation becomes simpler: X wc(5) = L .5 VIR $ ucAc arc a PIPc + P P h 1 + iJ arc (15.28) 15.3 Reboilrrs-Open-Loop Dynamh 357 where L($Rvr a AP + 1) 1 + ( Y .R .7. as R. + Kco Fc $2 + P P C a a s + l X i c ( g R v s + 1) 1 (15.3 REBOILERS-OPEN-LOOP DYNAMICS We wish to make an analysis of column bases with associated reboilers where there is signrficant liquid holdup. The various equations may then be written as follows. Such an analysis also applies to vaporizers with associated separators or knockout drums.27a) Case Where R.15.

358 condmrrr and Rcboilcr Dynamurr a U C a l C 8 -8 U G 8 IC 0 3 7 L 8 E E P U 6 8 2i 7i -c W P a* E=I?! &&I .

15.3 Rthih~--Opnr-Loqp QVUUU~ 359 i U C 0 0 5 3 0 e 1c 0 O s 8 E m I U L e e B I I 2s w-fn C W%! gs gz .

31) (15.7 v Schematic representation of column base and reboiler holdup .36) (15.33) (15. ) P 4T- cp @BU TBu(s) - cp T B U wBu(s) cp WB TBdS) - cp T B U fPg(s) = cp 5 [ T B U W B ( S ) + W TBdS) 3 B (15.(s) = -PCS(5) 8PCS (15.35) (15.T ~ 4 s ) l (Note that heat storage of the reboiler metal is neglected.32) [vi($) .hrp f B .34) (15.wBu(s) - WB (J)] 1 Steam-Side Dynamics The following are taken fiom Chapter 25 of reference 1: aC TS Tc.360 Gmuhcr and R e w m b t By using perturbation techniques and Laplace transforming this equation we get: cp Ti fpi(s) + cp @i Ti(s)+ qT(s) .) qT(5) = FIGURE 15.30) Material Balance or j [wi(t) - wBU(t) - wE(t)] dt = WB(t) .37) Pcs(s) = Qi(5) Qi<s> - Q4s) CR5 -ps(s) Qv aps = + -pcs(~) + -Qv v ( s ) aQv X dPCS Z V Q&) = -- qds) 1 h t Plt URAR [Tcs(s).= WB(~) 5 (15.

41) . where V. lbf/fi? Q = steam flow rate. aQv/aP.8. pcu/sec p/"C UR = reboiler heat-transfer coefficient.9 where three new functions are defined: 1 (15..38) column acoustic impedance. This can be partially reduced as shown on Figure 15.40) xPsc As2 1 (15.. lbf/fi? From the preceding equations we can prepare the signal flow diagram of Figure 15. = reboiler shell volume. "C B = rate of steam condensation. lbm/min column-base pressure. upstream of control valve = valve stem position The terms aQv/aP.3 Reboilnr-Open-Loop LlynamS 361 In these equations: Tu@)= steam condensing temperature. ft3/sec qT = heat transfer. f? T U = boiling temperature of process fluid. lbm/ft3 rate of boilup. fi3/sec = steam latent heat of condensation.. and aQv/dXv may be evaluated by the methods of Chapter 15 of reference 1. "C P&) = reboiler shell pressure. ft2 sec i A R = heat-transfer area. pcu/lbm = steam density Ibm/ft3 at Pa and Ta = acoustic capacitance of reboiler shell. Ibf/fi?. fi5/lbf = VJT' . ft3 = steam supply pressure.15. The column-base pressure dynamics may be represented by: (15. looking up from the base. lbf sec/fi5 density of vapor boilup.

A(5) has been so small.362 G m h m and Re- Dynumia For all reboilers examined to date. FIGURE 15. Heat-transfer lags are typically only several seconds. It therefore will be omitted in the remainder of this book.8 Preliminary signal flow diagram for heat transfer dynamics . We have also found that the sensible heat effect of the liquid mass in the r column base o separator is small. vapor flow finm the separator blows steam flow almost instantaneowly. as to be negligible. It is advisable. however. both statically and dynamically. to calculate A(5) for any new system as a check.

T4s) = 0 . fi? A B PL = liquid density.15. We may then prepare the signal flow diagram of Figure 15. FIGURE 15. lbm/fi3 Since Ti varies slowly or not at all.10.3 Reboden-Open-Lmp Dynamia 363 Base Level Control Cascaded to Steam flow Control We assume here that averaging level control is desired. Note that: KCf Gcf(s)= flow controller transfer function = liquid-level transmitter gain K mh KchGc&) = level controller transfer function = cross-sectional area of column base.8 .9 Partial reduction of figure 15.

Note that: 1 - CR1 W) = 1+-- 1 1 Pa CR5 A. then: 1 1 . 4 aTm (15.43) FIGURE 15. =0 Flow control loop gain and dynamics are determined entirely by the instrument characteristics.10 Signal flow diagram for base level control cascaded to steam flow control .42) ap.W) since v - = 1 (15..364 condenrer and Reboiler Dynmzia Noncritical Versus Critical Steam Flow If steam flow is critical. (b ( ap. QV aQ arc.

we can prepare the signal flow diagram of Figure 15. one may substitute l/K+for the flow control loop.12. . aQ. Practically speaking ~ ( s ) . Further Mathematical Simplification It has already been indicated that for averaging level control cascaded to steam flow control. a Q V / e S Signal F o Diagram Simplification lw Since we have called for averaging level control. As we will see in Chapters 16 and 17.44) This leads us to the final signal flow dagram of Figure 15. Gf (4 B(s) ax. and A($) reduce to constants. Note that: K V GV(4 Kcf aQ.W) has lead characteristics and a static gain of less than unity. To avoid problems with flow controller tuning.9 and by assuming A(s) = 0. permitting higher loop gain. 1 1 . 1 1 K+G*(s> K+ (15./dPa = 0.11. the system should always be operated in one flow r e p e or the other. X 5 - Base Level Control by Direct Manipulation of Steam Valve From Figure 15.. Note that if steam flow is critical. 444.. the natural frequency of the level control loop will be much lower than that of the flow control loop.365 If steam flow is noncritical. it usually will be possible to use much simpler reboiler models than that discussed in this section. Use of the mathematical models discussed here on commercial reboilers indicates that typical time constants range from a fraction of a second to 5-10 seconds. faster flow-control response. In most cases steam flow is noncritical. and much higher flow controller gain.

5 for a flooded condenser may be used as a starting point.4 PARTIALLY FLOODED REBOILERS The partially flooded reboiler is similar in many ways to the partially flooded condenser. FIGURE 15. which in t r varies the condensate level in the shell and thereby the heat-transfer area for condensation. it is a vertical thermosyphon reboiler. In this case pressure dynamics are essentially negligible. Usually. As discussed in Chapter 4. Section 2. If the steam supply pressure is constant. That area covered by liquid permits only sensible heat transfer h m the condensate. The signal flow diagram of Figure 15. it is controlled by throttling the steam un condensate.366 CDnrinrer and Rebode-Dynmnia 15. the steam condensing temperature is also constant.11 Final signal flow diagram for base level control cascaded to s e m flow control ta . t i is a small heat load compared hs with that of the condensing steam and is treated as neghgble. although not always.

P i mn = average exposed heat-transfer area.4 PartiaUy F h U Rebodem 367 The signal flow diagram of Figure 15. ft2 = condensate on shell side. pcu/lbm p/"C = heat-transfer coefficient. lbm/min rate of steam condensate withdrawal.45) where w c w o Ast UR x7 i aA.ART ~WSHI= total heat-transfer area of reboiler.13 may be prepared next.MR UR ATa sf wI wo(s) (15.12 Signal f o diagram for base level control by lw direct manipulation of steam valve . lbm/min = steam latent heat.14: WC(S) -- - 1 A* s + l . aS wH ART WH S rate of steam condensation. . As shown by the reduced form of Figure 15. f ? for steam condensate . lbm = = FIGURE 15.15.

-0 0 0 E e E z U L E G 5m .368 Chndenser and Reb& Dynamia L a . a 0 n 2 U a . Gi? w m c la- m c g E5E a a E2 .

15.g r 3 UJZi Mal U 5s La SZ .4 P d y F h h d Reboden 369 L 0 n z U 0 0 0 2 e L E F '3 1 E 0 G .

Solving equation (15. From Figure 15. and therefore in wc. we would intuitively expect a step increase in AT to cause an initial increase in heat transfer.49) In equation (15.49).Tsu. .46) Commercial lead-lag compensators commonly have values of a between 6 and 30. pcu/lbm TT . The former are usually much less expensive.47) wc( t )is a perturbation variable (deviation from steady state).Condensate level. however. heat. of boiling liquid. "C S Design experience with flooded reboilers is limited but indicates that typical time constants are of the order of 2-5 minutes. were held constant. As shown by equation (15.Condmer and Reboiler Dynmtia weight of condensate that will fill shell side of reboiler.47) If w. Ibm/min laten..47) for a step increase in AT shows this to be true: so that: (15. Simulation studies show that substantial improvement in response speed may be achieved by lead-lag compensations with transfer functions such as: TD --s+ a 1 where we let: (15. It is interesting to look at the response of wc to a change in AT. Some provide a fixed a and some have adjustable a. Ibm boilup. would eventually increase (thereby decreasing&) and eventually wcwould have to equal w.14: (15. it decays to zero as time goes to infinity.

(5> P L (BL/Bc) (15.51) = ---&) aT.we throttle the steam instead of the condensate. .1): HLPL BL = Hs pLBL + P. if we assume that the line loss Mhne negligible.50) (15. (15. WBW -41.wc(5) ) 5PY (VRJ (15.54) q ~ ( t= UR R ( ~ ) ) A AT(t) AT = T. ap. ABW .5 Partially Flooakd Rebodem fm Low-Boiling Materials 371 15.55) (15. H45) = . feet Pc.TB. we discussed a variation in flooded reboiler design for low-boiling materials. feet L Hs = condensate level in the shell.56) (15.52) ( 15. From equation (4. (15.57) (15.5 PARTIALLY FLOODED REBOILERS FOR LOW-BOILING MATERIALS In Chapter 4. lbm/ft3 The vented loop seal maintains P.53) ar H ART [Hrlrnax &(5) = a R~Hr(5) U . The static force-balance relationship is given by equation (4. lbf/f? abs p L = condensate density. then: 8.58) wc = @ A . = steam pressure in shell. is and that HL pLBL is constant.R a =- -P.15.just a little above atmospheric pressure. Section 2. We can now write the following equations: PcS(s) = Z(5) ~ 4 5 . As shown by Figure 4.4.1). B C B C where + APLine (44 H = loop seal or standpipe height.

15 Signal flow diagram for flooded reboiler for low boiling point materials FIGURE 15.372 Cbnhser and Reboiler Dpunia FIGURE 15.16 Reduced signal flow diagram for flooded reboiler for low boiling point materials .

63) Also: (15. Hempel. A. Ibm/ft3 V = average free volume in shell above liquid level. 1964. A.60) (15.61) We may now prepare the signal flow diagram of Figure 15. New York. Chemiurl PromControl. P. 3.. Ibm/min p I = steam density in shell.244 (1961). . which may then be reduced to the form of Figure 15. To get equation (15. Ibm/ft3 Other terms are as defined earlier in this chapter. Wiley. 2.55) into the s domain. write: (15. From the latter we see that: (15.373 where ws = steam flow. if there were a step change in Tsv. Gould. "C p L = condensate density. REFERENCES 1.16. S. TramASME. L. Techniques o Pmcea f Control.62) where (15...64) As in the previous section. Ibm/min wc = rate of steam condensation.15. ft3 Tu = condensing temperature.59) where (15. which would slowly decay to zero. there would be an immediate spike in wc. Buckley.

Reading. D. H. June 5. J. 19 (5): . JaUffiett. 22. Condensers. ..374 W e r and Reboiler Qynamaa Addison-Welsey. M s ..” presented at ACC meeting. Luyben. San Francisco. L. a s . P. 1969. Tyreus. 1983. P. Archambault. ASME paper 59-Adation of Vertical Subcooling 117.. and J. 6.AIChEJ. B. “Modeling and Sim4. 923-928 (1973).Calif. Thal-Larsen. W.

and sometimes steam condensate receivers.’ As stated in Chapter 1. the chief function is not that of holding level constant.2 LEVEL CONTROL OF SIMPLE VESSELS For a simple vessel such as &own in Figure 16.1 INTRODUCTION or distillation columns level control refers. column bases.1where the level is controlled by outflow and there is no sipficant level self-regulation effect. radians/minute. or (2) the tank is to be sized and the level control system designed to achieve flow smoothing adequate for downstream composition controls. On new projects the engineer is confronted with two alternatives: (1) the tank or holdup size is already specified and the problem is to get maximum flow smoothing. Before getting into specific applications of level control on distillation columns. Maintaining inventory or accumulation between an upper and a lower limit (not at a fixed value). Providing for smooth and gradual changes in manipulated flows to avoid upsetting process equipment. Balancing inflows against outflows at a point in a process. 2. in most cases. to overhead condenser receivers (reflux accumulators). let us review briefly the theory of averaging level control on simple vessels.1 6 f Liquid Level Control 16. where wR is downstream closedloop resonant fi-equency. For most of these applications. surge or feed tanks. 16. but rather of achieving the smoothest possible transitions in manipulated flows in response to disturbances. the b c t i o n s of averaging level control are: 1. This is “averaging” level control. 3. we need only a few equations: 375 . For the latter we make rH3 10/wR.

= valve gain.2) (16.4) Q ~ J=)4 2 0 64s) de. = controller output signal fi3/min a / d 8 .9 and 4. feet of process fluid Kd G&) = controller transfer h c t i o n 0. fi3/min = vessel cross-sectional area.1 Level control of simple vessel .(s) = K&G&(s) x ed(4 (16. fi? (vertical.1) = LbH(s) e.3) (16.Q4-t) = H($) Ltquid Lmel Control As &PI&) (16.7) For a cascade level-flow system: a Ql e d o c = 1 L f where K * = flow measurement gain of linear flow meter = ($3 FIGURE 16.376 Q4s) . fi3/min outflow. feet k b = level transmitter output signal L b = 12 psi/AHT for pneumatics M . where = = inflow. H = level transmitter input span. cylindrical vessel A assumed) H = liquid level. psi (see &scussions in Sections 3. or flow control loop gain.

16.4) may be combined into the signal flow diagram of Figure 16.9) QFS This last equation is very useful for finding the desired holdup. Then equation (16. then: A 4 0 [ h & A KJchdB.2 Level Control of Simple = Vesseh 377 where m].7) and equation (16. Equations (16.5) becomes: (16. as are transmitter dynamics. and if the valve has an installed linear flow characteristic with a wide-open capacity approximately equal to four times flowsheet flow.5) and (16. flow of flow-meter span. GS.. from which we may write by inspection: (16.2. for example. 3-15 psig). [TI 4Q3 - A M T Kch .1) through (16.6) Proportional-Only Control For this case KchGch(s) becomes simply Kch. A AHT.6) becomes: If the input span of the valve positioner is the same as the transmitter output span (as.TH (16. ft3/min maximum The analysis employed here is further simplified in that the effects of variable valve-pressure drop are omitted.. provided a value of rH is specified. Usually we choose Kch = 2 and bias the .

378 Liquid Level G m m l % E z? & c . E C I I 9 f 8 4 E VI Q) g L P NU $B W = E E 55 urn Gii .

01 @Sample Vesseh 379 proportional controller so that the tank level is midscale on the level transmitter. Since the unenhanced PI controller with Kch< 1 does not ensure that the tank will not run dry or oveiflow. 3. however. 3. Note that the level transmitter span is usually less than the nozzleto-node spacing. Level control is cascaded to flow control. a large tank will be required since for a given rH. Although in theory there exists a number of controllers that permit one to use Kd < 1. the PI controller enhanced or augmented with auto overrides has provided an almost foolproof way of keeping liquid within the vessel. “Nozzles” here refers to &ose used for connecting the level-measuring device to the vessel. the valve may not be closed when the level is at zero.2 Level Conh.9b) For level control cascaded to flow control: A AHT = V T = THKch (Qo)rnax (16. With floating pressure columns and with the trend toward small control-valve pressure drops for energy conservation. Normal set point for the PI level controller is midscale of level transmitter span. This allows for some variation in liquid specific gravity. inexpensive. If tank size is not specified but is to be calculated from equation (16.16. If tank size is specified. Consequently we usually specify KA = 2. Nozzle-to-nozzle spacing is so chosen that process operation will be satisfactory with the level at any location between the nozzles. if an orifice flow meter is used. much more flow smoothing for a given size vessel than will a proportional-only controller. In recent years. this is virtually mandatory to counteract the effect of control valve up. use of Kch= 1 is risky. past practice has been to have high and low alarms or high and low interlocks.9~) Proportional-Reset Control Although the proportional-only control system is simple.and downstream pressure variations. it must be followed by a square-root extractor. the flow smoothing is limited by rHbecause one cannot safely use Kch< 1.9c). 2.9a) or (16. Because mechanics ofien do not calibrate valves precisely. 2. under most circumstances. Then: (16. m T / 2 . and almost foolproof (at least when implemented with fixed-gain relays). This is required for proper functioning of the auto overrides. it has limitations: 1.VTis proportional to K d . that is. This leaves the top 25 percent of the transmitter span for overrides.” Before exploring the theory. the PI controller has been most popular. It provides. . The flow measurement must be linear.9a) A AHT = v~= THK*( 4 % ~ ) = STWQFS (16. let us make some additional design assumptions: 1.

.10) ?-as+ 1 (16. but the former permits more accurate. (16. Other overrides are shown in both the level controller output signal path to the flow controller set point and in the flow controller output signal path to the control valve. This means that the controller output at its maximum value verges on being taken over by an auto override.” This assumption permits us to use Laplace transforms and frequency response. But regardless of override location (other than auto Overrides). we get: K-. As a result of a number of studies (unpublished). The latter arrangement (overrides in the signal path to the control valve) has been far more common.12) . This may or may not be desirable. The transfer function for a PI controller is: where rR is the reset time in minutes. quantitative design. but a level indicator is desirable./QC by 1/K4.380 Liquid Level Control A schematic for PI level control on a simple tank is given in Figure 16. Substituting this into equation (16. we provide a switching design that causes the level controller reset to be bypassed (i. The former also implies that whenever the flow control station is not switched to “remote auto.5) and replacing aQ. we have concluded that auto overrides with gain 2 and a controller tuned for a damping ratio of one are optimum for most situations.25.3.” This virtually eliminates “bumping” when the flow controller switches fi-om either “manual” or “local auto” to “remote auto. we must resort to digital simulation. level controller has very fast reset) whenever the flow control station is not switched to “remote auto.11) (16.” A primary control station is not necessary.” the overrides are out of service. If it is desired to predict system behavior when forced by disturbances large enough to cause an override to take over.e. We have also found for most cases that the PI level control system so designed functions in a h e a r manner for step changes in in ut flow of up to 10 percent of span of the manipulated flow if Kub3 Po 0. In the a d y s S that f i l h we will & w e the role o ovevrides and will assume f that the PI level controller is always “in mmund.

transmitter and valve dynamics become sigdicant.12) and (16.14a) Let us also define: (16. The damping ratio. ~ . The damping ratio preferably should be at least unity.16. one increases Kd.causes severe peaking in the frequency response in the vicinity of the closed-loop natural frequency.13). while holding TR constant. and the loop eventually becomes unstable. E. as shown by equations (16. from equation (16. on the other hand.15) we can see that if TR is fixed. usually they are designed to be much faster. approaches zero and the control loop becomes very resonant. TQ becomes small.12) and (16.13) Now the denominator of equations (16. TQ becomes very large.2 Level Control ofsimple Vessels 381 Similarly. This resonance is sometimes called a “reset cycle” since it would not exist if the controller did not have automatic reset.6): (16. Since the loop approaches instability for both very large and very we small values of Kch. l / Flow and level regulation in that frequency range will be very poor. This is commonly called a “gain cycle” since it is caused by excess gain. say that it is umdzthnahy stable. In desigmng a level control system with a proportional-reset controller several practical considerations must be kept in mind: 1. then as one decreases Kchtwo things happen: 1. 2. The loop therefore becomes slower and less stable at the same time.15) (16. It is a quadratic whose damping ratio is : or TR = 4 t2TH (16. A low damping ratio. 2.15a) From equations (16.13) has some interesting characteristics not widely appreciated.14) and (16. approaching instability. Adjacent or related process controls must be designed with closed-loop natural frequencies much different from that of the level control. 5.

Kd. for a damping ratio of unity. 5 = 1 should suffice and places less of a burden on available controller settings. ai.0 or use a proportional-only controller. For a given dampmg r t o reset time must be increased as Koh decreased.It has transfer functions very similar to those ~’~ of the PI level control. is It frequently happens that for a desired and specified ICd. To protect upper and lower permissible level limits. For most other applications.18) TQ = 2 TH = and TQ = 4 TH = TR/4 (16.14 for5 1. For and 5 = 1 it can be seen from equation (16.19) One of the major disadvantages of PI controllers for liquid level is that they always cause the manipulated flow change temporarily to be greater than the disturbance flow change. It has the feature. cannot guarantee that level will be held within the vessel. we obtain the following: (AQo)m AQi = = = 1. At this point it may be appropriate to note that a viable alternative to PI level control is PL level ~ o n t r o l . For such applications we should choose 6 = 2. It is not a standard commercial item but usually can be assembled with various standard devices.17) (16.38 for 5 1.14a) that: TR = 4 TH 7R/2 (16.048 for5 = = = 0. The major instrument manufacturers can usually furnish modification kits or modified controllers with a larger rR. For example. one cannot readily obtain the necessary rR with a particular commercial controller. by a factor of 2. of not needing antireset windup. useful in some circumstances. we have found two approaches useful: .382 Liquid Level control 3. For Kd < 1. if the system we have been considering is subjected to a step change in inflow.4 1.16) (16. Augmented PI Controllers A plain PI controller. the transfer functions are identical if one reduces the PL level controller gain. even if tuned for 5 = 1.0 2.0 The fact that the outflow swings more than the inflow can create serious problems if the process is running close to capacity. it requires auto overrides just as the PI controller does.

then it is desirable to have a control valve with a linear installed flow characteristic.2 Level Control ofsimple Vmek 383 1.16. This is so because manipulated flow changes more rapidly so that level moves away from the top or bottom of the tank more quickly.0 psig (zero level). This design does not provide flow smoothing quite as good as that of the PI plus override scheme.2 As shown by Figure 16. with an equal-percentage installed flow characteristic. As the level deviates signhcantly from set p i n & Kd increases and TR decreases.0 psig. Effect of Installed Valve F o Characteristic lw If the PI level control system (or proportional-only) is not a cascade levelflow system. = constant. then: a. and b. When the level becomes too h h or too g low.4 whenQ. If the valve has an equal-percentage installed characteristic.0 psig when the input (level transmitter signal) is 9. For this case it is best to find controller settings for 5 = 1 at the minimum expected flow. then from Table 15.or low-selector. Nonlinear PI controllers for electronic^.14).3. = 2Gs Thus we see that.0 psig when its input is 15 psig. = (eFs/2). if its input goes down to 3. our studies show that it rarely permits excessive deviations of level. Auto overrides for pneumatics. . its output is 15 psig. that is. a propomonal-only controller (usually a fixed-gain relay) takes over through a high. is the flow through the valve in its wide-open position.7 whenQ. the relative stability is decreased at low flow and increased at high flow.1. 5 = 1. and then only by a small amount. reference 1: @o = e12 o kEP Since kEp a = 50): = 3. 4 = 0. Then its output is 15. we see that if a controller is set up correctly with 5 = 1 a t e = Gs. 2.9 for a 50:l equal-percentage value (kEp = In a where Referring to equation (16. Although in theory this is not quite as foolproof as auto overrides. the high-level gain 2 auto override is so biased that its output is 3.0 psig.0 psig. t Q / d 8 . its output is also 3. Correspondingly the low-level gain 2 auto override is so biased that when the input is 9. Then control-loop dynamics would be independent of flow rate and where (ep).^ A preferred version has long reset time and a s d ICd in the vicinity of the set point.

3% Liquid Level Control s C 0 8 G g c1 0 8 5 v) I vg S8 W’ s e 5g E: .

= 0. the e n p e e r will be constrained by either [&. If choose [TRIdesign = [TRImax and find . Alternatively. the control-valve installed-flow characteristic should also be linear. [THIOR = Tuning Procedure for PI Controller From the preceding it may be apparent that to get maximum flow smoothing with a gven commercial controller and specified damping ratio. To find out which. Let [&I. This is larger than the minimum available gain of any commercial controller with whch we are familiar. we suggest that it not be less than one minute. and to determine [K&]d=..25 (see reference 10). flow.16. In this regime it is not absolutely necessary that control be stable. For some applications a much larger value will be desirable. The slope of this curve at average rate For a cascade level-flow system. On the basis of experience. with permission from production supervision. of that controller.’ a Auto-Override Time Constant For large swings in the disturbance flow that will drive the level far enough that one of the auto overrides takes effect. The proportional-only auto-override control system ier has the characteristic time constant: A AHT KOR (Qo)rnax To date we have been unable to come up with a truly rational way of specifying the desired numerical value of [T&R.gn or [TR]d=ign. 2.Imin or [TR].stable r e p e . the system temporarily will be under control of the proportional-only auto override. we can experimentally make a plot of valve loading signal versus is aQJaYlC. the auto override will drive the system back into the l n a . and if the hydraulic resistance and static heads of the process equipment are known. we can usually calculate the installed flow characteristic.2 Level Control o Simple Vessels f 385 If pump and valve curves are available. the following procedure is suggested: 1.

The amount of vapor condensed in the column by cold reflux is: where wR = external reflux flow. of process vapor .4 LEVEL CONTROL OF OVERHEAD CONDENSER RECEIVER VIA REFLUX MANIPULATION For this application it is necessary to take reflux subcooling into account. For proportional-only control. one should make TH 1 2 minutes. one should make TQ I2 minutes. "C A = vapor latent heat of condensation. pcu/lbm "C To = vapor condensing temperature.3 LEVEL CONTROL OF OVERHEAD CONDENSER RECEIVER VIA TOP-PRODUCT WITHDRAWAL Ordinarily this is a simple system to design since it almost always fits the preceding analysis. Chapter 3. Section 9.386 Liquid Level Control 3. It is also more convenient to use weight units. 16. If and 16. discusses some of the practical details. for a PI controller. If choose and find 4. Ibm/min cp = reflux specific heat. "C T R = external reflux temperature. pcu/lbm.

22) where wtwc WD = = pL AT vapor t o top tray.4 Level control of Overhead Gmhmer Receiver Pia Repm Manipulatum r 1 387 Then (16. By malung TH 2 5 minutes.21) (16. f (vertical cylindrical design assumed) ? Strictly speaking we should include condenser dynamics but these usually amount to only a half-minute to a minute lag.4 for propomonalonly level control. Ibm/min = external reflux density. Ibm/min. The larger TH or TQ is.23) + l PLAT Let us now define: (16.20) (16.16. By inspection we can see that: fP&) %(S) 1 wt. we can usually ignore condenser lag.25) where . We may now prepare the signal flow diagram of Figure 16. lbm/ft3 = cross-sectional area of tank.1(s) - [ 1 + (To . the less signhcant will be condenser lag. Ibm/min vaporfim top tray to condenser. also rate of condensation = top-product flow.TR) X 1 I 1 (16.24) Then (16.

E 3 5 0 C 0 8 CI m 8 0 p.388 L i p i d Level Control c m s a c 0 c . E L 4 2 *= E B E 28 W = 529 L W sz . U C 2 s 8 P m h C 2 0 C 0 p. S E E 0 3 2 m > E c . cn 8 c .

5 pcu/lb"C. Typical values are in the range of 3-8 seconds.-n +m> (16. from lowest downcomer into column base. Some of the practical details are discussed in Chapter 4.29) where flow.5 COLUMN-BASE LEVEL CONTROL VIA BOITOM-PRODUCT MANIPULATION This is usually a fairly straightforward system to design since it fits the analysis of Section 16. Some idea . T Subcooling therefore can have a signhcant effect on TH. For most applications one should make rH 2 10 minutes. For PI level control we can readily show that: (16. flow due to reflux or feed change wV = boilup. Let us suppose that there are n trays between these two points and that hydraulic lag of each tray is first order with a time constant rm (see Chapter 13). Ibm/min. A = 100 pcu/lb. 16.27) (rm5 + 1)" Such an approximation simplifies considerably either hand or computer calculations.5.2.26) 16. and cp = 0.28) and (16. Then Ksc = 1. then wl(5) = e-nTmswF(5) (16. It has been shown elsewhere (Chapter 12 of reference 1) that a number of equal lags may be approximated by dead time: 1 = . lbm/min wl = . If the feed enters at its boiling point.2 must be extended to take into account the lags between the column feed point and the column base.16.6 Column-Bme Level Control via Feed F h Manipulatirm 389 Note that rH contains Ksc and is inversely proportional to it. Ibm/min w B = bottom product.the magnitude of Ksc may be obtained by assuming as an example To of ' = 100°C.6 COLUMN-BASE LEVEL CONTROL VIA FEED FLOW MANIPULATION For this case the simple analysis of Section 16.

9 would help greatly. The levels in the overhead condensate receiver and in the column . 0. For PI control choose: 1 = 20a T R Z 5 x (16.390 Liquid Level control The equations for the remainder of the system now follow the analysis of Section 16.32) defines a minimum value of rH.25/a) The level and inflow responses to a step change in outflow of an averaging level control system with dead time and a PI controller tuned as recommended above are given in Figure 16.7 COLUMN-BASE LEVEL CONTROL CASCADED TO STEAM FLOW CONTROL Ordinarily it is satisfactory to use liquid level as a measure of inventory. nb J = K L (16.32) and U . this measurement is ambiguous to some extent. may be represented by the signal flow diagram of Figure 16.32a) where u.5. By inspection we can see that: wds) . however.2.33) (0.1 KIK. by means of Bode and Nichols plots. level control cascaded to feed flow control. rad/& Equation (16.31) Without going through the mathematics. = closed-loop resonant frequency. 16. we will simply state that. we found that for well-damped response (Mp = 2 db): a =-H= PLABKWF. Recently we have done some work (unpublished) that suggests that incorporation of a Smith predictors. dimensionless (wR) = closed-loop resonant frequency.[w&) + tPg(31 PL AB5 Let us next define: a = tarm (16. The complete system.for greater stability and filtering one may use larger values of rH.25 =aoR (16.6. In the case of distillation columns.

7 Column-Bme Level G m C m d d to Steam F h Conml m l 391 c 0 c .16. 3 ! 2 s! 0 0 c 8 2 3 z 8 3 9 m c 3 C 2 0 x g L 4 2 E n" F Ljg a= e 53 2s bm . m a c = I E b: g e m L c .

392 Liquid Lmei Gmmi 0 C s 3 0 3 5 m & e t ! i *r 0 +r : ' 2 g z 3 E c 6 e +r 8 8 f - C E i: i 2% -a3 W E rn F cn E S J a o 3 a .

as heat load increases. percent vapor in the tubes increases. = change in tube vapor volume. lbm/min B.(s) = where kd ~ S T ( 4 (16. Put another way.16. A. fi3. level changes can occur that do not result from inventory changes. and the two “swell factors”: . At a given operating point: V. Heat Load For a constant liquid level in the column base.7 Column-Base Level Contvol Cascaded to Steam Flow Control 393 base do not accurately reflect either total column inventory or inventory changes. an increase in base level normally decreases tube vapor volume: vH(s) = ~VZHB(~) (16. due to heat load change change in liquid volume in column base.due to heat load change steam flow. The effect is most marked at low heat loads. but they can be very important when base level is controlled by throttling heating-medium flow. due to base liquid level change = change in liquid volume in column base. VB(s). Column-Base Liquid level For a constant heat load. fi3.34) V. Thermosyphon Reboiler Swell In Chapter 4 it was mentioned that two factors can affect the volumetric percent vapor in the tubes and thereby cause a liquid displacement or “swell” in the column base. = wsT = = change in tube vapor volume in tubes. fi3. the column base is now seen as attributable in to three factors: liquid flows entering and leaving the column base. due to change in tube vapor volume k 2 = * a HB The total volume change. So far we have ignored these.35) where V. fi3.

No change in liquid downflow.37) where K m has the u i s nt Ibm/min x min and T~ is the individual tray hydraulic Ibm/min time constant in minutes. Overall Control System If we are not limited in column-base holdup and can design for reasonably well-damped control.This may be partially reduced to the form of Figure 16. the mathematical relationship was shown to be: (16. This is so because the reflux flow momentarily exceeds the boilup.8we can see that for stability with proportional-only level control we must have: (16.7. A temporary decrease in liquid flow down the column.394 Liquid Level Control where w1 wm wB liquid fi-om last downcomer due to reflux or feed. and that boilup follows steam flow without lag. an i w e a s e in vapor flow up the column may cause the following: 1.38) . Ibm/min liquid downflow due to inverse or direct response. 2. This we call “neutral” response. which we term “direct)’ response. A temporary zwease in liquid flow down the column. From this last illustration we can see some of the loop’s characteristics as they are affected by reboiler swell and inverse response. Reboiler Swell From the denominator of the upper-right-hand term of Figure 16. Ibm/min bottom product flow. Ibm/min = = = Inverse Response As pointed out in Chapter 13. This says that steam flow responds to the flow controller set point immediately. A.8. 3. Note that when K * / T ~is both positive and greater than unity. which is called “inverse” response. Ibm/min wv = vapor boilup. For either inverse or &rect response. We may then prepare the signal flow diagram of Figure 16. then we can treat reboiler dynamics as negligible.Note that Ke = steam flow-meter gain = 12/(wfl)-. there is inverse response for base level as well as for base composition.

The predictor contains a model of the inverse response and cancels its effect from the measured variable./& > p L B. when holdup is adequately large. . But as pointed out in Chapter 4. The time-domain equation for the predictor is: ~ ( t )K~ = p.9. is to make loop calculations without the inverse response term and to design for a closedloop natural period equal to or greater than 1 0 n ~ ~ .39) where It can be shown by equation (16.H TTR) .t 1) (16.e.8 becomes: An examination of the mathematics suggests that for stability we should make A B .9 after the predictor eliminates this. Inverse Response If Km is positive. the control system designer must be carell about the term: At startup some control systems of this type will simply drive to a high base level and wide-open steam valve because of a large k91 or improperly chosen values of the other parameters. If the control loop has to be tightly tuned because holdup is small. kVl is often large at low heat loads. we have inverse response and the effect is unstablizing. A common approach to design. The inverse response predictor must be implemented in a microprocessor controller or control computer that has storage capability for the term 8 (t . as is swell.n T ~ ) . .39) that P ( t ) will have a slight offset from set point.16.7 Column-Base Level Control Cascaded to Steam Fknv Control 395 The term b92 is normally both negative and small. we can design a compensator for inverse response called an "inverse response predictor.(t . the denominator of the upper-right-hand box of Figure 16.KV1> kVl Kmh and 7 R A. If a PI controller is used instead of a proportional-only controller. allowing us to tune the controller as though inverse response were not present."8 It is analogous to the Smith predictor for dead-time compensation. The impulse function shown in Figure 16.(t)l At + ~ (.' The control loop containing the predictor is shown in Figure 16.

E c 3 2 CI c e s G z m E m > x 0 2s 3 c 0 7-zi w% a- ss ern .396 Liquid Level Control E z cn 0 n 2 W s i3 0 3 L 0 a 3 % .

gg Ln. 0 .16.7 Column-Base Level Control Cascaded to Steam Flow Control 397 2 r 2 3 c m IC 0 C as3 m11 -2 e.

8 s 8 e n 8 8 L 3 e E 2 e& P) zs UP) 5 : gg .398 Liquid Level Control E 8 E E e m m > c .

will be assumed to be resistive only.8 Column-Base Level Control via Cmhmate Tbrottltng 399 If K m / ~ m u i y or close to it.40) The system now behaves as though it has dead time.44) where K. wu VB pL = = = = = wl = kVl = ~~1 = steam condensate flow-meter gain 12/(tt7.16. It is assumed that column impedance is fast enough to reduce to a constant. let us look at the control of column-base level by throttling condensate flow fiom the reboiler. 1bf&/fi5. We need in addition the following relationships: (16.8 COLUMN-BASE LEVEL CONTROL VIA CONDENSATE THROTTLING FROM A FLOODED REBOILER (CASCADE LEVEL-FLOW CONTROL) As a final example.)max steam condensate flow. Column acoustic impedance.+ = = wo w.8 is n t becomes simplified: K m (1 T m . We will ignore possible column inverse response but will take swell into account.1 . Ibm/min boilup.42) (16. lbm/fi3 liquid downflow fiom column. lbm/min liquid volume between level taps. . the inverse response term in Figure 16. Ibm/min fi3 liquid swell coefficient. lbm/min boilup column acoustic resistance.13. seen fiom the base. ft3 liquid density.Ibm/min rate of steam condensation. 16. open-loop signal flow diagram for the flooded reboiler is given in Figure 15.41) (16.e-nTrnJ (16. The basic.

45) (16.46) PL5 Next. it is neglected in this analysis.11. Ibm/fi3 For a small number of flooded reboilers examined to date.27a): (16. the kvl term is essentially constant from low to high heat loads. that is. The preliminary signal flow diagram of Figure 16. The compression effect due to process-side liquid-elevation changes has not been checked but is believed to be small.47) From equation (15. Some consolidation of terms of Figure 16.50) (16.52) . let: (16.10 may now be prepared. lbf/fi? process vapor density.51) (16.48) (16.400 PB Liqutd Level Control = pBu = column-base pressure.11 may now be accomplished: kv1 1 --= PLS kvl P L J - 1 (16. It may be reduced readily to the form of Figure 16.49) so (16. the relationship between heat load and tube vapor volume is a straight line.

60) (16.54) or or (16.59) (16.58) For 5 = 1: 4K~72 1 = Now let - 2KL71 + K Z < (16. the damping ratio should be at least one.56) that for stability we must make KL 71 6 1.8 Column-Base Level Control via Cma!ensate T r t l n botig 401 From Figure 16.56) It can be seen fiom equation (16.52) we can see that the characteristic equation for this system is: For proportional-only control this reduces to: (16.61) (2 kR KL = + 1) k 2 d 71 m (16.57) or (16. But this is not enough for a practical design.11 and equations (16.62) . where: (16.48) and (16.16.

al 3 8 m w I 2 E +.402 Liquid Level Control 5 0 n : 0 2 e m 0 E & m 0 e - E 5 c .' ! ! 8 al E 2 5 3 E - 2 0 V L 4 2 E z U E m E m Z-FI e - $l rdi? r m W = 5s && .

-z c 0 z!z m U a B a u-8 5s .8 Column-Base Level Control via Cmhnsate Throttling 403 2 5 rE 5 m b 0 .16.

. Buckley. TX. Houston. Buckley. “Designing Override and Feedfonvard Controls. F.” Cont. 3. L. ISA. we should choose from equation (16. to use auto overrides or a controller with nonlinear gain and reset. Cheung. Cox.” presented at ISM 84 meeting. 4. Eng.. Fehenwi.” DTIECH. J. P. St. 43-48 (Jan. K.. R.. T. “A Further Study of the PL Level Controller. 5. Smith.62) the solution that gives the smallest value of KL. P.” presented at ISA Svmmsium. - (16. 1977). Houston. ISA J. and W. P. “Nonlinear Control of Liquid 9. well-damped response we want = 7JKL to be as large as possible. . 18-21..” presented at ISA Joint 8.... P. Louis.63) With such a controller it will be necessary. P.”ISA Trans. and W. 65-68 (Dec. M. “Some Practical Considerations in the Applications of Overrides. “A Propomonal-Lag Level Controller. Buckley. Apr. Wilmington. Spring Conference. Shunta. Luyben. Del. 18 (2): 7377 (1979). May 4-5.. as indicated earlier. 1984. S. S.” Proceedings. 22. 82-85 7. Level. 2. choose: . “Selection of Optimum Final Element Characteristics. L. 1971). New York. 1983 Joint Symposium. 10. “Recent Advances in Averaging Level Control. 1964. 1971). J. Control. Buckley. “A Compensator for Inverse Response. Wiley. Techniqzles o Process f 6... (Oct. P. Shunta.. 48-51 (Aug. S. Buckley. 1973. Apr.” INTECH. Oct. 6(2) (1959). S. S. For a proportional-reset controller.404 Liquid Level Control Since for slow. REFERENCES 1. Tex.. Luyben. J. 1976). P.. 1964. 0. and W. 1983.

If the vapor flow must be throttled to each load. It may be of interest that the only applications of tight pressure control we have found are in heatrecovery schemes where the vapor from one column serves as the heating medium for the reboiler of another column. l 2. and perhaps hrnishes heat to other loads.5-2 cpm. affected by heat storage. Latent heats at the top and bottom of the column are nearly the same. 2 HEAT-STORAGE EFFECT ON COLUMN PRESSURE The stored heat in the liquid in the column and its base can exercise quite a leveling effect on pressure and differential pressure. the reverse happens-the rate of boiling is decreased and pressure rises at a slower rate. as mentioned earlier.and downstream pressures help good flow control. and for some of them it is undesirable. w. 405 . a rapid change in pressure can cause flashing or cessation of boiling in the column. 1 7 . The former might result in flooding and the latter in dumping. 3. If pressure starts to drop. Column A' is small compared with absolute pressure. If pressure starts to rise. which mostly treat condenser and reboiler dynamics as negligible. usuallv can be of the "averaging" type. constant up.. And the temperature difference across the column is not too large. more liquid flashes. Most columns do not need "tight" pressure control.1 INTRODUCTION olumn Dressure control. For the upper value one should make at least a rough check of condenser and reboiler dynamics. is A simplified analysis can be made if: 1. Most of the tight column pressure controls we have studied have closed-loop resonant fi-equencies in the range of 0. are usually adequate. the mathematical models in this chapter. which tends to reduce the rate at which pressure falls.1 7 C I Pressure and AP Control 17. For tight pressure contol.. For well-damped pressure control. In this section we will see how vaporization rate. we should use these models with caution.

we may make a partial signal flow diagram as shown on Figure 17. Tcp.wB(s)] (17. to lump the column contents together and to assume an average temperature. lbf/fi? Q.(~ p ~ T ~ p W ~(17. at least for annospheric or pressurized columns.1. Reboiler with Direct Throttled Steam For the case where the steam valve is manipulated by some variable other than flow or flow ratio. as shown by Figure 17.8.C p [ ( W c + R ) + sJK011 T&) + c~T'wR(s) cp@RTR(s) . may be reduced to the form of Figure 17. pcu/lb UR = reboiler heat-transfer coefficient. we may combine the above equations as shown in the preliminary signal flow diagram of Figure 17. Pep. = reboiler hot-side acoustic capacitance. We may now write a heat-balance equation similar to equation (15. actual fi3/sec .4 where: C. = heating-medium latent heat. fi5/lbf A R = reboiler heat-transfer area.3.wc(s) . lbm/ft3 A . = heating-medium flow. in turn.and an average pressure. Referring to Figure 15. we may need to account for reboiler dynamics to calculate qT. pa = heating-medium (vapor) density.2.1) + g ~ ) ~l(s) ( A p + C p T p ) %(-r) 1 -[wF(s) + 5 The material-balance equation is: wcol(s) = wR(s) . fi? Pa = reboiler hot-side pressure.30) but with terms for reflux added: CPTFFwF(S) = + CP@FTF(S)+ qdJ). pcu/sec "C fi? .2) Reboiler with Steam Flow or Flow-Ratio Controlled If we treat reboiler dynamics as negligible. Note that the reboiler is lumped with the column base but the condenser and overhead receiver are not lumped with the column top. This.406 Pressure and AP Control These assumptions permit us.~ p T ~ p f ~ . and if we assume that steam will be flow or flow-ratio controlled.

1 Simplified treatment of heat storage effect on column pressure dynamics .2 Heat-Storwe Efect on Column Pressure 407 FIGURE 17.17.

2 Preliminary signal flow diagram for column heat storage dynamics . 17. we often control pressure in a column by a pressure-dividingnetwork such as shown in Figure 3.7.408 Pressure and AP Control This may be reduced to the form of Figure 17./aP.. = 0.5 where: URA.3 and 17.5 may now be combined into the signal flow diagram of Figure 17. (17.aQv/apa For critical flow.3 PRESSURE CONTROL VIA VENT AND INERT GAS VALVES As mentioned in Chapter 3. Figures 17.6. note that aQ.3) P(4 = URARX aTaIaPa 1+LPa c R s . FIGURE 17. The control-valve inputsignal spans are usually the same but one valve opens while the other closes.

17.3 Pressure Control v a Vent and Inert Gas Values i 409 m g U E z - i? e rc B 0 a . s WZ . E m m U E bz m e F m v - E 8 5 e.

3 and 17.6 Combined signal f o diagram for figures 17.5 lw .410 Pressure and AI? Control FIGURE 17.

Figure 17. the valve gains. we may reduce Figure 17.8 and 17. Note the pressure feedback on w.4) Similarly.7 to the form of Figure 17.9. this commonly leads to large controller gains (small proportional bands) and control valve saturation for fairly small disturbances.(s).8. lbf/fi? PR = pressure downstream of vent valve./aP.G.. lbm/sec Pq = column pressure.1 term.10 where: (17. K. Together with the first-order dynamics. If we can assume that PIG and PR are sufficiently constant.9 can be reduced to the form of Figure 17. and the condenser. lbf/@ 6. Note that the valve gains. the equation for the vent valve is: (17. awIG/d6.G.17. vapor line to condenser. and the addition of the pressure measurement./aO.6) we can see that open-loop pressure dynamics are essentially first order.(s).3 Pressure Control via Vent and Inert Gas Valves 411 The equation for the inert gas valve is: (1 7. are both assumed to be positive. .. The term C.and aw. through dT. lbm/sec we = vent flow. For the case where reboiler steam is flow or flow-ratio controlled we can now combine Figures 17. dwIG/aO.3 into the signal flow diagram of Figure 17. and the controller.5) where WIG = inert gas flow.7. K. Ibf/ft2 PIG = inert gas supply pressure. = controller output signal These equations may be combined into the partial signal flow diagram of Figure 17. Since in most cases the inert gas bleed and the vent flow are fairly small.6) From equation (17. reverse action of the inert gas valve is obtained by the . is the acoustic capacitance of the column.and awe/aO.. tend to be small.


Pressure and AI’ Control

Partial signal flow diagram for column pressure control via manipulation of iner gas and vent valves

FIGURE 17.8 Reduction of signal flow diagram of figure 17.7

17.3 Pressure Control via Vent and Inert Garr Valves











m >



c ,

E al


m m m m


5 .C

0 ._

0 C

m -

c ,

a S

E m e


c , C
0 0

g m g

€9 2z sg


E" 0


2 3 a-0" 63,



55 $

2 g;


Pressure and AP Control


= &g z G

2 s u l E + 0



3 s E a


17.4 Premre Control Pia F h h d Conhnser




Here we consider four cases:

1. Flow- or flow-ratio-controlled steam to reboiler, signhcant inerts. 2. Flow- or flow-ratio-controlled steam to reboiler, negligible inerts. 3. Steam to reboiler not flow or flow-ratio controlled, signdicant inerts. 4. Steam to reboiler not flow or flow-ratio controlled, negligible inerts.

Flow- or Flow-Ratio-Controlled Reboiler Steam, Significant lnerts
For this case we combine Figures 15.4 and 17.3, which leads to the signal flow diagram of Figure 17.11. This reduces to the form of Figure 17.12 where:

KFC6 + a)

This will clear to an expression that is first order in the numerator and second order in the denominator.

Flow- or Flow-Ratio-Controlled Steam, Negligible Inerts
This system may also be represented by a signal flow diagram similar to that of Figure 17.12 except that Kp3Gp3(s) replaces KpzGpz(s) has a slightly and different definition [see equation (15.26)]:
G C ( S


+ a) + 1)


cp 1 + Kf-,(s+ a) x -[(we + WB) + s , aT,W,] , S ( 4 J + 1) Ap apLp


Steam to Reboiler N o t Flow or Flow-Ratio Controlled, Significant lnerts
Pressure controls via a flooded condenser with signhcant inerts and where reboiler steam is not flow or flow-ratio controlled may be represented by a signal flow diagram (Figure 17.13) formed by the combination of Figures 17.6 and 15.4. This diagram may be reduced to the form of Figure 17.14 where: KP4GP4(S)




+ 2 (TQS + 1 x - x - [ P ( s ) + cp (Tii, + WB) + sWco,] a, &

+ a)

7$s2 + 2 (7Qs aT* 1


+ a)


(17.9) Note that the numerator comes from equation (15.25) while P(s) is given by equation ( 17.3).


Pressure and AP Control

z m

8 0 3


i 3







E m

s I:




t G 8 c C




e U s 111 5



U 0

G" m >

2 + C
% ;




5E 23

17.4 Pressure Control v a Fhakd Conhmer i






2 1


0 C 0


q '3


Zrn 3-




Pressure and AP Cmml

Q) +r cn

L 0



2s 4s g!E






lg2 0
P C - 0 0 0


t o

m -6:


! 3




17.4 Presrure G m t d viu Fhaka! Condenser















E? o!



a3 2s

r a a


Prerrure and AP Control

Steam to Reboiler Not Flow or Flow-Ratio Controlled. Negligible Inerts

The signal flow diagram for this case is the same as Figure 17.14 except that Kp5Gp5(s) defined differently [see equation (13.26)]: is


+ a\


This method of controlling pressure, although once popular, has fallen into some disfavor in recent years. This is particularly true for once-through coolant. Since its flow rate cannot be allowed to go too low-which would lead to fouling as well as excessive exit coolant temperatures, which, in t r ,contribute un to corrosion-it permits only limited control of pressure. Tempered coolant, which avoids these problems, is a better choice for column pressure control via coolant flow manipulation. The signal flow diagram for pressure control via coolant flow manipulation is given in Figure 17.15. Once-through coolant and steam flow or flow-ratio control are assumed. The lefi-hand side of this diagram comes tiom Figure 17.3. Note that wc is the rate of condensation, Ibm/sec.

The control of column AP by throttling steam to the reboiler was once very popular in the chemical industry, particularly for small columns. The usual practice was to run at a boilup that would give considerably more reflux than called for by design. This would usually provide a product purer than specification. In an era when it was common practice to overdesign columns (lowffactors, bubble-cap trays, and extra trays) and there was little concern about saving energy, this approach to control did have the advantage of usually providing a good-quality product with simple insmentation. For today's tightly designed columns, it is technically less satisfactory, and with the rapidly rising energy costs its wastage of steam is economically unattractive. Nevertheless we still have an interest in this control technique for override purposes; an override controller is now commonly used to keep column AI' from exceeding the maximum value speciiied by the column designer or determined by plant tests. Before looking at the overall control scheme, let us discuss what is meant by the column impedance, Z&).

17.6 Column AP Control Pia Heat
to Reb&


c ,


g p

m C



g5 22
= W

c z

9 %

s o





€ 23 a




a c




then we must treat the impedance as that of an RC chain as shown in Figure 17. Here each RC section represents one tray. Mathematically the entire network may be studied by the methods of transmission-line analysis. If ZT = 0. If the individual RC sections are equivalent.(s). The terminal impedance. Z&). ] (17. or nearly so. then: ( 17.16. respectively. Hamett.(s) tanh nl + ZT(s)tanh nl + Z&) (17.12) ( 17.13) and (17. and Rose2 If. as would be true for an atmospheric column or for a column with tight overhead pressure control. . between the bottom and top trays as a pure resistance approximately equal to 2 (2 z.11) where (17.16) As can be seen. The validity of this assumption has been shown by tests run by Stanton and Bremer' on a 90-tray column and by the computer studies of Williams.14) where R and C are the resistance and capacitance. the impedance becomes IR at low fiequencies. This means that we treat the acoustic impedance Z . we are interested in the high-frequency behavior of the column. and 1 is the total number of trays. .15) -2 R C r2 1 4 2 15 2 -.422 Pressure and AP Conml Relationship Between Boilup and Column Pressure Drop So far we have assumed that there is negligible lag in vapor flow between adjacent trays. the resistance is that of the tray and layer of liquid to vapor flow. is simply P&)/Q. then the impedance looking up from the reboiler is approximately: Z+) Z. however. Two cases are now of primary interest. and the capacitance is the acoustic capacitance of the space between the trays. for each tray and vapor space.) =+_- where the subscripts s and Y refer to the stripping and rectification sections of the column.*.

16 Equivalent network for vapor f o and pressures in column lw .6 Column AI? Control Pia Heat to Reb& 423 FIGURE 17.17.

C ou A e i z P. t-n YZ 3 Ea 52 w e gz .424 Pressure and AI? Control I Z c C . z 3 m 2 2 c . 0 u E B B E 0 c.

2 C 8 A 2 VI 2 E $ a n m r W E 5z sa L O !5E .17.6 Column AP G m t d Pia Heat to Rebe 425 2 8 -5 $! 0 g e z 0 9 w a a 3 w s E 2 m 5 9 c .

17) A more detailed. ISA 1008-1019 (1956). As in the case of level control cascaded to flow control.T. REFERENCES 1.J. the flow loop must have a linear flow meter or an orifice meter followed by a square root extractor. Eqg. New Yo& Rose. L. T. as shown by Figure 17.. Wilmington. R. 3. J. . the secondary or slave loop. I d . 48(6): 1957. Stanton. Day. “Plant and ProcessDynamic 1958. more rigorous treatment of this subject will be found in Day.. Note that P(s) is defined by equation (17. Bremer.8. we need a somewhat different analysis.8 and add the AP controller as shown in Figure 17. columnbase pressure control and AP control are equivalent. Paper W-2-58. D. Hamett.3). B. We may now simpMy and rearrange Figure 15. Af Control Via Direct Steam Valve Manipulation If. Academic Press. Wlim.. B. 2. is tuned to be much faster than the primary loop.426 Prmure and AP control The other case of interest is that of a tall column (many trays). Characteri&qJJ edited by A. Note that for constant top pressure control. the AP controller is connected directly to the steam valve positioner... and A. Then at high fiequencies: (17. ila s Young.18.3 Af Control Cascaded to Steam F o Control lw Let us assume that column AP control is cascaded to steam flow control and that the latter. Del. and A. Chem. as is usually the case.17. This calls for another rearrangement of Figure 15.

In the following discussion.. more rigorous nonlinear equations should be used.. For binary distillations with extremely high purity product and for multicomponent distillations where feed composition varies widely.. a vapor stream leaving.2 BASIC TRAY DYNAMICS Let us visualize a single tray such as that of Figure 18. 427 . we limit our attention to some of the more basic aspects. V.1 8 i Composition DynamicsBinary Distillation 18.1 with an entering vapor stream fiom the tray below. . L. V. and where the column has a number of sidestream drawoffs. There are five equations of primary interest. Since this is an extremely complex subject if discussed M y . where there is decidedly unequal molal overflow. an entering liquid stream from the tray above.. L. moles.. First. We now consider composition dynamics. we use a linearized treatment of the column equations for binary distillation since limited experience to date indicates that linearization does not lead to serious errors for a conventional column whose feed composition does not vary too much. the assumptions are those for an ideal binary distillation. .1 INTRODUCTION n Chapter 13 we looked at the material-balance dynamics of trays. 18. The holdup on the tray is M . equal molal overflow. let us take up the dynamics of a single tray (not the feed tray) in the interior of the column. and a liquid stream leaving.

j’n-1 and the vapor-liquid equilibrium equation: E = j’n * (18.Lnxn d + vn-lj’n-1- vnj’n (18. = 1 + (a .1) Then we have the Murphree tray efficiency: .2) y.428 Composition Dynamkr-Binury DirtiJlation The first is the tray material balance on the more volatile component (perfect mixing assumed): z(Mnxn) = L+1xn+1 .3) The next two equations concern the total tray material balance: dM. axn (18. .L.j’n-1 j’n .5) FIGURE 18. -dt and (18.Ln+l.l)xn + vn-l.1 Flows t o and from basic tray .v.4) aLn L ( t )21 -Mn(t) aMn (18.

. and yn.11). is a h c t i o n of both x. (18. we get: yn = 1 (18.(for any tray efficiency less than unity).9) Partial differentiation of equation (18.1)ZJ2 ( 18.7).11) As discussed in Chapter 17. and (18.2) gives: yn = Ey: + (1 ._ ax.6).10) (18..(S)= V(s) (18. [l and (a .E)yn-l Substituting for yi from equation (18.18.12) We can now combine equations (18.12) to give : .2 Basic Tray L?ynumtks 429 This last equation must be used with caution and only after a test made for inverse response (see Chapter 13) shows that response to a change in boilup is neutral.-. Terms in the preceding equations not already defined have the following meanings: VandL = mols/hr x = mol fraction low boiler (light key) in liquid y = mol fraction low boiler (light key) in vapor Then since the vapor composition y. (18.8) + (a .9) leads to: aY.7) A rearrangement of equation (18. most evidence available today indcates that there is no s i m c a n t lag in vapor flow between adjacent trays.3).l)xn + (1 . (18... we can write: (18. so: V&) = V.E)yn-l Ea - Eax.

1 - 1 (18.14) may be combined (after some reduction) into the signal flow diagram of Figure 18.1 ( ~ ) n For most trays Vne1 = V.17) Typical values of rm calculated by the authors for sieve tray columns are in the range of 2. = and so the last term cancels out.l2 (18. The transmissions V(s)and y.9).5) may be transformed and combined: - + + on-1 (vn-1 .-l(s) (18. Finally equations (18.430 Composition Dynumiu-Binu~ D & M .-%JMn(J) . as mentioned previously.13) v. although the feed tray and two terminal trays have slightly modified diagrams. ft2 change in flow over the outlet weir per change in height over the weir. But.15) where A. which is usually the column cross-sectional area minus the area of two downcomers.E) V(J) . the transmissions L.(s) go to the tray above.14) Harriott’ has shown that: (18.16) we obtain: ( 18. and (18.. Next equations (18.8). .. fi3/sec fi By using the approximate form of the Francis weir formula: Q = kf$.13).W L ( 5 ) + E + 1xn + 1 ( ~ )+ TnEyn . (18. = active area of the tray.(s) and xn(5) go to the tray below. In this way a signal flow diagram for any number of trays can be prepared.5-8 seconds.. (18.4) and (18.Vn)y.. the derivation of rm is valid only if the column does not have sigmficant inverse response.2 for a basic tray.

18.2 Bmu Tray l?ynam# 43 1 + P .0 6 9 3: sg Giii . 1 V 0 0 e E E b P N= . .

the feed flow.27) .r) (18.18) (18. and (2) the thermal condition of the feed may cause the vapor and liquid rates above the feed tray to be sipficantly different from those below the feed tray. this becomes: ZfmdS) + ZfSMf(S> = Z+IXf+l(S) + Zf+Jf+i(S) + F z F ( s ) + Z F F ( S ) - ZfXf(S) - - ZfLf(S) + Tf-IYf-l(S) + Yf-lVf-l(.22) and (18.Lf + Vf-1 dt - v .7): f.18) and rearranging gives us: Vf= F ( l .q) + Vf-1 (18.YfVf(4 We can write an equation for y(F analogous to equation (18.Lfxf + Vf-1Yf-1 . When linearized and Laplace transformed.3 FEED TRAY DYNAMICS The feed tray differs from the basic tray in two ways: (1) it has an additional flow.VflfN .24) where zF = low boiler mole fraction in feed. .25) .21) (18.) (18.20) from which.26) Next the total material balance for the tray is: a34 f = E: + Lf+.19) into (18.23) We can next write a material-balanceequation for the more volatile component (perfect mixing assumed): d -(M-f) dt = Lf+IXf+l + Z F F .18. by partial differentiation. Let us first write an equation for the summation of steady-state flows at the feed tray: ( 18. (18.VfYf (18.19) Substituting equation (18. we obtain: (18.

q is approximately the heat necessary to vaporize 1 pound mole of feed divided by the molar latent heat of vaporization of the feed.3. it is usually possible to control internal reflux rate. so the reflux enthalpy does not change significantly. Lo. minutes e .(a1 + a2 )r -xR(S) (18.29) . in most cases leads to small errors in the calculated values of static gains and top-tray mixing time constant. This is not usually true.28) Note that From equations (18.4 Top-Tray and Overhead System Composition Dynamiw 433 which may be Laplace transformed to: S f S= F(S) M ( ) + Lf+l(J)Lf(J) Vf-l(S) Vf(S) + a L LfN = --fMf(J) af M (18. Further. This we are doing more frequently today.28) we can prepare the signal flow diagram of Figure 18. We can then write the following transfer function relating the vapor composition yr and reflux composition xR: e .2) if we make the simplifying assumption that the reflux. but if the reflux is subcooled only a few degrees. We assume further that the vapor from the top tray is totally condensed.20) through (18. Note that q is a measure of the thermal condition of the feed. the reflux temperature is sometimes controlled. Ignoring subcooling. Then LR = Lo.4). For those cases where it is not practical to control reflux temperature. enters the top tray at its boiling point. the slope of the so-called q line is q / ( q .1). therefore.18. minutes liquid-flow transport delay from condensate receiver back to top tray. subcooling has only a small effect on reflux enthalpy. the intercept on the 45" line is always zF. On a McCabe-Thiele diagram (see Section 2.4 TOP-TRAY AND OVERHEAD SYSTEM COMPOSITION DYNAMICS The top tray can be represented by a signal flow diagram similar to that of the basic tray (see Figure 18.%(S) TDs + 1 where al = vapor-flow transport delay from top tray to condensate receiver. minutes TD = = mixing time constant of well-mixed condensate receiver. 18.

434 Composition Dynamia-Binaty DktUation w L P c L B lg m)'cI E P E 5% w= L W 2G 52 .

This has been shown experimentally by Aikman to be true of a plant column. The mathematical explanation is simple. (18. The term LR($)/6$) relates reflux flow to controller output signal.5. Note that: . Note that y .30) ER + 1) m$ + ( + 7aXT ayT) Rippin and Lamb' point out that if we provide a perfect feedforward control system to adjust LR so that feed changes in flow and composition do not change yT.34) (18. If dead time is negligible. Then: (18.4 Top-Tray and Overhead System Compostth Dynamia 435 The signal flow diagram of Figure 18. By collapsing the overhead composition loop and the two transmissions for Y . then the transmission xR(s)is broken and there is no feedback of xR(s)down the column.~ ( S )we are led to the signal flow diagram of Figure 18. the set point to the secondary flow controller.32) and rT = m (18.l is the mol fiaction of low boiler in the vapor leaving the tray below the top tray. for the moment there are no implications as to whether the manipulation is simply a valve.33) This is the mixing time constant of the top tray. P(s) 1 = zRay' 1 (TD$ .35) .31) where (18.18.4 may now be prepared. then we may write: (18. It is possible to accomplish the same thing by makmg 1/70 sufliciently smaller than the resonant fi-equency of the closed-loot composition-control system.-(Ul+U2)5 ax. or the ratio in a ratiocontrol system. Note that a composition control loop is also shown.

436 Gmpositbn DyMmiEs-Binury Dh-tiflutbm 4 2? E 3 c $ ! z m 0 *r P a 4J 0 e E 2 m m L. s= 2s 2: C 3 o m G 5 .

. E 0 +J a L 4 2 E cn m 5 e m G i3 cn - C I I I sjs W E a h m 3 U sz iip.4 Top-Tray and Overhead System Composition W a m h 437 E E x 9 E 8 ! i U 5 ) I c.18.

39) and the smaller is: (18.K ) is large enough.[ p ( j v ) ]has a gain of 1/( 1 . if the resonant frequency of the closed-loop composition control system is well above l/rD. Actually the control of top-tray vapor composition is not as important as the control of composition of condensate to the next step.35).Otherwise conditional stability might occur if 1/(1 .42) At frequencies below (1 . it should be noted that the first procedure-feedforward control-is more desirable than the use of a large condensate receiver.35) may now be factored into two firstorder lags such that the larger is: (18. the lowfrequency phase-shifi bulge of a proportional-reset controller should be so set that it does not coincide with the phase shift bulge in p(jw).40) Substituting back into equation (18. transmission. .438 ConapwitMn Dynunaia-Binuty DktiUahn where (18. It can be shown that blending is much more effective if done outside the composition control loop.K ) while at frequencies above l / r Dit has a gain of unity. As a practical design consideration. Although we have indicated two methods of breaking the xR(s) thereby simp-g composition control system design and improving the control of?&).37) (18. then there is in effect no transmission of xR. we get: (18. Therefore.38) The denominator of equation (18.36) and (18.K)/rD.

The chief difference is that the reboiler has no entering vapor flow from a lower tray. is that of the reboiler itself plus that of the column base. 18. See. L1 and x1 are the liquid flow and its composition from the lowest tray.1). The holdup. but for the moment there are no implications as to whether the steam flow controller set-point signal comes directly from the composition controller or from a steam-to-feed ratio controller that is reset by the composition controller.5 REBOILER AND COLUMN-BASE COMPOSITION DYNAMICS If the reboiler is either a well-agitated kettle type or a thermosyphon type with a high recirculation rate. This means that transport times in the vapor and reflux lines should be as small as possible. it occurs in some columns when an increase in vapor flow causes a momentaq i w e a s e in liquid flow down the column. The condensate receiver. should have as small a holdup as possible. Usually. This momentary change of concentration in the wrong direction gves rise to the term “inverse response. although in the long run a vapor flow increase will decrease low boiler concentration. It is the subject of a paper by L ~ y b e n .” The signdicance of inverse response in the design of feedforward compensators and feedback controllers for composition is discussed later.4 and 18. We therefore may prepare the signal flow diagram of Figure 18.2-18.30) that it is quite important to keep the dead time in the overhead system to a minimum.6 Inverse Response 439 Finally it should be noted from equation (18.6. This is due to a decrease in foam density on the trays and a consequent momentary overflow into the downcomer. too.18. the reboiler has considerably more holdup than a typical tray. It is assumed that steam flow is controlled by a steam flow controller.(s)is the transfer function relating vapor flow to composition controller output.6. MB. the reboiler/column-base composition dynamics are essentially those of a simplified basic tray (see Figure 18. ~ . however. 18. if unagitated.6 INVERSE RESPONSE The phenomenon of inverse response was apparently first noted by Rijnsdorp? As mentioned in Chapter 13. in the short run it increases it. Therefore. A composition control system is shown in which V(s)/O. the discussion in Chapter 3. an increase of liquid flow onto a tray increases low boiler concentration. As may be seen from Figure 18.

ki - z 8 0 n 2 e E L E 3 u p 2g G $3 a m Giij .440 CmpositMn Dynamia -Binmy DistiUation z E TI m U g C 0 c .

convenient. This was primarily to see how feed disturbances enter the column and how the design and arrangement of auxiliaries can s e c t column dynamics. Binary Distillation-Rippin and Lamb Model For a binary distillation column. and (2) by preparing all of the column equations in matrix form. the "stepping" technique.43) where p is the top pressure in psia and where the individual transfer functions are determined by one of the methods indicated above. or a third method-simulation on an analog or digital computer. and bottom system. zF.7 OVERALL COMPOSITION DYNAMICS In previous sections we examined in some d e t d the composition dynamics of the feed tray. For this reason more recent studies have usually been based on matrix methods. Symbolism such as . Forcing functions in the original Rippin and Lamb equations were limited to LR. one can combine the various equations affecting column composition dynamics and thereby calculate overall column behavior..' A small amount of work has been done with hybrid computers. and F . The signal flow diagram approach to deriving individual transfer functions is exceedingly tedious even for a short column. and even when the shorthand notation of signal flow graphs is used.18. V. From the material previously presented. compact representation of overall column composition dynamics in two equations has been suggested by Rippin and Lamb:2 yds) = pa] z&) OL + zF(s) p$] F(s) OL (18.7 Overall Composition Dynamirs 441 18. Teager6 and Campbell7 suggest that this combining might be done in two ways: (1) by connecting the signal flow diagrams of the individual trays and auxiliaries. the authors have added terms for p and q. overhead system.

Feedback control for top and bottom composition could then be designed with ease. The two models give essentially the same results except that the Waller method leads to higher order-and intuitively more accurate-transfer functions. then the open-loop transfer functions are relatively insensitive to variations in terminal volumes. Rijnsdorp. Applymg asymptote techniques to the Bode plots.sL14 computing these transfer functions in the frequency domain for with a digital computer. 4 ) are valid. however. nonideal separations. Other Models Two other models for binary columns are those proposed by Wahl and Harriott16 and by Waller. predicted behavior checks with that of more rigorous models. This topic definitely needs more study. however.18 Tolliver and Waggoner” have published an exhaustive review of more recent additions to the literature. suggest that if these holdups are no larger than the total holdup on the trays. strictly spealung. Rippin and Lamb developed a stepping procedure. 4 ) may now be represented in signal flow diagram form as shown in Figure 18. For columns with moderate relative volatitities-say 2 to 5-and for terminal purities not greater than 98 to 99 percent.4-42 Cmpacitim DynamiLs-Binuty DirtillutiOn means the open-loop transfer function of yT with respect to 2. of more p o w e m and less expensive digital computers has shifted the emphasis to digital simulation. for only one combination of holdup volumes in the condensate receiver and the column base.43) and ( 1 8 . Low-order transfer functions are calculated &om steady-state data. A comparison of the computational efficiency of s t e p p y and matrix-inversion techniques has been carried out by Shunta and Luyben.” These are similar in that they are both based on circulation rates and liquid holdup in the column. For large columns in particular. later extended by Luyben and others. The advent. Since these transfer functions were mostly of relatively low order-second to fourth order-they could be simulated on an analog computer with much less hardware than would be necessary with more conventional distillation-column simulations. Studies to date. (18. and Maarleveld. 12-14 It should also be noted that the open-loop transfer functions given in equations (18.7. accuracy falls off. . In the meantime we should determine terminal volumes by material-balance and protective-control calculations before calculating column composition dynamics. But for highpurity. A comprehensive review of the literature of distillation dynamics and control through about 1974 is given by Rademaker. Equations . More comor~Iy today the individual tray differential equations are combined and solved in the time domain.43) and ( 1 8 . they determined approximate Laplace transforms. It should be noted that some of the work of modeling column composition dynamics has been concerned with multicomponent rather than binary separations. it is shown that the former is much faster. More work needs to be done to check the range of validity of these models.

7 Overall CmposritiOn Dynumh 443 z m U 3 g 2 I s c z $ c n 8 0 c .0 L E a63 WE: 52 a m e3j .18. I m U e 4 n L E t a a a 5 e e P t-= .

C. and P.. 5(4):386-394 (1966). T. Sc. 32. Elsevier. D. et al. J. ISA Paper 56-28-3. Rijnsdorp. . P. Anderson. Vol. 1( 1):7481 (1967). New York. 15. E. and A. and W.. 7. 18. 12:5-14 (1961). T. 4. McGrawHill. Birmingham Univ... Dynamrczr and Control of Continuow DtitillatMn Columns. 1963. J. Rothfus. Harper.” Institute of Chemical Engineers Symposium Series no. 17. Nos. 1975. Tex.” Bulletin. Thesis. Waller.E. W. and R. Campbell. 6(3):421-431 (1967).. “A Theoretical Study of the Dynamics and Control of Binary Distillation. Rothfus. Cadman. M. 8. Maarleveld. W. Rijnsdorp. University of Delaware. Baber. L. 7.. Ph. Cadman.. Luyben. P r g ... O. Proms Dynamics. Wahl. Tex. Massachusetts Institute of Technology. 3... Tolliver. “Simple Models for Distillation Dynamics. Houston. Series.. 1964. Cadman. 1960.D.9(3):396-407 (1970). “Distillation Column Control. 1955. Kermode. 57. L. 11. T. Waggoner. 36. Chem. London.. “Comparison of Stepping and General Complex Matrix Inversion Techniques in Calculating the Frequency Response of Binary Distillation Columns.. 1980..444 Cmpaitirm Dynamics-Binay Dktillutk REFERENCES 1. E.. Symp. W. Process Cuntrol. Eng. E. E. and R. Apr. En&. Carr. T. 1969. H. Harriott.. E. ISA Trans. Thesis. New York. P.. Kermode. “Dynamics and Control of Continuous Distillation Columns. 12. K. C. L. Dm. 19.” ISA Paper C. L. W. 80-508. R. Luyben. and J. 59-79. Luyben. “Feedback and Feedforward Control of Distillation Columns with Inverse Response. Aikman. D. Shunta. 149 (1961).” presented at American Institute of ChemicalEngineers Meeting. 2. Rippin... R. G:39.C. J. pp.. and R. and N. Fund. Rademaker.” I. Des. 6. A Review and Perspective from the C. D. A. W. I. and D. 1-5. Rijnsdorp. R. 13. New York. L. R. 16.D. 10. R.I. Lamb. University of Delaware. L. L. 1979.. Ph...” Fifth World Petroleum Congress. 9. University of Delaware.. J.. T. Cbem. presented at Houston. Wiley.. Edwards.E. 5.P.. 8:838 (1969). 1959. E. ISA Trans. T. 1964. W. M.E. Thesis. O. P. Harriott. 1958. Fund. I. E. Teager. F. Rademaker. IEC Proc. Oct. I. 14.

while Buckley. Cox. for feedback control purposes we are interested in two composition gains at each end of the column. As a minimum. but bottom composition as well.. As pointed out in a paper by Douglas and Seemann. plot the points of top and bottom purities versus reflux ratio. i.2 and Bauer and OrrY3 who made trial-and-error use of McCabe-Thiele diagramming. Correspondingly. it is only necessary to make a number of top and bottom composition calculations for different reflux ratios. Changes in either of the two top manipulative variables will affect not only top composition.1 9 m Calculation of Steady-State Gains 19. and either boilup or bottom product to control bottom composition. but is implicit in terms of the control variables. changes in either bottom-product flow or boilup will affect both bottom and top compositions.1 INTRODUCTION ost composition control schemes for a binary column involve manipulating either riflux or distillate to control top comp&ition.e. The calculations to predict how a column already designed or built will operate are intrinsically more difKcult than are column design calculations. For example.’ ‘The number of plates required to achieve a given separation at a specified reflux ratio can be calculated directly. Other workers in the field seem primarily interested in such calculations to determine the economic penalty of either overrefluxing (via excess boilup) or of specifymg product purities better than needed. To obtain column gains.” An appreciation of the labor required may be obtained by a perusal of papers such as those of Uitti. More recently Wood4 used a digital computer for trial-and-error calculations via the tray-to-tray method. and draw curves 445 . the reflux rate required to achieve a given separation if the number of plates is specified must be obtained by trial-and-error procedures. therefore. and Luyben5 performed similar calculations on a programmable calculator. we need go only one step further.

As is shown here. 2. The former has the disadvantage of requiring the assumption of constant relative volatility but has the advantage of being much faster when programmable calculators or small computers are used.2 DESIGN PROCEDURE Since we will use the tray-to-tray method. Simple column with a single feed and top. on a binary basis and relative volatility at each tray. and so forth. Binary or pseudobinary separation.446 C&&k o Stedy-State Gains f through the points. it is appropriate to b e p by stating the required assumptions: 1. 2. we may calculate the effect of changing reflux ratio. The basic approach then consists of defining a base case. and reads out any or all tray compositions. 2. Some workers have used the Smoker analytic method rather than the trayto-tray method. calculate mol fi-action low boiler. reflux ratio. 4. Such an approximation greatly reduces computation and time. Sometimes a multicomponent system . The following procedure is suggested: 1. Using printouts of multicomponent design calculations. the second problem must be dealt with first. surprisingly good accuracy can be obtained in many cases by approximating a multicomponent system as a binary system. However. to firm up feed-tray location (number of trays above and below the feed tray). As shown by Strangio and Treybal. We feel that the advantages of the trayto-tray method outweigh the disadvantage of longer run times. 3. Most industrial columns are not binary and do not have constant relative volaulity fi-om top to bottom. the tray-totray method easily handles relative volatility as a function of composition. usually the design case. and boilup ratio. Go through binary design calculations to check the total number of theoretical trays and feed-tray location. Then the slope of either curve is the “gain” of top or bottom composition as a function of reflux ratio.and bottom-product drawoffs. 19. xu. Two fimdamental problems therefore conflont the engineer at the outset: 1. If the match with the multicomponent calculations is not adequate. How to derive a relationship between liquid composition on each tray and relative volatility. How to make a reasonable binary approximation of a multicomponent column. With these in hand. Pick a light key and a heavy key-or a low boiler and a high boiler. go to step 3 and use a higher order function for curve fitting. Use curve-fitting techniques to relate xu and relative volatility at each tray.

l)xn axn (19. If trays are numbered from the top down. 3. and enthalpy or q.and bottom-product compositions and reflux ratio. The tray-to-tray calculations come next. Equimolar overflow.Running a design program is not absolutely necessary for our purposes. three choices were available for relative volatility: (1) constant. but it is one way of initializing subsequent programs and of checkmg assumptions. Partial reboiler and total condenser. it proved inadequate.4) This is the equation defining the vapor-liquid equilibrium relationship. If trays are numbered from the bottom up. LR/D. and (3) a quadratic relationship between relative volatility and liquid composition on a tray. 3. the base is the usual starting point. composition. or may be mixed phase in nature.1 Starting at the column base and working upward. One bottom product that is all liquid.2 Dabn Procedure 447 can be modeled as one binary in the bottom section and as a different binary in the upper section. Feed that may be single-phase vapor or liquid. Find: (19. . 5.1) (19. it is customary to start at the top. The design program (see also Chapter 2) starts with data provided by the column designer or plant. In the first programmable calculator programs we developed. We now have an HP-41C program that fits a quadratic to each of six segments of a column to relate relative volatility and tray number. 3. 1. Find: (19. It also includes top. One top product that is all liquid. find: yn = 1 + (a . 6 . We will follow the second practice. For optimum design it is also the composition of vapor above the feed tray.19. (2) a linear relationship between a and liquid composition x. This includes feed flow rate. But the first time we tried these on a real column for MeOH-H20. The following steps are involved.3) This is they corresponding to the intercept of the operating and q lines. 4. 7.2) 2.

2 Find: (19.2 have been repeated is NR. .5)] is one less. A partial condenser would count as a stage. it produces bottom-product purity slightly better than specification).3 Repeat 3.1 and 4. The first x calculated is that of the first tray above the feed tray. the number of stages there are in the stripping section. liquid composition on the bottom tray. 19. we switch operating lines. the number of stages in the rectification section. The number of times that we perform 3. Starting with the tray above the feed tray. Note that the first x is xB. and therefore represents the number of theoretical trays. and so on until y 2 y j . Calculate P from equation (19.1 Find: (19.3 Repeat 4. Then use yl in 3. What we need to do next is to find a slightly different.1. including feed tray and the reboiler (assuming there is one). Since the condenser is assumed to be a total condenser. 4. “exact” value of R that causes the column to produce top and bottom purities that exactly match specifications. the value o f y used is that of the feed tray. Assume a new value of R slightly smaller than that used for design. it does not count as a stage.1 [equation (19.2 Find: Xn+l = Pyn + XB P + l (19. using x1 to calculate yl. 4.5) This is the material-balance equation and its initial application gives xl.. 4. including the feed tray. The number of times that we perform 3. 3.6) This is the material-balance equation for the rectification section.1 and 4. This may be determined by the following procedure. 2.448 Calcuhbn of Steady-State Gains It is the x versus y curve for a McCabe-Thiele diagram.3 EXACT R PROCEDURE The “design” program usudy leads to a design that produces top-product purity slightly better than specification for an integral number of trays (if we start at the top instead of the bottom.7) 4. 3.2 to calculate x2.2 [equation (19.4)] is N.2 untily 1xD (total condenser assumed). designated “Type A”: 1.2). The number of times 4.

5. used a simple “proportional controller. go back to step 1. Compare last y = yr with specified xD. we multiply the “error” yT 7 xD by a constant with the proper polarity for negative feedback. 1 + (a . 2. feed composition. but for preparing programs.7) utl they have been used Nr ni N s = NR times. assume a new value of R. Column terminal composition sensitivity to changes in feed rate.” that is. and repeat steps 2-6. Here. This reqmres some trial and error to achieve a balance between oscillation and excessively slow response. (a .l)x. If yr does not match within a preselected tolerance. we must keep in mind at least four needs: 1.4 COLUMN OPERATION PROCEDURE . Originally we used 0.5) Repeat u t l the number of times equation (19. we need to know the change in top composition due to a change in distillate or reflux flow with other variables held constant. (19. A number of convergence procedures are available for step 6. for example.7) If the calculated x. Repeat equations (19.19.6) and (19.l)xn (19. and column pressure (usually defined at column f N S 19... This proved to be too large for some purposes so we switched to lo-’.4) (19. 6.. A better method is a modified interval-halving technique that provides moderately rapid convergence without instability. assume a slightly larger value of R and start again at step 1.+ < x. Many different modes of column operation and control can be visualized. Starting at the column base and working upward. Then switch operating lines and find: (19.1. Use they thus calculated for the feed tray. for example.0002 as a tolerance: [3”r .xD].4) has been used equals ni 4.4 Column Operation ProMdure 449 3. we must be carell to provide program logic to prevent “taking off to the left of the equilibrium curve.6) yn = ax. As indicated by Figure 19. feed enthalpy. find: y n = 1+ ax. We have. Column gains for feedback control.

00000 . Guess for R is too small. or B. Guess for x.450 Calcrrlation o Steady-State Gains f FIGURE 19. is too close to 1.1 Effect on calculation of rectifling m i o n when: A.

assume a new value of xD calculate a new x i : = x. The first of these we have labeled ‘Type A” and the second we call ‘Type B. we change a terminal condition such as distillate flow and find the corresponding change in terminal compositions. Ability to estimate the cost penal? of producing excessively pure product by excessive boilup and reflux. Shinskey and Douglas and Seemann’ have been particularly interested in this.19.4 Column O p e r h Pmmdure 451 overhead).The new distillate flow is: D’=D+AD and B’ = = B - AD AD (since F is fxed at 1. we wish eventually to control column top composition by manipulating distillate flow. The starting point is a set of “prep” equations that reflect the desired change in distillate. Let us assume that feed rate. feed composition. the top-product rate. These sensitivities ofien give good clues to the need for feedforward compensation or for terminal composition feedback control. column gains for feedforward compensation. For example. and boilup are fixed and that we wish to find the changes (i.e. for most of these concerns.” Gains for Feedback Control case 1 Consider Figure 19. and . if feed composition changes. different values fiom those of design. and bottom-product flow is determined by column-base level control. we use one of two types of programs: (1) Column terminal conditions are fixed at new. Again. or (2) with fixed reflux or boilup. feed enthalpy.. steady-state accuracy is required. With these equations in hand. reflux flow is set by condensate receiver level control. As it turns out. 4. and reflux is varied by trial and error u t l the materialni balance equation and tray-to-tray calculations converge. what changes in boilup and reflux do we need to hold terminal compositions constant? Steady-state accuracy is required.00) p‘ LA R’ VJB’ = LR = LAID‘ 1. AD. 3.2 where top-product flow is set by flow control. “gains”) of top and bottom compositions in response to a change in D. As shown by the dotted line. boilup is fixed by flow control of steam or other heating medium.

Calculate: yn = 1 + (a .. Calculate: Xn+l = pyn + x.452 Calculatum o Steady-State G f ’ 2.l p’ +1 Continue calculating y and x alternately u t l the number of times y ni been calculated equals N. bottom product via base level control .2 Reflux via reflux drum level control.l ) x n axn starting at the column base and working up. 3. FIGURE 19.

that is.Ns = NR. The “prep” equations for ultimately finding new base composition. Note that we call the preceding Type B. we have also calculated the change in xB in response to a change in D.4 Column Operation Promdure 453 4. Determine when the calculation is complete. we would start by assuming that both feed and top-product rates are fixed. it is necessary.l and repeat steps 1-5. (a . This second gain is an interaction term.l (Figure 19. If bT . = 1+ 4.! = V. assume a new and larger value of xB and resume at step 4. when the Type B program is to be used. case 2 If we desire to control bottom composition by changing boilup as shown on Figure 19. Note that changing B is equivalent for this case to c h a n p g D. when the estimated xb I at the top tray. The m a p t u d e s of the interaction terms give clues to the need for decouplers. is another interaction term. Since the original design calculations are unconverged. and new top composition.1). &. To control base composition via boilup. we need the gain of xB with respect to V.. y choose new x.1 Calculate: xn+l = (R‘+ l)yn R’ ax. are therefore: V. . We will term the procedure for steps 1-5 the Type B procedure in contrast to the Type A procedure. + AVf LA = L R + AVs p’ = V:/B R’ = LAID Then use the Type B routine. 4. 5.19. Continue until N = NT ..2. first to obtain converged values of xB and xD for the specified reflux and boilup.2 If the first calculated x shows that xk < x k . The reader may have observed that although we need only the change in xD for a change in D to design the overhead composition control system.x D ] is greater than the chosen tolerance. Switch to r e d c a t i o n operating line.The second gain. This means that bottom product is also fixed. change in xD with respect to V.l)xn and y. x i .

perhaps eventually cascaded fiom top composition control. while reflux is flow controlled.454 Calnclatron @Steadystate Gains ’ case3 Another commonly encountered control scheme is that of Figure 19.3.x. I f f d and boilup are fixed at design conditions.3 Distillate via reflux dNm level control.2 except that overhead condensate receiver level is controlled by throttling distillate. This is similar to that of Figure 19.l response to reflux changes may be found by starting with the following “prep” equations: L A B’ =L R = = B + ALR + ALR ALR p’ D‘ R’ VJB’ LAID‘ =D = FIGURE 19. top and bottom composition &. bottom product via base level control .

V: V. x. perhaps ultimately cascaded fiom overhead composition control. perhaps ultimately cascaded fiom bottomproduct composition control. is required. the effect of subcooling may be considered as a change in LR: ALR = :(To . is then required. + AVs V:/B’ p’ D’ R’ =D = + AV. we wish to find the responses x.TR) If the base case assumes some subcooling (TR= TR& and later the subcoohg is different (TR = Tm): case4 For the control system of Figure 19. and if it is desired to find the responses xh. while reflux is flow controlled. if feed and reflux are fixed. Note that if the original design calculations were done on the assumption that reflux is not subcooled. If feed and bottom rates are fixed. the following “prep” equations are required: B’ =B = = AV. x. described in the discussion of Case 1. described in discussion of Case 1. Here overhead condensate receiver level is controlled by throttling distillate flow. &/D’ The Type B routine.l to a change in reflux flow. would then be required. described in the discussion of Case 1.4. The following “prep” equations are required: LA = L + ALR R p‘ = V:/B R‘ = LAID The Type B routine. . case 5 A third basic control scheme is that of Figure 19..4 Column Operation Pmxdure 455 The Type B routine. Base level is controlled by adjusting heating-medium flow control.3.19.l to changes in boilup. Bottomproduct flow is on flow controi.

x. we may wish to find the responses A&.456 Calculation ofsteady-State Gainr Case 6 If. FIGURE 19. is required.4.4 Distillate via reflux d u level control: boil up via base level control rm .l to a change in bottom-product flow. described under Case 1. for Figure 19. The following “prep” equations are required: B’ = = B + AB V: V.AB V. ./B’ - p’ = D’ = D AB R’ = LJD’ The Type B routine. feed and reflux are fixed.

(Aq)F R LAID R’ Use the Type B program as described under Case 1.th Promdure 457 Column Terminal Composition Sensitivity to Various Inputs case 7 For the control system of Figure 19. B‘ might better be labeled B”. Since D is fixed: B’ B + AF - But our program requires that all flows be relative to F = 1.3.B’ = R‘ LAID’ = In the first equation above. we may wish to determine overhead and base composition responses to changes in feed rate.19. Therefore let: B‘ = B+AF F+AF B+AF l+AF For simplicity of symbolism. The following “prep” equations are needed: B’ = = = B + AF 1 + AF VJB’ p‘ LA D’ L + (1 . The following “prep” equations are needed: q’ L A = = = q + Aq L . case 9 For the control system of Figure 19. we represent B” by B ’ . we may wish to determine overhead and base composition responses to changes in feed rate. The following “prep” equations are needed: B’ = B + qAF 1 + AF .q)AF R = 1 .4 Column 0pera.2. Case 8 For the control system of Figure 19.2. we may wish to find the responses xb and x i to changes in q.

case 10 For the control system of Figure 19. The following “prep” equations are required: . For a change in F to F’ = F + AF: B’ But: = L R = + q(F + AF) L R - V. we may wish to find the responses of xD and xB to changes in q. we use B’ instead of B . B’ . we may wish to determine overhead and base composition responses to changes in feed rate.458 Calnrlation $Sttady-Statc Gainr In the first equation above. The following “prep” equations are required: !?’=!!+A!? B’ = B + (Aq)F p’ D’ R’ = Vf/B’ =D - (Aq)F = LR/D’ The Type B program (see Case 1) should then be employed.V. B and therefore: + qF = . should be used. case 11 For the control system of Figure 19. for simplicity of symbolism. A Type B program. discussed under Case 1.4. let us define: B” B” = qAF + B F + AF qAF + B 1 + AF = Again. B‘ again might more properly be written B .B or qAF B’ = qAF +B Since OUT programs are based on 1 mole feed per unit time.3.

Column Gains for Feedforward Compensation Here the objective is to find required feedforward compensator gain to hold terminal composition constant as various external factors vary. We wish to find the changes in LR.19. bzF. or 19.3. and C. This dormation could be used to design feedforward compensators to minimize transient changes in terminal compositions.. case 12 For the control system of Figure 19. the effect of a column pressure change fiom P to P’ on xD and xB may be found by recalculating a’s.4 Column Operation Procedure 459 v: = v. we may wish to determine overhead and base composition responses to changes in feed enthalpy factor. and V.4. rerunning the program for the quadratic coefficientsA. The following “prep” equations are needed: F q‘ = q + Aq V. B. 19... + (Aq)F = 1. q.3.4. B.2. B.and entering the new values of A. and C..! = V. ..2. For any of the control schemes.4assume that top composition and bottom composition are held constant. D. Case 14 For any of the control schemes. required to hold compositions constant in the face of a feed composition change.00 p’ = v:/B Case 13 Use the Type B program (see Case 1). or 19.19. the effect of feed composition change fiom zF to zk on xD and xB may be found simply by entering the new zk into data storage for the Type B program (see Case 1). in the Type B program (see Case 1). The variables chosen for feedforward compensation will depend on which feedback control scheme is used-Figure 19.+ qAF p’ D’ = = = V:/B’ 1 - B’ R’ LR/D’ Use the Type B program (see Case 1). case 1s For any of the control systems discussed-Figure 19.

. B. say xD. If the column control scheme fixes either D or B.: . Then if one new terminal composition is chosen. described in Section 19.. case 19 If we have a situation similar to that of Case 18 except that both a new x.. the other is readily calculated: X. is set at a new.460 Calm- OfStedy-State Gaim The equations required to recalculate the other variables are: 2: = ZF + AZF F = 1. + Economic Penalty o Overrefluxing f Case 18 If product purity at one end of the column. Case 17 This is the same as Case 15 except that column top pressure is changed to P = P AP. Case 16 This is the same as Case 15 except that q is changed. to ensure that product purity is always at least as good as the original specification.3. the other will also be fixed. The only equation needed is: q' = q Aq + Use the Type A program (see Section 19. we will get changes in D and B as well as LR and V. Relative volatilities should be recalculated and the quadratic ' coefficients A. and C. reevaluated. constant. Then run the Type A program (see Section 19.00 B=F-D' Use the Type A program. and a new x. are chosen.3). = XD + AxD Use the Type A program (see Section 19. higher value. we may wish to estimate the cost penalty of the increase boilup.3). We will need to know LR and V.3).

the following “prep” equations apply for constant V.: R’=R+AR LA = = R’D’ LR LA or + D’ + D’ +D +D RID’ so: = RD B’ = = 1 . For a column with automatic control of R equations apply for constant R: = case 21 L/D. - + LA = VA = Vi (1 LA = q) RD’ so: D’ or + RD’= v: (1 . Case 20 If a column has automatic control of R = LR/D (it does not matter whether condensate receiver is level controlled via reflux or top product).q ) l+R .D’ p VJB’ Use the Type B procedure (see Case 1). Shinskey’ has pointed out that the differences between the design value of VJF and those required for higher values of xD are measures of the cost of overrefluxing.19.3).q) D’ v: (1 .4 CoLumn Opera&n Procedure 461 Use the Type A program (see Section 19. + AV. the following “prep” Vi D’ But: = V.

which is Case 1.469 0. A = 0. Although we chose small changes. so for the design case.001. D = 0. It is also a function of F.46 0. 19.80 R = 2.1.80 a t x = 0.00. Mer running the design and exact R (Type A) cases.7 = XB = q = a = 2./dD and d%B/dD. and once for D slightly larger than design. D The gain is a very nonlinear function.d+Bx+C The results are tabulated in Table 19.was held constant. we made a data card.20 a = 2.306.53608. decreasing as xD increases and as D decreases. .01 0.98 0. dx. V. We then chose to find the column @IS. once for D slightly smaller than design.462 Calculation @Steady-State Gains Use the Type B procedure (see Case 1). we fitted the two Sets of three points-one set for xD and one set for xB-to a quadratic h c t i o n : y=A. we found large changes in gain with changes in D.46392 and B = 0. With this card.50 N.20 atx = 0. We arbitrarily chose F = 1. One could use this information to design a suitable nonlinear or adaptive control scheme.5 EXAMPLES Example 1:Calculation o Column Gains f Let us use a test case where the following conditions apply: ZF = XD 0. = 7 N T = 16 Our HP-97 program predicts: (NR = 9) p = 2. we ran a Type B program twice.9930 (xg)find = Next we found the exact R to be 2.

98113 0.5 f i a m p h 463 Example 2: Economic Penalty o Overrefluxing f Douglas and Seemann' have investigated the consequences of overrefluxing a deisobutankr. predicts N.332 = 5.780 -0. Again we made a data card at this point to minimize subsequent calculations.9520. = 21 stages and N. D I xo x* 0. and NR = 31. = 26. For the design case. reflux is set by condensate receiver (reflux drum) level control.65 trays. Design conditions were: ZF = 0. that N.98000 -1.314 .575 XD XB = = = 0.009273 dx 0 dD dx.00 q R a = 1.670 0.87 They assumed a control system in which distillate is flow controlled.01000 0. as well as the tray-to-tray method.46292 0.46392 0. dD -0.97875 -1.050 1. The exact R turns o'it to be 5. Douglas and Seemann calculated by their analytlcal method. and bottom-product rate is controlled by column-base level.30 stages.2 shows the results obtained with various values of xD greater than the design value of 0. based on a simplified Smoker equation. The exact Smoker equation.19.950 0.852. = 18 stages. Table 19.891 0. including the reboiler.010835 0. Predicted overhead composition by our HP-97 program is 0. Overhead composition is controlled by boilup.46492 c 0.189 -0.063 -1.950.

997 4.” ISA Paper 49-9-2.33 31.1880 6.774 5. Cox. Des.” presented at American Institute of Chemical Enpeers Meeting. 50(6) (June 1954). E. K. 64(1) (Jan. Uitti. Eng. L. and R. C. Dirtillation Control. “Effect of Control Point Location Upon column Control. and R. Wood. Eng.. P. P.1834 8.965 0. Pmc. Ey. 1968). Luyben.036 0. Cbem.. L. New York.72 11. Strangio. Seemann. 7.970 0. Cbm.050 XI3 0.12 19. 2. 4.SEEMANN ANALYTICAL METHOD 0 1.960 0. 6.442 4.65 REFLUX RATIO R=LR/o DOUGLAS.. 1976.043 5.S EEMANN ANALYTI CAL METHOD % ENERGY INCREASE DOUGLA S.13 6.6145 7. M. McGraw-Hill.. G. 3.93 5. 1977. Cbem..261 0 5. Pny. E. A. Etg..87 4.2 Energy increase due to over-refluxing deisobutanizer TOP PRODUCT COMPOS ITION BOTTOM PRODUCT COMPOSITION REFLUX RATIO R = LR/D T RAY-TO-T RAY VAPOR BO1 LUP-TO-FEED RATIO VS/F TRAY-TO-TRAY % ENERGY INCREASE TRAY -TO-T RAY XD 0. and C. Treybal.022 0. 13( 3 ):279 ( 1974). D o . Shinskey.193 4. S.464 C&ulatMn of Steady-State Gains TABLE 19.Pmg. “Steady-State Control of Distillation Columns. R. Buckley.61 3. Pmg.44 6. and W.950 0. Douglas.62 6. Chicago.0185 3. I d .98 11.8521 6. .47 REFERENCES 1. V. J.. Orr. K..70 6. 5. F.029 0. 49-55 (June 1978).955 0. R. Bauer. Cbem.

1 INTRODUCTION truly d e h t i v e treatment of composition control. 9 465 .20 a Composition Control-Binary Distillation 20. even for simple binary distillation. we strongly recommend a simulation study using a nonlinear model to compare different control schemes. usually involving “modern control theory. however.” These papers are largely aimed at multivariable control. Unfortunately most of the studies are concerned with product impurity levels that are typically in the range of 1-5 percent and the authors assume constant relative volatility and linear column models. Tyreus” discusses multivariable control of an industrial column whose design was performed via a technique called the inverse Nyquist array. For a laboratory column Waller and associates” studied six approaches to dual composition control. Some very interesting papers have been published that demonstrate exotic techniques. a laudable objective. for example) apparently are not badly affficted with interaction. The last consideration results in very nonlinear behavior. Most of those reported in the literature (references 8 and 9. Real-life industrial columns are often highly nonideal (nonlinear) and often must produce products with high purities-0. some columns are so equipped. has not yet been published-and the reader will not find one here. But particularly for composition control at both ends of the column. Although most existing columns do not have composition control at both ends. At least dual composition control was accomplished without decouplers.1 mol percent impurities or less. Some very perceptive insights into the Merences between petroleum refinin distillation and that in the chemical industry are offered by Tolliver and Waggone in an extensive literature review. we will assume linearity in this chapter to give some insight into column-composition control principles. including those of modem control theory.’ For simplicity‘s sake.

FIGURE 20.1 we may prepare the closed-loop signal flow diagram of Figure 20. reflux.466 Composition C ~ t r o l . were represented graphically in Figure 18.1. then for the system of Figure 20. If we assume that boilup. Then to go from the primary controller output to flow we need only multiply by l/K+where K+is the flow-meter static gain (linear flow meters assumed).B i ~ tDiniuation y 20. Base composition control cascaded to steam flow control For all four loops we assume that the secondary flow controls are fast compared with the primary level and composition loops. Condensate receiver level control cascaded to &stillate flow control 2. This permits us to ignore flow control loop dynamics. and top-product flow are the manipulated flows.1 Distillate via reflux drum level control b t o product via base level control otm . bottom-product flow.2 for: 1. Base level control cascaded to bottom-product flow control 4. based on the Rippin and Lamb model. Overhead composition control cascaded to reflux flow control 3.2 FEEDBACK CONTROL OF COMPOSITION In Chapter 18 overall composition dynamics.

. ZP.2 Feetiback Conml ofCompo~tbn 467 Y 0 N E ! E g E P U 2 E L G B CY1 aq !2= s.20.

as shown by the signal flow diagram of Figure 20. The technique here leads to stable. sometimes requires some interaction for optimatity. KcrGcr(s). is shown as a PI controller with external reset feedback. If. with its more sophisticated approaches to multivariable control. It also means that if the operator puts the receiver level control loop on “manual.1) Note that the feedback controller. Since column composition does not change u t l ni reflux flow changes. we use it only occasionally for very high reflux ratio columns. Modern control theory. .468 Cmpmition Control-Binaty Didlutim Figure 20. but not necessarily to optimum control. we cascade (1)top composition control to distillate flow control and (2) condensate receiver level control to reflux flow control (Shskey’s “material-balance control”). the top level control is “nested” in the top composition control loop. noninteracting control. A possible implementation is shown in Figure 20. The decoupler for canceling the effect of bottom composition controller output changes on top composition has the transfer function: (20. however.4 for the system of Figure 20.” there is no composition control. we must use tight level control.3.2) Note that the impulse function time constants are the same as the reset time constants of the loops to which the decouplers are connected. 20.2 shows that the two level control loops are not nested in the composition control loops. The decoupler for canceling the effect of top composition controller output changes on the bottom composition has the transfer function: (20.2. We also suggested that decouplers could be designed to compensate for interactions in the same way that feedforward compensators are designed. then. It also accommodates feedforward compensation and advanced control techniques without interfering with either normal reset or antireset windup. This means that the condensate receiver may not be used for distillate flow smoothing to the next process step. Since none of our studies have ever turned up a case where this scheme offered advantages for composition controls.3 INTERACTION COMPENSATION In Chapter 12 we proposed a particular control-loop structure that incorporates overrides and antireset windup.

20. CL E 3 m E G 3 2 f '5 8? L 8 E 3 E E P a l E 0 NP m 8 mg ZZ s.3 Interadon Compens&km 469 si 8 U E E 3 a 3 2 - I t.. . EP.

470 COmparitiOn C ~ n t d .B i t ~ t Dtitdhtk y E p. 3 8 U 8 E 3 3 2 N 2 iE g IC 0 D L 42 G E e 3 TU E aB W = 52 gg .

xT.20.3) we find that the compensator.5.7) .4 may now be reduced by means of signal flow diagram transformation theorems to the form of Figure 20. has the transfer function: LR/fpR KmGFT(s) = - -[I xB(s) OL Ke LR(s) (20. KFTGFT(s). By substituting equation (20.5) We can also find the other compensator transfer function from equation (20. xB. By inspection we can see that the eftkt of reflux flow changes on bottom composition.6) Since the two loops are now decoupled.3 Intmactiun COmpenratMn 471 The signal flow diagram of Figure 20.4): (20.5) into the large block of the top loop. we get: (20. we may prepare the partial signal flow diagram of Figure 20. may be canceled out if we make: In solving equation (20.6 that shows that the two loops are now independent. may be canceled out if we make: while the effect of vapor flow changes on top composition.

2 cn m NU 2 Til % W E ab =E !2m Kbp. .472 Composition Control-Binary Dirt- N x 5 E cn IC 0 c P B U E G .

3 Inter& Compensatk 473 0 N u! m Q= + 0 s P a g e I e E m C Ul m =8 U gs W E F$ 5 r.k .20.

6s +1 [EloL + 0.0.6s 1 .12) 21.0000036 2 : (20.0. From these we find that the large top block of Figure 20.5s + 1 .7)] : [%IOL [IOL = (?e)( ) 0. the large block in the bottom loop of Figure 20.13) Correspondingly.474 Comparitiun Control-Binay Dictdmkn As an example let us consider a binary distillation column designed to separate water from nitric acid. The following transfer functions were derived via the stepping technique: [%IoL p$IoL 0. Top composition is controlled by manipulating reflux (see Figure 20.6 becomes [equation (20.000215 EloL= X 28s-k 1 The time constants are in minutes.6s 1) + (20.0000097 = 21.000207 31.11) (20.6 [Equation (20.10) (20.8)] becomes: .0000062 21.1) while the base composition is controlled by adjusting steam flow.9) (20.

as shown by Figure 20. Then.4 FEEDFORWARD COMPENSATION Feedforward compensators may be added for almost any disturbance. Individual feedforward compensator functions may be determined in a very simple fashion. as dscussed in Chapter 12.1s + 1 (20. Let us. and KFBGFB(S) 20. we can see by inspection of Figure 20. But we would expect. Later Luyben and Vinante13 tested decoupling experimentally on a semiworks column.000 18 28.13) and (20. in any event. we had no opportunity to test operation with decouplers.4 Feedjmvard G m t p e m h 475 = (e)( ) 0.7 that: . including one he terms "ideal.16) Equations (20. and there will be no contribution fiom the feedback controllers.15) 1. F. for example. Since the column had little interaction. benefits of decoupling were minimal. Luyben12 has discussed two approaches to decoupling.14) indicate simpler process dynamics than probably exists in reality. with an impulse function and a summer connected inside of the composition feedback controller reset circuit. The decouplers for this problem are easily implemented with analog pneumatic or electronic devices. Niederlin~ki'~ Wallerl5 have also studied decoupling.20. If the compensators do a perfect job." but that offers limited benefits and is more complex to implement. then yT(s) = 0 and xB(s) = 0.14) The decouplers themselves become: (20.02 (20. Since the column as built had a composition analyzer at only one end. look at feedforward compensation for feedrate changes.7. that a PI controller with fairly high gain could be used. Physically each may be implemented. to make yT(s) = 0.

476 Cmpmition Control-Binuty Dirtillation I f a 3 U 3 U B C w * 3 0 a 5 E 8 2 e I -z L '5 5 E 8 3 * = w C 0 m I 3 's I 3 E 0 8 k s 8z 3% sg bu .

20.23) (20. we can derive the following feedforward compensator functions: (20.24) .22) By a similar analysis we find that.18) we find that (20.19) Similarly.) (4 LR = 0 (20. we may write: + Kf4Gf4(5) x a ["'I. to make Y ~ ( s ) = 0 and ~ ~ ( 5 ) 0 in = response to disturbances zF(s)and q(s).18) and (20.17) and (20.4 Feea'jcmard Compensdm 477 where (20.21) and (20.. to make xB(s) = 0.20) By simultaneous solution of equations (20.

in some instances. Usually.478 (20. the incremental improvement obtained with the various G(s) terms is comparatively small. It is likely. Shinskey and McAvoy6 have been assiduous in exploring the applications to distillation columns.~ caused by inverse response in the design of feedforward compensators as well as feedback controllers for composition. one finds in practice a static gain term with a first-order lag or simple lead-lag dynamic compensation. One of the most lucid and concise treaunents we have seen is that of Ray. not implicit in the mathematics.7 The implication is that by proper "pairing" of variables one may arrive. some sort of on-line identification procedure is necessary to tune the various feedfonvard parameters adaptively. was oripally suggested by Bristol" as a means of determining the steady-state interaction between process variables. there has been a tendency to concentrate on compensating for them. that more exact dynamic h c t i o n s will be beneficial in at least some cases. that the steady-state gains gve a true indication of interaction.26) In practice. since feed flow rate changes constitute by far the major disturbances to most distillation columns.25) (20. For columns that are fairly nonideal. Some of the practical problems in feeding forward from feed compensation have been discussed by L ~ y b e n Luyben4 has also discussed the problems . since a distillation column is apt to be just one equipment piece in a sequence of process steps. at a control loop structure that promises less interaction than other feasible structures. For real systems dynamics effects may be just the opposite of steady-state effects and may be dominant. Like other "linear" techniques. there are usually fewer choices of control system structure than . therefore. In the published papers and books there is also an assumption. or relative-gain array (RGA). Further. however. the relative-gain array assumes that the principle of superposition holds. Further. it has been found that major improvement results from usmg only static feedforward. 203 RELATIVE-GAIN MATRIX The relative-gain matrix.

27) and (20.29): A According to the theory: = A3 A4 (20.aYT at. Condensate receiver level controlled by distillate. if one is going to use decouplers. Top composition controlled by reflux. as is entirely practical today. AV. constant xs If V. = 0 and from equation (20. (20. is constant. Some work has been done on defining a dynamic RGA. ReMve-Gain M a r k 05 479 implied by the literature. in matrix form: The relative-gain array is related to the first term in brackets on the righthand side of equation (20. 2.27) we get: -. 3. there may be little preference between various control system Structures.28) whence. Finally. 4.30) .29a) constant v. By the principle of superposition we may write two equations: (20. Bottom composition controlled by boilup (usually steam flow). Base level controlled by bottom product withdrawal. See reference 6. Relative-Gain Matrix for Binary Column-I Let us consider our normally preferred scheme: 1.2 .

34a) (20.27). = 0 and from equation (20.28) we get: (20. On substituting this back into equation (20.34b) L av. (20. we obtain: (20. (20. av.32) L o r av. then AX.34) Next: ay.31) av. is constant.33) Hence : (20.If x. J .

. AX --AD+-AV. the theory says that: A2 A3 aL.aD a5 v so: (20. J = = 1- A1 1- A1 A4 = A1 = A3 = A4.35) and ax.36) (20. If AI zz A2 Relative-Gain Matrix for Binary Column-I1 This scheme is the one that Shinskey" terms "material-balance'' control: 1. The two starting equations are: (20. Condensate receiver level controlled by reflux. 3. 4.38) .20. We can see that each column and each row must add up to 1. 2.5 Rehive-Gain M& 481 L Further.37) Here: (20. Base level controlled by bottom product withdrawal. Top composition controlled by distillate. Bottom composition controlled by boilup (usually steam flow). ax. interaction is severe.

2. Top composition controlled by manipulation of LJD. Condensate receiver level controlled by distillate flow.39b) L av. Relative-Gain M t i for Binary Column-Ill arx The last scheme we consider here has the following features: 1.ax. . av. ax.- Next: (20.39a) (20.

we find: n (20.41) (20.45a) A2 = .-AR . avs (20. 4.5 RCld~e-Gain MatrrjC 483 3. once more: A = Here: & B ax.. av. + -AV.43) (20.aR ay.aR whence.45) Next: (20. Column-base level controlled by bottom-product withdrawal.20.40) Then: ?YT AyT . [: :1] (20. First let: (20. Bottom composition controlled by boilup (usually steam flow).AR .42) AX . + -AV.44) By a procedure similar to that used i the two previous sections.

45~) Again: L aR J &mpfe. Oripally lps these slopes were drawn by eye.10). The h l results.47 A3 = -1. _ . The design specification for yT is 0. and V versus xB (Figure 20. An HP-41C program was used to calculate curves of LR/D versus yT (Figure 20.0000422 - mEA m/hr mEA ax.985710 mEA and for x.000483 m/hr From equation (20.47 2. is 0. are as follows: a * avs aV S Hence: = -0.11).-0. As an example let us consider a binary system with low boilerA and high boiler B.8). s o e were drawn on each of these curves to obtain the four gains. determined by digital Merentiation.45) we find: & and A2 = 2.1.9).47 1 . V versus y T (Figure 20.47 = so: . At the specified operating conditions.484 (20. but this led to serious errors in calculating A’s. LJD versus XB (Figure 20.000165 mfA.

5 Rehive-Gain Matrix 485 FIGURE 20.20. R .8 YT VS.

9 X . R .486 C~mPmiriOn Control-Binaty Dktdlath FIGURE 20. vs.

5 Reldve-Gain M & 487 FIGURE 20.10 YT vs-vs .20.

11 .~ x.488 Compmition C h d . =.B i n a t y Dhil&&. v s FIGURE 20.

Z F ) X B (1 . considerable attention has been devoted to selecting measurement locations for composition control. column-located measurements should be as close to the top tray as possible. and (2) it has been shown that the closer the measurement to the point of manipulation. 20. For some columns overhead vapor or condensate composition measurements may be feasible. One might t i k hn offhand that to control terminal composition. but not on the top tray where there is usually poor mixing.X B ) YT(1 . the better is the control. we found that for terminal purities of 95 mol percent or less.33.solution of equations (20.Y T ) where (20. For overhead composition control.XB h1 = (YT . In making comparisons of the two methods with other columns. one should measure compositions at the column top and bottom. In view of the ease of calculating steady-state column gains rigorously with programmable calculators or small computers.” S . the measurement should be located in the liquid line to the reboiler or in the vapor space below the bottom tray.486933 d. .47) so For the illustrative problem.46) and (20. = 0.33 These results are clearly quite different from those obtained by the more rigorous method. This turns out indeed to be the case.Y r ) h S + (ZF . The range of validity of Shinskey’s separation factor unfortunately has never been published.47) gives hl = 6. This may be partly due to the high terminal purities. we recommend the latter approach. For column bottom composition control.6 COMPOSITION MEASUREMENT LOCATION In the distillation literature.6 COmparitiOn Measurement Lo& 489 Shinskef has derived simplified equations for column gains based on what he terms a “separation factor. Then: R A = V 6.20. For this system his equation for X 1 is: 1 1+ Y T . Measurements on any interior tray of the column should be avoided if at all possible for two reasons: (1)the relationship between composition on any interior tray and composition at the top or bottom of the column is ambiguous.33 -5.X B ) Y T (1 . the two methods gave comparable results. z.

.? 122:361 (1983). K. Pm. 1977. Shinskey. This was true even though he provided no feedforward compensation for these changes. L.80-508. Luyben..” Chem.” Proceedings. Tolliver. the concentration of the same component in the overhead decreases. Tex. W... 1972. New York.I. McGraw-Hill. C. 4. Advanad Pmcess Gmtml. Ray. London. Interactiota A d y s k . McAvoy. T. Furthermore. Fuentes. REFERENCES 1. presented at Houston. D s e Dev. C. Research Triangle Park. ISA.. S. More recently Wood17 has shown that with the aid of a digital computer one may quickly calculate for a given control tray the steady-state variations in terminal composition for a specified change in feed composition. 61:74 (1965). G. New York. C. and W. and K. Luyben.490 Gmpitim Conmi-Binav DirtiuatiOn The last statement warrants some emphasis in view of the various schemes found in the literature for selecting a “control” tray. Ellis. 1983. N. he got f&ly good control of overhead composition for feed composition changes.. and this is seldom the case with today’s measurement technology. and D. m i McGraw-Hill. Uim16 has shown that for an ideal binary system.C. “A Comparison of . Ryskamp. L. Pmg. O t c. e. Fagervik.I. an increase in feed concentration of the more volatile component led to an increase in overhead concentration of the same component. mas ell^. W. J. Eng. 5 . For the same system.” Institute of Chemical Engineer Symposium Series No. H.. “Distillation Column Control: A Review and Perspective from the C. New York. and (2) that tray where the change in temperature per change in overhead (or bottoms) composition is a maximum. Since holding interior tray temperatures constant does not hold terminal compositions constant in the face of feed-composition changes.” ISA Paper C.. W. “control of Hgh Purity Disdlation Columns.. L. ISA Conference #27. O t c. “New Strategy Improves Dual Composition Control. the control of composition at a point above the feed tray and below the top tray leads to an odd result: for an increase in the feed concentrationof the more volatile component. Uim showed that by maintaining a constant reflux-to-feed ratio. 10. 8. “Steady-State Gains in Distillation Column Fdorward Control. 2. has swested a feedforward scheme for changing temperature controller set points.d. J. Hammerstrom. 6.” IEC Pmc.” Wyahu. C. 9. 32. it has occurred to some authors that one should vary interior tray temperatures as feed composition changes.. L. 7. Luyben. Waggoner. 1968. T. “Feedback and Feedforward Control of Distillation Columns with Inverse Response. and R. Waller. 1980. “control Both Ends for Profit. G. 1981. Di-tdlaziota G m.P. 3. 51-59 (June 1980).’6 None of these schemes has any validity for feedback control in binary system unless it is physically impossible to make a measurement at the ends of the column. L. Luyben.18 for example. F.. Two common ones are (1) that tray where the temperature change per tray is a maximum.

“Decouphng in Distillation. 1979). Six Control Approaches for Two Product Control of Distillation. Sn’. Luyben.” ISA Paper 49-9-2. Ab0 Akademi. 1971). Eng. AC-l1:133 (1966). A. W.” Kemian TeoUiruw 29 8:499-514 (1972). 17. D. Process Control Laboratory.”LEE T m . K. Auto. 1968). 20(3):592-594 (May 1974).” IEC Im D s ’ . Cone. Niederlinski. 13.” MChE J . ”Expenmental Studies in Distillation Decoupling. D. 19. Tyreus.” Chem. 24:997-1007 (1969). ‘Two Variable Distillation Control: Decouple or Not 15. W. C . e. Wood. Luyben.” AIChE J. B. Vinante. 18. Luyben. 12.. Eng. Uitti. 14. 1980. W. 16. “Multivariable Control System Design for an Industrial Distillation Column. 17(5):12611263 (Sept. L.. E. DO. Chem. L. Bristol. Waller. . E. L. Decouple.” Report 80-2. “Effect of Control Point Location Upon Column Control. Pmg. 16:198 (1970).” MChE J. Department of Chemical Engineering. “Distillation Decoupling.181177-182 (Jan. K. “Distillation Feedforward Control With Intermediate Feedback Control Trays.491 11. “On a New Measure for Multivariable Control..64(1) (Jan. and C.

For large columns time constants in the range of one-half to one hour are not unusual. Special-purpose algorithms can be constructed in the sofnvare to deal with the multivariable. Most sampled-data control systems employ discrete versions of PI and PID control algorithms although computers are certainly not limited to only these types. their sampling time is less than 1 second. Besides being used in gas chromatographic loops. nonlinear nature of a distillation column. Adaptive control. for example. Microprocessor-based control systems are also discrete devices but. The controller output changes each time it receives a measurement. and then holds that value until a new measurement is received.i INTRODUCTION ampled-data control defines control action that is executed at discrete time intekals rather than continuously. Employing continuous control is not necessary because the composition or temperature responds slowly to upsets. Digital computers are ideally suited to this type of control because they operate at discrete intervals. making them almost continuous devices. because they are designed to replace analog equipment.21 S Sampled-Data Control of Distillation Columns 21. sampled-data control is also useful in loops containing sigdicant dead time. and where process gains are small and time constants are very large. AII example is a composition control loop in which a gas chromatograph analyzer provides the composition measurement at 10-minute intervals. updates the parameters in control algorithms and sampling rates to compensate for nonlinearities in the process.' Optimal control is a 493 . Tray temperature control is one example where these guidelines apply.

e. and how their parameters are selected. not only affects the variable it is controlling. The majority of articles written about interaction compensation have been for continuous control (Chapter 20).. while Fitzpatrick” and Shunta” designed sampled-data interaction compensators. Control engineers meanwhile have devised simpler methods to handle nonlinearities and the multivariable nature of a distillation column. selects the parameters to cause the process to line out smoothly in the minimum time following a load disturbance or a change in set point..7-” Much still needs to be done to bring some of the new. but also may afTect controlled variables in other loops. Therefore. A measured load disturbance is used to adjust the manipulative variable before feedback control occurs. modem control theories to the place where they are convenient to implement. called “minimal prototypeyy control. It is worthwhile here to review some of the more common types. More thorough discussions on this can be found in several texts on &@tal ~ 0 n t r 0 1 . ~ ~ 3 ~ ~ Analog Types The most common algorithms in use today are discretized PI and PID algorithms in which the continuous b c t i o n s of integration and differentiation are approximated by numerical methods.1) . Feedforward control is yet another type of multivariable control. interaction compensation attempts to eliminate the effect one manipulative variable has on other controlled variables.494 Sampled-Data CmmlofDiitdh&m Columm technique in which the parameters in the control algorithm are determined so that the process operates to minimize or maximize some index of performance. according to a built-in logic that selects the controller to meet the most urgent need of the process (Chapter 9). A change in one manipulative variable. for example.’3714 chapter discusses the application This of some of these concepts to sampled-data control <. T . one at a time. A PID controller that is approximated by simple rectangular integration is: ek 70 + -(e. Interaction compensation is a case in point. to cancel the effect of the disturbance on the controlled variable (Chapter 20). Override control allows a valve to be manipulated by several different controllers.Another kind of optimal control.-l (21.2 CONTROL ALGORITHMS Control algorithms have been the subject of numerous papers over the years and are still evolving today. 21.f a distillation column. Some simple adaptive control schemes have also been used. their forms. Adaptive sampling rates and gain tuning are typical examples. The types of algorithms people have proposed range from simple proportional and discretized analog types (PI and PID) to highly complex optimal control algorithms.

Others select the parameters to minimize an index of performance such as. in tum. Controller tuning on line determines the dynamics directly by manually changing the set point and observing the process transient response. designing for single-loop control may be inadequate and some sort of interaction compensation will be required. Provision should be made to adapt the algorithm to process nonlinearities.. Some criteria are similar to those for continuous systems. of course. Because these algorithms approximate the continuous ones. An alternate approximation is called the “velocity” form whose output is the change in valve position or set point. Sampled-Data Algorithms An alternative to discrete PI or PID algorithms is one that is determined by sampled-data techniques using the z-transformation.. From the dynamics. that is. The bias term B is to initialize the output when the algorithm is put in service. is the absolute value of the valve signal or the set point in supervisory (cascade) control. if process interactions are sipficant.” The crux in implementing these algorithms is to select the parameters K. closed-loopMp. and T to achieve satisfactory control. usually & 10-15 percent about the initial steady state. changes in process gains and time constants as the operating conditions change. for example. In addition.. but is expressed as a ratio of polynomials in powers of 2 whose coefficients are specified to achieve a certain response. in the fastest time without exhibiting oscillations.TD. Controller tuning off line needs a dynamic model of the process.21 2 Control A&nitbmr 495 This is the “position” form because the output M. Gallierl3discusses a typical method. Dahlin’s method28specifies the response For to a step change in set point to be a first-order lag with dead time. and TD. the controller parameters are selected to meet certain performance criteria. This algorithm does not have parameters K. zero error. This. integral-error-squared. rR. The velocity form does not have to be initialized. requires knowing what the actual process dynamics are. The methods discussed thus far for selecting parameters apply to singleloop control that is limited to a linear region. for example.21-28 example. The “minimal prototype” sampled-data algorithm is a type of optimal control in which the output is specified to reach set point.positions the valve or set point. Some criteria for selecting coefficients are like the methods described in the previous section while others select the coefficients to obtain a specified type of dosed-loop response. their performance deteriorates as the sampling period T increases. . rR. that is. Its output changes the position of a stepping motor or other integrating device that. and so forth. damping ratio.

3 SERVO AND REGULATOR CONTROL Having discussed some general points about sampled-data control techniques and algorithms.1 Sampled-data control .or bottom-product composition-control loop. A typical loop is shown in Figure 21. Let us consider a top. we now look specifically at how sampled-data control is applied to a distillation column. and that the temperature undergoes a tolerable deviation when a load disturbance FIGURE 21. Fundamentally we are interested in achieving good control in the face of set-point changes and load disturbances-traditionally called servo and regulator control.1 where steam flow is manipulated to control the temperature on a stripping tray.fDhiUatMn Columns 21. and its output is the set point to the analog steam-flow controller. A common approach for binary systems is to control the temperature on some tray in the r e q i n g or stripping section in the absence of an on-line product analyzer.496 Sampled-Data Control . The temperature control loop is in the computer. The performance requirements for this loop are that we get a smooth transition from one temperature to another when a set-point change is made.

21.3 Servo and Regrrlator Control


in feed rate or feed composition occurs. A set-point change requires a different response of the controlled variable than a load disturbance requires. Conssquently a controller tuned for good response to a set-point change may be unsatisfactory for load disturbances. When a conventional discrete PI or PID algorithm is used as shown in Figure 21.2, the parameters can be specified for either good servo or regulator control, but not for both. Some have proposed that two separate sets of tuning parameters be stored in the computer. This approach requires additional computer logic and space.

Dual Algorithm
A better approach is to implement a “dual” sampled-data algorithm that is structured to handle both set-point and load disturbances simultaneously, as shown in Figure 21 .3.29The DL(z)part of the algorithm is designed to achieve good regulator control and the D&) part to achieve good servo control. These algorithms are derived in the following manner. The equation for the controlled variable in sampled notation is derived from Figure 21.3. We assume the dynamics of the continuous-flow controller are fast enough to be neglected.

G&I(z)is the z-transform of the product of the process transfer function and zero-order hold. The output of the zero-order hold is the last value of the computer output, which is held constant u t l the next sample time. GLL(z)is ni the z-transform of the product of the load transfer h c t i o n and the load variable. DL(z)is determined from equation (21.3) by setting P ( z ) equal to zero.

GLL(z)and C(z) have to be specified before DL(z)can be calculated. Therefore, some knowledge of the type of load is necessary-whether it is a step or ramp function. The time response of C ( z ) to the load disturbance can be specified to meet any number of criteria as long as DL(z)is physically realizable. C ( z ) is in the form of a series of negative powers of z. The power of z corresponds to the number of sample points following the disturbance. The coefficient of z is the deviation from the initial value of C(z). The set-point compensation part of the algorithm D&) is determined from equation (21.3) by setting G J @ ) equal to zero. This says that the load variable does not change.

The appropriate terms must be substituted into equation (21.5) to calculate D@). DL(z)has been calculated already fiom equation (21.4). C=(z) is the z-


Sampled-Data GmtrolofD)ricillationCdumns


C 0



a qp za
w '

un r




and Regulatm Cmml



3 mg







Sampled-Data CmztrolofDiadutkm Columnr

uansfbrm of the --point change, typically a step h a i o n . C(z)is the specification of the controlled variable response to the set-point change.
Numerical Example Cox and Shunta3' report the transfer functions for a simulated 20-tray

nonlinear binary distillation column in which the second tray temperature is controlled by manipulating steam. Let us determine the dual algorithm for this information (Table 21.1) and compare its performance with that of a discretized PID controller. We use the minimal prototype criterion in selecting the closedloop response. The minimal prototype criterion specifies that C(z)must return to set point in the fastest time possible following a load disturbance without having oscillations. C(z) therefore is specified according to equation (21.6) for a step change in the load variable, feed composition. We express C(z) as the mole fraction of the more volatile component on tray 2 instead of temperature. C(z) = 0.13 Axfz-l (1

+ 0.65 z - ' )


C(z) is the numerator of GLL(z).Axf is the m a p t u d e of the load change. Equation (21.6) says that C ( t )will have a deviation of 0.13 Axf from set point

at the first sampling instant following the feed composition disturbance and a deviation of 0.0845 Axf two sampling instants past the disturbance. C ( t )will be at set point for the consequent sampling points. This assumes that the lsturbance occurs immediately following a sampling instant. Therefore, C ( t ) responds open loop for the sampling period. The appropriate terms from Table 2 1.1 and equation (21.6) can now be substituted into equation (21.4) to solve for DL(z).

-730 (1 - 0.3682-') (21.7) 1 - 2-1 Now let us solve for the servo part of the algorithm D,(z). The minimal prototype criterion says that C(z) must reach the new set point as rapidly as possible without overshooting and having oscillations. C(z)can be specified to reach set point in one sampling period for a first-order process, two sampling periods for a second-order process, and so on. This assumes that the process is linear and that we are not limited in the magnitude of the manipulative variable. G,(s) is almost first order, so spec@ C(z) to reach the set point in one sampling period. DL(4

C(2) =

AC*z-' 1 - 2-I


AC" is the magnitude of the set-point change. Substitute the appropriate terms in Table 21.1, and equations (21.7) and (21.8) into equation (21.5).


21.3 Servo and Rgduttm Control


r c 7-z


2 0

WC $2



Sampled-Data control ofDidl&km Columns

The output of D&) in the time domain is a function of the present value of set point C-(t) and the last value of Ds(t).

Ds(t)= 0.5187 C-(t)

+ 0.48 Dl(t - 7')


The output of DL(z) in the time domain is the required change in the manipulative variable.
DL(t) = -730 [Ds(t)- C ( t ) ]

+ 268 [Ds(t- 7')
- C(t - T ) ] + DL(t - T)


Comparison of Dual and Discrete PID Algorithms
A nonlinear computer simulation30was used to compare the dual algorithm with a discretized PID algorithm for a set-point change and load disturbances in feed composition and feed rate. The PID settings were selected to meet a minimum integral of absolute error criterion for a set-point change as per Smith" (page 176).The sampling period is 1minute, approximately 10 percent of the process time constant. Figure 21.4a shows the response with the two controllers to a step change in set point fiom 0.057 to 0.01 mole fiaction. Performance of the dual algorithm is slightly better than the PID algorithm. Figure 21.4b shows the response to a step change in feed composition fiom 50 to 60 percent. The disturbance occurred immediately after sampling at t = 0. The response with the dual algorithm meets the specified response in equation (21.6) very well and is markedly better than with the PID controller. A disadvantage of the minimal prototype criterion is that the algorithm is designed for one type of disturbance and may not be as good for others. This is illustrated for a disturbance in feed rate in Figure 21.4~. The feed rate is increased stepwise 30 percent and the algorithm derived for a feed-composition load is employed. The response with the dual algorithm exhibits some overshoot and oscillations, but is markedly better than the PID algorithm. In practice the load algorithm should be designed for the predominant disturbance. It has thus been shown that the dual algorithm can handle both set-point and load disturbances satisfactorily without having to adjust the parameters. It is markedly better than a discrete PID algorithm tuned for set-point response.

Feedforward control adjusts the manipulative variable as soon as a load disturbance is detected in an attempt to cancel its effect on the controlled variable. Sampled-data feedforward al orithms are designed readily fiom knowledge of the dynamics of the process!' Feedforward control is best employed in conjunction with feedback control to correct for any offset due to inaccuracies in the process model. A typical feedforward/feedback sampled-data loop is

21.4 Feea@ward G m m i



E 8
w C

Ns W
(N0113Vtld 3 l O W ) I X





aL e



Sampled-Data Conml OfDtidldim Columns


3 t;












2 2




n +E



W a&


21.4 Fcc-ard



3 %


5 z


e e


c, C

r u





aw: ss ZP


SampU-Data Cmtd OfDirtiUation Columnr

shown in Figure 21.5. Theoretically any number of feedforward algorithms DF(z) be added. D&) is derived in the following manner. can A control-loop equation for C(z) is written from Figure 21.5.

C(z) is set equal to zero, which says that there will be zero error at each sampling point. DF(z) solved fiom equation (21.12). is


DF(z) be determined from equation (21.13) once the nature of the load can and the process model are known. Let us calculate the feedforward algorithm for a feed-composition disturbance from the information in Table 21.1 for a step-load change. DF(4

49.3 (1

+ 0.65 ~ - ‘ ) ( 1- O.368~-’)
(1 - 0.48 Z-’)


The time output DF(t)from equation (21.14) is the required feedforward change in the manipulative variable after a change in the feed composition is detected.
DF(t) = 49.3 [ A . ~ f ( t ) 0.28 A ~ f (- T ) + t

0.24 A ~ f ( - 2T)] t (21.15)

+ 0.48DF(t - T )

If the load form is not known exactly, L(s) may be approximated by a
staircase function L*(s)N(~).~’ asterisk denotes that L(s) is a sampled variable. The

This simply says that the load is fictitiously sampled and the value is held constant for the sampling period. The numerator of equation (21.13)becomes:

(21.16) Substitution of equation (21.16) into equation (21.13) gves a different form for DF(z).
= L(z)GLH(z)

Equation (21.17) can be solved for the feedforward algorithm without knowing the actual form of the load. I f L ( s ) is a step function, DF(z), from equation (21.17), will be equal to equation (21.13). The nonlinear simulation was used to illustrate the closed-loop response of the controlled variable x2 following a 30 percent increase in feed composition. The results are shown in Figure 21.4b with the feedback-only dual and PID algorithms. control is immensely improved with the feedforward action. The slight deviation in x2 with feedforward control is due to inaccuracies in the linear model and the long sampling time relative to the process dead time. The

4 Feea’jhward Control 507 E g 0 8 5 s % ! e 2 e I qg e 3 K q !zE a Lrn .21.

19) defines DIS(z): (21. and it causes a further deviation in the rectifying temperature. and stripping tray temperature by manipulating steam. 21. Consider first the stripping section control loop. Basically the interaction compensators will function like feedforward algorithms with reflux and vapor boilup treated as load disturbances on the opposite controlled variables.19) Equation (21. DLS. If the loops are tightly tuned. The load and set-point algorithms DSS.21) .DLRare designed according to Section 21. The controlled variable X . the necessary change to make in V ( z )is: (21.6.508 Samphd-Data Cmml OfDirMatim Columns feedforward action initially disturbs the controlled variable in the direction opposite fiom the direction of the load disturbance so that the combined effects will cancel each other and x2 will be at its set point at the following sampling points. The interaction algorithms DIS and Dm are designed as follows. The steam flow is adjusted to compensate. A change in reflux causes a deviation in the stripping tray temperature. The control system is illustrated in Figure 21.18): (21. DsR.(z) when a change is made in LR(z). The sampling time for this example should be equal to or less than the dead time in GL(s)so that the manipulation will occur when the effect of the load occurs. is given by equation (21.20) A similar calculation in the re+g loop leads to Dm(z): (21. Therefore. The purpose of this section is to illustrate how interaction compensators can be achieved by sampled-data algorithms for the control system discussed above. A classic example is a distillation column in which a rectlfjrlngtray temperature is controlled by manipulating reflux.18) We desire zero change in X.5 INTERACTION COMPENSATION Interaction occurs in multivariable systems when a change in a manipulative variable causes a deviation in more than one controlled variable. Chapter 20 discusses methods to eliminate interaction for analog controls.3. this interaction can result in unstable control.'2 The compensators are used in conjunction with the dual feedback algorithms.

5 Inter& Gmm ? p & 509 FIGURE 21.6 Interaction compensation .21.

A tracking mechanism to prevent controller saturation or "windup" is needed for sampled-data systems as well as analog. This section. therefore. can be neglected for this analysis.8 illustrates a conventional loop. when a change in reflux or vapor boilup occurs. hard-wired circuits are still. for safety considerations. The advantage of computer implementation is the ease with which control schemes can be modified by simply reprogramming. Analog systems require changing hardware and rewiring. The secondary variable measurement is the proper feedback for a primary controller in a cascade loop. therefore. Figure 21. The improvement in control with interaction compensators is sipficant. The flowcontrol dynamics are very fast compared with the composition dynamics and.9.7d for a 10 percent change in feed composition. however. With no means of tracking the output of the secondary variable (assumed equal to the output of the signal selector).taControl ofDimU&im Columns Thus. but requires a different approach.preferred for critical overrides. is devoted to the design of sampled-data algorithms that prevent windup. The principles of override control discussed in Chapter 9 suffice for sampleddata systems and will not be restated here. Therefore. for this discussion.6 SAMPLED-DATA CONTROL FOR LOOPS WITH OVERRIDES The use of overrides or protective controls as a way to deal with the multivariable nature of dstillation-column control is discussed in Chapter 9 for analog controllers.7b show the closed-loop responses of composition on trays 2 and 18 for set-point changes and Figures 2 1 . The secondary . applies to a DDC loop as well. the dual algorithm in Figure 21.30 Tracking Dual Algorithm We assume. the algorithm is structured to track an external feedback signal. This discussion. Shunta12 compares the results with and without sampled-data interaction compensators for c o n t r o h g both ends of a 20-tray distillation column. The override control concept can be implemented readily in the computer as well. However. that the overrides are implemented with analog hardware.510 Sa?nphd-Da. for sampled-data algorithms. D&) and Dm(z)make the appropriate change in the manipulative variable in the opposite loop to compensate for it. Overrides can enter the loop in the set-point path of the flow controller through the signal selector. Similarly.7a and 21. 7 ~ and 21. Figures 21. 21. We are again controlling the temperature in the stripping section of the distillation column by manipulating steam flow and using the dual algorithm. the algorithm must be restructured as in Figure 21.8 saturates when an override occurs. Chapter 9 points out that an analog PI or PID controller saturates when an override takes over control of the valve unless the controller reset is made to track the valve signal in simple loops.

7~ FIGURE 21.6 Sqled-Data Controlfm Loops with Openides 511 FIGURE 21.7d .21.

8 Conventional “dual” control in loop with overrides .5 12 Sampled-Data Conml @Dinillation Columnr FIGURE 21.

6 Samphd-Data G m fb.9 Tracking “dual” control in loop with overrides .21.Loops with Ovewides m l 513 FIGURE 21.

23) . C(s) = [ C " * ( ~ ) D W .DL(z)) = (21.5 14 Santphd-Da$a Control ofDktdL&m Columns variable is fed back to the computer at each sampling instant and is the input to the D part of the algorithm. M*(s) = M*(s)D.C*(s)l M(s)Gp(s) +M*(s)Dt*(s)H(s)Gp(s) L(s)G.27).C*(s)]K 1 .26) K is specified to make DL(z) physically realizable.(s) + (21. Dt(z)is related toDL(z)by equation (21. so that zo terms in the numerator ofDL(z) are zero. and the controlled variable.C*(S)] KH(s)GP(s) Equation (21.22) The asterisk (*) denotes a variable that has been sampled.H(z)KC"(z) (21.22). The algorithm D is solved by first writing the loop equation. (21. The manipulative variable M*(s) is solved and substituted into equation (21. K is a static gain and its input is the difference L between the set-point algorithm 0.24) is z-transformed and rearranged to solve for the controlled variable C(z).[C"'*(s)D?*(s) . C ~ ( Z ) D S ( Z ) G P W ) (K 1 . . Ds(z) is solved from equation (21.28) Ds(z)is the same for both loops.22) by setting L(s) = 0 Ds(z) = c"(z) C(Z> + C(Z)[l .27) This relationship holds true for any form of DL(z).DL*(s) Therefore: C(S)= [C"*(s)Df(s) . The overrides L are assumed to be zero.25) Dt(z) is solved by setting C"(z) 0.*(s) + [C"*(S)D~(S) C*(s)]K (21. (21.DL(z)) + G J ( 4 C(z) = + K GPH(z)(l . .DL(41 G. that is.

T~ is the reset time.13 show the closed-loop responses to a 20 percent step increase in feed composition.Dt(z) is solved by substituting the appropriate terms in Table 21.11 show the closed-loop responses to a set-point change in x2 fi-om 0.21. The sampling period is 1 minute. Figure 2 1.6) for a feed-composition disturbance.10).10 and 21.30) K. A discrete version of a PI controller is found by approximating the continuous controller by rectangular i n t e g r a t i ~ n ~ ~ known as “implicit (also Euler integration”).27). Figures 21.8 and 21. Dt(z) alternatively could have been determined by equation (21.29) Figures 21. C ( z ) was specified as in equation (21.1 into equation (21. is the controller gain and D ( z ) and equation (21.730 to make DL (2) physically realizable for a step-load disturbance. D‘(z) is calculated from . The conventional algorithm saturates and takes almost an hour to unwind (Figure 21.14 compares the conventional loop and the restructured tracking form.12 and 21.01 for the conventional and tracking dual algorithms. (21.6 Sampled-Daa Control j&rLoops with Om-rhks 515 Cox and Shunta3’ compare the tracking dual algorithm Dt(z) and Ds(z) with the conventional algorithms DL(z) and Ds(z) in Figures 21.7) and (21.057 to 0. The manipulative variable M (vapor boilup) is limited by an override to a maximum change of 11 moles/min while the controller output CO (expressed as moles/min vapor boilup) initially calls for a greater value.27). Again control is improved with the tracking design. This same technique can be applied to restructure a discrete PI algorithm so that it does not saturate. DL(z)andDs(z) are p e n by equations (21. The tracking algorithm does not saturate and resumes control quickly. (21.1. Tracking PI Algorithm The tracking structure is not unique to the dual algorithms.9 for the 20-tray distillation column whose transfer functions are given in Table 21.K has a value of .26).9).

516 Sampled-Data Control of Didl&timr Columns FIGURE 21.10 Conventional control o X. with setpoint disturbance f .

with setpoint disturbance .1 1 Tracking sampled-data control of X.6 Smnpled-Data G m fiLoops with Ovemdes m l 517 FIGURE 21.21.

518 Sampled-Data Control o Didhtiim Columns f FIGURE 21.12 Conventional control o X. with feed composition disturbance f .

21. with feed composition disturbance .13 Tracking sampled-data control of X.6 Sampkd-Data Control fm Loops with O v e v d e s 5 19 FIGURE 21.

520 Sampled-Data Control of Dtjtllation Columns FIGURE 21.14 Comparison o conventional and tracking PI control f .

and W. et al.Refireraas 521 K is specified to make D’(z) physically realizable by setting the left side of the numerator equal to zero. C.J. 1972). Mori.. 18(3) (May 1972). “An Experimental Evaluation of Kalman Filtering... K = Kc (TR + TR r> (21. Wright. “Advanced Computer Control Improves Process Performance. H.. Stevens. “Sampled-Data Processing Techniques for Feedback Control Systems. 1954).. 73 (Nov. 13(4) (July 1967). Chem. The antisaturation capability is designed into the algorithm and requires n o additional logic that takes up space and time in the computer. T. B..32) The h a l form of D’(z) is found by substituting equation (21. G... for that matter.” presented at 74th National American Institute of Chemical Engineers Meeting. 6. 9.. 1962). Bergen. Y. 4. and J. “Process Control by Digital Compensation.Erg. “Optimal State Changeover Control of a Multivariable System Using a Minicomputer.” Iw. 2.. “Application of Modem Control Theory to Distillation Columns. 14(3) (May 1968). 50 (Aug. Mosler... 17(2) (Mar. andD. 1973). I U ~ a n d h e r i a ~ ~ that the feedback found signal needs to be free of noise. C. REFERENCES 1. and the A/D and D/A signal converters must be zeroed properly for best performance. Jarvis. et al. (Sept. or almost any other break in the control loop.” AIChE J. D.” AIChE J.. Ragazzini. Mar. Brosilow. and W.. R. (Jan.C ( t ) ]+ D’(t) (21. A. A. “Adaptive Control of a Chemical Process System. 1973. Ahlgren.” AIEE Trans. and K.”ATChEJ. R. R. La.32) into (21. F. Seborg. Hu. R. Tech. the tracking algorithm prevents saturation when an override occurs. D.31). “Optimal Control of a Distillation Column. F. 5 .. 1971).33) The time-domain output of D’(z) is a function of the past values of D’(t) and the manipulative variable. Hamilton. . The complete controller output is: + CO(t>= (T TR) [ c y ). 8. J. Handky.” m.35) TR Therefore. C. “Discrete Compensator Controls Dead Time Process.” AIChE J. Ramirez. E. Ew. Fisher. 7. D’(z) = Tz-’ T + TR - TRZ-~ (21. 3. C. F..” Can. M. D.. and J. New Orleans.

” Inst. 31.. Kuo-Cheng Chin.. J. A. J. B... H. W. 1 1963). and W. . J. Sutherland. E y . 18(5)(Sept. (Dec.. Part 1 . (Jan. Law. Tou. Chem. (May 1969). 1959. E y .. Cundall. 1972). E y . L.. 27.” Can. W. Chao. Shunta. 1 (1970). Sci. et al. C . W. Cont. et al. “Compensating for Dynam~cs Digital Control. C . G.”AIChEJ. K. and V. and W. . 1972.. Sa. “variable Sampling Frequency-A New Adaptive Technique. and J. Marroquin. Khandheria. 1968). A. “Digital Control Algorithms-Part 11 1. P. Traclung Action Improves Continuous Control. et al.” Inst. 28.” EC%.... L. Luyben. J. (June 1968). Syst. 30..” Inst. and R. Dev.. “Application of Conventional Loop Tuning to Sampled-Data Systems. Contr. 32.” Gmt. Cont. Latham.. “Designing and Tuning Digital Controllers. “Evaluating Digital PI and PID Controller Performance.. Neumann. “Sampled-Data Noninteracting Control for Distillation Columns.E y . Mosler.. 1973). Eng.W. 15 (Apr.. ‘Tuning PI and PID Digital controllers. (Apr. A.. M. W. “Dgital and Sampled-Data cmrtrd System’’ McG~~w-HLU. Shunta.” Inst. 20. Slaughter. 16. A. Tech. et al. Fitzpamdc. 11. 27(6) (June 1972). J. 1973). 13... E. Time Domain Specifications of Digital Controllers. 17. 22. Luyben. and Huang. “Noninteracting Control for Multivariable Sampled-Data Systems: Transform Method Design of Decouphng Controllers. (Sept. Da.” Chem. et al. 24. L. Cont.” presented at 4th Annual Conference on the Use of Digital Computers in Process Control..” W. 14. 19. Luyben. (May 1964). 69(9) (Sept.” Chem.. 1962). Dahlin. L. 1969). E y .. A. E. “Sampled-Data Proportional Control of a Class of Stable Processes.522 Sampled-Data Control o f D i d l & n a Col~mns 10. Dgital Computer Process control. C.” Chem.” Inst. Tech. 29. Do.. Luyben.. L.”Inst. R. (Feb. 1973). Cont.. Tech.. J. A. 1971). “Feedfonvard Control by Digital Compensation. Corripio. T. et al. P. Mosler. P. Baton Rouge. Gallier. P. H. (Oct. J. (Feb. “Improving the Performance of Digital Control LOOps. International Textbook Co... 1971). “Sampled-Data Feedback Control of a Binary Distillation Column.. Lopez. Sp. (July 1973). L.. New York.’yIEC Des. 1973.. 25. Luyben. Yung-Cheng. 1968). Prog. P. In&. “Frequency Domain Synthesis of Sampled Data Controllers.” Part I. 21. La. Hsiao-Ping. L. Chem. in E y .. “Designing Digital Computer Control Systems. (Oct. 5(3) (July 1966). et al. Otto. E.” AIChE J. Moore. L. T. P r o m Modeling. and V. 25(5) (May 1970). 26. Simu l h m and Controlfm Chemical En@ern. Tuning PI and PID Controllers. and W.. and W.” Chime J. 6(1) (Jan. B. NW York. Shunta. 1967). Luyben. Smith. “Root Locus Plots for Sampled Data Systems in the Lnz-Plane. 18. McGraw-Hill.”Inst. F. Cox. Syst. 50 (June 1972). “Damping Coefficient Design Charts for Sampled-Data Control of Processes with Deadtime. W.. F. 1976).. Luyben. 15. SF.. J. 12. “SelfTuning Computer Adapts DDC @rithms. Syst. 33. (Apr.. B.” Inst. M. L. ”Experimental Evaluation of Digital Algorithms for Antireset Windup. 23.

522 Cutler. 135 Fuentes.374 Arnold. 192 Chin. 305. C. R.. 242 Boyd.. 23. C. R. F. 292 Bolles.. 241. 297. O... W. J. L.. 292 Billet. H.. 441. A. E. K. A. B.. L. R. T. 292. 292 Holland.. 522 6.. J. E. 2 3 Harnett. K.. 192 Fehervari. R. 444 Carr. 107 Bonilla. C. E. S. 495. 445... F.. R. D. 135 Griffin.. 135 DOSS.. M. R. S. 373 Grabbe. 347.. R.. J. L. T. W. M... 227 Gallier. 522 Cheung. 311. T. 404 Chiang. W. D. 242 Hamilton. 311 Baber. 444 Bauer.. 373 Hengstebeck. 521 Harbert. 522 Day.. 306.. 242 Grote.. R. D. P. P. 521 Buckley. 340. 135 Compio.... 311 Dobratz.. 292 Arant. 135. 242. T. 66 Hollander.. 107 Church. 490 Gaines. C. R. C. Cundall. 292 Archamboult. 79. 310. L.. 426 Bristol.J. H. 527 Garcia. J. C. 1 .. 464 Beck. 310. 426 530 . W. A. J.. 240. M.. 522 Cox. A. 347. 313. D. 521 Fitzpamck.404. 442. 180. 444 Harriott. G. 310 htrom. G. 297... M. E. T. D.. A. C. M. G. 430.. D.... L. W. B. J. B. 23. A. H. 490 Fahmi. 404 Fisher. L. D. 490 Handey. 444 Campbell. T. 444 Ellis.404. 192 Bergen. 227 Gilliland. 258. R.... 180 Driskell. 23. B. 167... 310... F. R. F. K. K. 230. E.. C. A. E. P. D. C. 192 Giles. D.. 435. A. 242 Bremer.. E. 464 Cadman.. 4 4 500. J. 23.. 445. L.. 310 Edwards. W. 107. 490 Fagervik. 464 D o h ... J. 310.. T. T. L. T. E. M. C. P. 227.. G.. 521 Berger.. 107 Dahlin. 444 Chao Yung-Cheug. I?. R. L. 444 ANSUISA. N. 23. J.Author Index Ahlgren. 66 Binder. D. D. 522 Frank.. V. 444 Hempel. L. 23 Hepp. R. 108. 66 Gould. 310 Beesley.. 311 Douglas. J. E. N. 445.... 310. C. 292 Edgar. R... B. L... 108. K. 135. 490 Brosilow.. 444 Anderson. 231. 426 Harper. J.. 521 Aikman. J. M. F.. 311 Geyer.. 373.445. 311. 521 Hammerstrom.

C. L. 442. L. 135 Schmoyer. L. J.. J. 464 Shah. T. C. C. S.. 292 Schnelle. 192. 180.. 23.. 292 Powers. 441. 481. W. F. 151 Oldershaw. 444 Ragazzini. 4 1 475.. Y.. 478. J. 242 . R. 464. 444 McAvoy. 521. 307.. H. 6.. 3. F.445. 241. 296.. L. P. 135 531 Rademaker. F.494. J. 0... J. 242. A.. 66 Kirschbaum. 23. 521. 135 Marroquin. I. 242.... D. J. R. 192 Mueller. E. D. 23. 426 Rosenbrock. 374. 13.. 521 Mosler. C.. A. 522 Maarleveld. 311. G. B. E... 444 Khanderia. 311 Seborg. H. H. G. 310. W. 310... J. 167. 167. G. 444 Rouse. 435. A. V... M.... 192 Ray. S.. 66 Smith. E. 502. 242. D. 522 Lupfer. 441. R. R.. R.. J. 325. D. R. 522 Slaughter. A. 22. C. 478... 310. R. 347. 522 Law.. 522 Lopa. 292 Sastry. C. 107 Rush. 490 McKee. L. M. R.. 444. R. 435.. M. 521 H u g . E. L. 310. c. 484.. 310. D.. 478. R. C. 521 Jaufret. 23. 522 Niederlinski.. 167.. A. 522 Smith. 510. 23. 292 Schnelle. 310 Rothfus. 311 Ramirez.... J. 404. 135 Jarvis. J. 227.. D. P.. 242 Patterson... 464 Otto. J. R. 445. E... 522 Parsons. 192 Ryskamp.. H. 66 Kline.. R. 192 Oglesby. A. C. 521 Seemann. P. 500.. 22. 135 Per+ Chemical Engineers' Handbook. O. D. 151. 310 Hu. F. 192 Prickett... 310 Rhinesmith. 66.445.. 310. D. 522 Maselli. W. C. J. R. Jean-Luc.490. H. 311 Mori. A. 522 King. R. 311 Schellene. 192 O'Brien.. K...Autbur I& H o o p . 311. P. 522 Lamb. 444 Rippin. 229. 23. R. 490 Mathur.. K. 490 Sanders. 192. E. M. 108. N.. 444 Robinson. W.. G. 527 Smith. 307. D. T. O. Hsiao-Ping. 192 Koppel. 242 Shaner. 444 Latham. 23... B.. B. S. Jr.. 135 Orr. 478 Renaud. E. 475. 192. 242 Kennode. 310 Rose. 66 Rollins. 11. P. 442. J. 242. 310 Moore. 521 Rathoye.. H.. A. V. 240. 107 Neumann. N.. 11. C. E. B. 442.. F. P. 522 Mostafa... C. C. H. 313.. 521 Ramaker. B. E. 23 Null. V. 292 Luyben.. F. 522 Morari. 23. S. P. 135 Mehra.. 151. K. I. 192 Rijnsdorp. 23 Kuo-Cheng Chin. 4 4 444. M. R. F. H. 522 Jacobs. M. 491 Nisenfeld. J.. K... J. F. W. 348 Joseph. 442.. 311 Hougen. D. R. 192 Shinskey.. K. 442. 24. D. H. 299... 442.. F. 311 Meyer. 490 Shunta. 491. L.

A.. C. 135 Smith. R. W.464 Sutherland.. A. 475. 242 Wild. 231. L. 444.490 . E. 444 Thal-Larsen. 306. B. M. J. N.. 135 Stevens. 475. A. 490 Tou. 444 Waller. C... J. 491 Wade. 310 Wright. 3 11 Treybal. R. A. T. J. 11.465. D. O... C.. E.. 446. A. 313. M.465.464. 404 Author I& Uim. 108. 66. 0.. 23. D. A.. 326 Tivv... 490 Webber. H.. 310. D.491 Van Winkle. K. 231.. 231.490. H. 310 Waggoner. F.. 446. J. 135 Speicher.445. 242 Weber.. 151 Stanton. 464.464 Tyreus. N.. R. H. 242 Tolher. K. 426 Stempling.. V. 522 Teager. M..465. L:. F.. 66. 192... 310. 521 Snyder. 348.. 490 Wahl. B. 521 Strangio. 107 Williams. H. 441. V.. 23. V.T. 426 Wood. W. 242 Vinante. V.. 192 Vermilion. E. L. K.. D.490. K. H... 326 Vanwormer. T. 444..465. C. 326 Williams. 444. 522 Touchstone. 374..532 Smith. 491 Wood.. T. W. J. 374 Thistelthwaite. J. R. 347.. B. 73. 325. E... 361. E.

100-107 holdup. 327-3% multivariable (MIMO). 273 Dew point. 181-182 Constraints. 390-399 Bottom product. 140-141 527 . 166 unreasonable schemes. 34. 299-303 Feedforward plus overrides. 70. 200 material balance. 279 impedance. 389-390 via steam to reboiler. 1 product quality. 385 Average temperature measurement. 5 15 External reset feedback. 141-143 Feedforward compensation. 336-337 via feed. 70. 99 Analyzers. 11-12. 166. 231-234 Distillate. 58-60 AP measurement. 202-205 Feed systems. 211-213 Autoovemde level control time constant. 153. 241-242 Euler integration. 144-145 Feed temperature control. 241 Averaging level control. 2 1 C. 39-40 Calandria. 2 proportional-integral (reset).167 Controller tuning. 295-297 design procedure. 100-107. composition. 233 Air-cooled condensers. 241-242 Computations. 6-10. 193-227. 80-84 Composition control. 104-107 Condensers. 307-308 Control averaging level. 240 Estimators. 137-151 Feed rank size.. 4. 34. 25 Bottoms. 71. pneumatic. 303-305 Conventions. 70. 375-404 composition. 110 Cascade control. 4. 11-12. 80-84 vacuum. 382-383 Automatic start-up/shutdown. 202. 126-130. 83. 4. 15-16 objectives.465-491 estimators.465-491 design approach to. 405426 Balancing energy and material handling capacities. 72-79 pressure. 200-202 Augmented PI level controller. 25 Bubble point.40-41 Differential vapor pressure. 70-72 Conservation. 19 integral (floating). 420-426 rating. 126-130. 303 Column design. 199 single-loop (SISO).Subject Index Activity coefficients 39. energy. 100-107. 308-309 Base level control via bottom product. 14-15 unfavorable schemes. 213-214. 375-404 Averaging pressure control. 25 Disturbances. 12-13 Divider. 255-256 Condensate receiver level control. 74. 199 proportional-only. 299 Feed enthalpy control. 229 Anti reset-windup. 1. 60-61 columns atmospheric. 250 Double-differential temperature. heat flow. 11-12.

waves. 394. 266 column base. 259-260 Material balance control. 110-116 Rectification section. 109. 260-261 delta P transmitter with double remote seals. 49-64 Median selector. 296-297.528 Feed tray location. 261-262 flush diaphragm transmitter and 1: 1 repeater. 4. 57 q-line. 383 On-line identification. 41-43 Flooded condenser. 195 High selectors. 193-227 Pump bypass. 195 Minimum number of trays. 100. 272-273 Low limiters. 154-166. 194 Internal r d u x calculation. 288-289 Gravity return reflux. 109. 63-64 Minimum reflux ratio. 189 Separation factor (Shinskey). 439 Level control. 262-263 purge system errors. 220-227 PI controller. 266-268 specific gravity compensation. 333. 256 damping chamber. 86-87 Interactions. 243-249 external. pneumatic. 313. 327-346 Shinskey scheme. 16-17. 6-10. 243-249 Reflux cycle. 195 Low selectors. 381 Sensible heat recovery. 169-180 composition control. 43-45 Reset cycle. 309-310 On-line models 305-310 Operating lines. 273 high viscosity fill AP transmitter. 114-1 16. 45-49 Nonlinear PI controllers. 51-54 Orifice. 143 Flash. 14 . 264-266 two flush diaphragm transmitters. 15-16 Murphree tray efficiency. 272 displacer-type. 87-90. 28 column overhead. 92-99 Relative gain array (matrix). 249-2 55 Overhead level control. 468-475 Interlocks. 260 external damping. 69 Maximum capacity overrides. 255-256 Heat recovery schemes. 428 Noise. 246-249 Multivariable control. 116-1 17 internal. 405-426 Product quality control. 54-55 Reflux. temperature and pressure compensation. 210 High column AP override. 357-365 Reboiler types forced-circulation. 349-359 Flooded reboiler. 210 High limiters. 385-386 Pressure-compensated temperature.107 Overrides. 301 PI level controller tuning. 27 internal. 214-215 q. enthalpy factor. 57-58 Raoult's law. 468 Multiplier. 46-91 Protective controls. 4. 170-180 Single-loop control (SISO). 39 Reboiler dynamics. 90-99 Heat flow computations. 117-119 thermosyphon. 375 Level measurement. 62-63 Subject I d t x Modern control theory. 258-259 Nonideality. 109. 119-122 kettle-type. 234-239 Pressure control. 195 Manometer. 3-4. 243-249 Inverse response. averaging. 182-192 High base pressure override. 215-217 McCabe-Thiele diagram. 489 Sidedraw columns. 193-227 for sidedraw columns. 195 Hot vapor bypass. 478-489 Relative volatility. 366-370 Flow and flow ratio conventions. 70. 256-273 characterized displacers.

97 Swell. 289-292 S t m p and shutdown. liquid. column base. 25 Total reflux. 195 pneumatic. 229-230 Tempered coolant. 275 installed flow characteristic. 28-30. 241 differential. 244-246 Sutro weir. 347. 428 Tray hydraulics. 393-394 Temperature measurement average. 75 529 Top product. 234-239 single tray. 189-191 Water saver. 51-54 Subcooling. 348-349 Thennowell installation. 13-14. 211-213 Steam condensate removal. 275 inherent flow characteristic. 279-288 pressure-compensated.. 122-126 Steam supply.258-259 Windup. 276-277 split-ranging. 305 . 92.Subject I m i a Smoker analpc method. 107 Summers. 273-279 flow regimes. 446 Split-ranging of control valves. 200-202 Ziegler-Nichols tuning. gas. 289-292 Vapor-liquid equilibrium. Murphee. 73. 30-49 Vapor recompression. 230-231 double-differential. 275 flow regimes. 100.. 213-214 Tray &aency. 77. 122-126 Stripping section. reset. 276 maximum flow and turndown. 313-326 Valves. 82 Wave noise. control C. 240 dynamics.

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->