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

10 Control-Valve Split Ranging Part 111 QUANTITATIVE DESIGN OF DISTILLATION CONTROL SYSTEMS Chapter 12 Approaches to Quantitative Design 12.5 10.6 Control Valves 11.2 12.8 Temperature-Measurement Dynamics 11.4 Mathematical Model for Combined Trays Chapter 14 Distillation-Column Material-Balance Control 14.3 Derivation of Overall Tray Equation 13.7 Column AP Measurement 11.1 12.1 Mathematical Model-Open Loop 14.8 10.2 Tray Hydraulics 13.5 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 13.9 Chapter 11 1 1.3 12.3 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 12.7 10.9 Flow and Flow-Ratio Conventions 11.1 11.6 10.1 Introduction 13.4 Heat-Flow Computations 11.5 Column-Base Level Measurements 11.viii GmtenB 10.2 Control in the Direction of Flow 320 323 327 327 333 .2 11.

8 Column-Base Level Control Via Condensate Throttling from a Flooded Reboiler (Cascade Level-Flow Control) Chapter 17 Pressure and A P Control 17.4 Material-Balance Control in Sidestream Drawoff Columns 14.3 Level Control of Overhead Condenser Receiver Via TopProduct Withdrawal 16.2 Heat-Storage Effect on Column Pressure 17.1 Introduction 17.c ne t o t ns ix 14.5 Top and Bottom Level Control Combinations Chapter 15 Condenser and Reboiler Dynamics 15.2 Basic Tray Dynamics 18.6 Column AI' Control Via Heat to Reboiler Chapter 1 8 Composition Dynamics-Binary 18.7 Column-Base Level Control Cascaded to Steam FlowControl 16.3 Control in the Direction Opposite to Flow 14.3 Reboilers-Open-Loop Dynamics 15.6 Column-Base Level Control Via Feed Flow Manipulation 16.4 Level Control of Overhead Condenser Receiver Via Reflux Manipulation 16.1 Introduction 18.3 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 .2 Level Control of Simple Vessels 16.3 Pressure Control Via Vent and Inert Gas Valves 17.1 Liquid-Cooled Condensers with No Condensate Holdup 15.4 Pressure Control Via Flooded Condenser 17.5 Pressure Control Via Condenser Cooling Water 17.1 Introduction 16.4 Partially Flooded Reboilers 15.5 Partially Flooded Reboilers for Low-Boiling Materials Chapter 16 Liquid Level Control 16.5 Column-Base Level Control Via Bottom-Product Manipulation 16.2 Flooded Condensers-Open-Loop Dynamics 15.

1 Introduction 19.4 Column Operation Procedure 19.4 Feedforward Compensation 20.5 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 19.1 Introduction 20.6 Composition Measurement Location Chapter 21 Sampled-Data Control of Disti1lat.3 Servo and Regulator Control 21.X contents 18.6 Sampled-Data Control for Loops with Overrides Nomenclature Subject Index Author Index .6 18.4 Feedforward Control 2 1.3 Exact R Procedure 19.m Columns 2 1.2 Design Procedure 19.4 18.1 Introduction 21.5 Interaction Compensation 21.2 Feedback Control of Composition 20.3 Interaction Compensation 20.2 Control Algorithms 21.5 Examples Chapter 20 Composition Control-Binary Distillation 20.5 Relative-Gain Matrix 20.

16 2.18 2.5 2. temperature Temperature vs.3 1.8 2.19 29 31 32 33 34 35 36 37 38 41 42 44 46 47 48 49 50 .15 2. pressure for pure component Typical method of plotting vapor pressure vs.11 2.12 2. composition of binary mixture at constant pressure Pressure vs.4 2.9 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 1.2 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.6 2.14 2.4 9 10 26 27 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 (2.10 2.1 1.1 2.5 2.17 2.Figures 1.2 2.7 2.13 2.3 2.

15 Horizontal condenser.3 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 3.24 2.1 3.14 3.11 3.12 3.6 3.27 2. vapor in shell Vertical condenser.2 3.17 3.23 2. 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 3.5 3.4 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.9 3.7 3.13 3.xii 2.16B Column pressure control by hot gas bypass 3.25 2.22 2.21 2.10 3.26 2.8 3.18 Column pressure control with flooded condenser Gravity flow reflux (flow controlled) and distillate (level controlled) 92 .16A Column pressure control by hot gas bypass 3.20 2.28 2.

12 4.1A 4.13 4.4 4.20 3.5 4.9 4.14 4.2 4.6 4. supply side liquid level 4.22 3.11 4.24 3.17 Distillation column base with thermosyphon reboiler Relationship between vapor volume in tubes of thermosyphon reboiler.16 4.7 4.8 4.27 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. surge tank with Sutro weir.21 3. 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.10 4.1B 4.15 4.3 4.F4qures 3.26 3.23 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 .19 3.25 3. cascade arrangement Proportional-only level control system for column base Column base level control by steam flow manipulation. heat load. 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.

5 5.8 6.9 5.5 7.3 6.1 5.2 6. overhead level control via reflux.9 7.4 5. 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. base level control via feed Distillate demand. overhead level control via top product. reflux drum level control via reflux.6 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.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.1 6. base level control via bottom product Material balance control in direction of flow. base level via feed Bottom product demand.3 7.5 6.XiV Fgures Feed system for distillation column 138 139 140 141 142 5. overhead level control via b i up.1 7. base level control via feed Distillate demand. (B) The five alternative sidestream tray positions and 175 176 .7 6. base level control via bottom product Material balance control in direction of flow.10 6. reflux drum level control via base level control via feed Material balance control in direction of flow. reflux drum level control via dstillate. reflux drum level control via reflux.7 5. reflux drum level control via distillate. base level control via boilup Like Figure 6.8 5.6 5.3 5.4 7.4 6. the toluene impurity content in the dlstillate producer is controlled by the reflux ratio.2 5.2 7.

3 8.8 9.17 9.20 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.9 9.6 9.1 9.3 9.7 9. xv 177 178 179 7. multiple loads 184 185 187 Scheme for establishing heat load priorities source.1 8.P.14 9.16 9. are shown in this blowup.2 9.5 8.12 9.2 8.15 9.7 7.Fipres their controls.19 9.21 D-scheme L-scheme Heat recovery scheme-single Heat recovery-single Heat recovery-single Heat recovery-split source.13 9.18 9. which regulate the benzene and xylene impurities in the sidestream drawoff. 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 . single load-scheme feed.5 9.4 8.8 8.6 9.10 9.4 9.11 9. single load source.

3 10.9 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 11.1 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 9.3 11.9.11 Spechc gravity compensation of head measurement of liquid level 11.18 Level measurement with A' transmitter with double remote I seals 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 11.1 10.2 10.17 Angled nozzle with dip tube 11.8 11.23 9.19 Level measurement with two flush diaphragm transmitters and a summing relay 11.16 Best gas flow purge system 11.24 10.15 11.6 11.5 11.2 11.20 Level measurement with flush diaphragm AP transmitter and 1. t and p Composition vs.1 repeater .7 11.12 11.10 Velocity error in head measurement 11.14 Typical gas flow purge system 11.

of f l d e d condenser 350 353 354 355 . derivative.5 14. PI controller.1 14. v = lO/sec 11.1 15.F&%WJ xvii 281 282 284 285 286 287 290 291 11.28 Split ranging reflux and distillate valves 12.3 Single loop with feedforward compensation.2 Final signal flow diagram for P.3 14. overrides.8 14.6 14.23 Effect on step response of annular clearance 11.4 Signal flow diagram-material balance control in direction opposite to flow Rearranged version of Figure 14.6 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 .2 15.2 13.7 14.25 Effect of annular fill on step response.24 Effect of fluid velocity on step response 11.26 Effect of annular clearance on step response 11.21 Effect of velacity on step response 11.2 12.1 12. First reduction of signal flow diagram of Figure 15. 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 13.22 Effect on step response of various annular fills 11.9 15.1 13.3 14.3 15.27 Split ranging of large and small valve 11.4 14.2 14.

2 16.8 15.11 Final signal flow diagram for base level control cascaded to steam flow control 15.5 16.10 Simplified treatment of heat storage effect on column pressure dynamics Preliminary signal flow diagram for column heat storage dynamics 407 408 .6 16.2 Reduced version of Figure 16. of flooded condenser Reduced signal flow diagram for w.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 16.11 17.15 Signal flow dagram for flooded reboiler for low bohng point materials 15.13 Preliminary signal flow diagram for flooded reboiler 15.6 15.3 16.xviii 15.4 16.7 16.5 15.14 Reduced signal flow diagram for flooded reboiler 15.8 FiguVes 358 359 360 362 363 364 366 367 368 369 372 372 15.10 Preliminary signal flow diagram for column base level control via condensate throttling from a flooded reboiler 16.1 16.12 Signal flow diagram for base level control by direct manipulation of steam valve 15.1 17.16 Reduced signal flow diagram for flooded reboiler for low boiling point materials 16.8 16.9 First signal flow diagram for w.7 15.10 Signal flow diagram for base level control cascaded to steam flow control 15.7 Inverse response predictor for base level control via steam flow 16. of flooded condenser Schematic representation of column base and reboiler holdup Preliminary signal flow diagram for heat transfer dynamics Partial reduction of Figure 15.

5 Partial signal flow diagram for column pressure control via manipulation of inert gas and vent valves Reduction of signal flow diagram of Figure 17.4 18.10 Reduced form of Figure 17.2 18. significant inerts 17.11 Column pressure control via flooded condenser drainnegligible inerts and reboiler steam flow or flow-ratio controlled 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 17.1 19.13 Column pressure control via flooded condenser.4 Combined signal flow diagram for Figures 17.6 18.7 19.7 17.15 Signal flow diagram for column pressure control via manipulation of condenser cooling water 17.00000 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 . reboiler steam not flow or flow-ratio controlled.5 17.16 Equivalent network for vapor flow and pressures in column 17.1I 17.12 Partially reduced version of Figure 17. Guess for xD is too close to 1.Fkures 17.9 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. Guess for R is too small or B.3 and 17.3 18.1 18.5 18.2 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 .Rippin-Lamb model for binary distillation column dynamics Effect on calculation of rectifying section when: A.13 17.8 17.6 17.3 17.18 Column P (base pressure) control via direct manipulation of steam valve 18.14 Partially reduced version of Figure 17.4 17.17 Column AP (base pressure) control cascaded to steam flow control 17.

3 20.5 20.7a 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 with decouplers Partly reduced signal flow diagram of Figure 20.1 20.10 Tracking “dual” control in loop with overrides Conventional control of X2 with set-point disturbance 2 1.4 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. R YT vs.8 21.7d Disturbance in feed composition Convential “dual” control in loop with overrides 21.9 21. 7 ~ Disturbance in feed cornposition 21.3 21.9 20. bottom product via base level control Partial signal flow diagram for Figure 20.7 20.8 20.4 20.2 20.7b Set-point change with compensators 2 1 .3 19.11 21.6 20.5 Composition control of distillation column with feedforward compensation and decouplers 454 456 20.xx 19.1 2 1.5 21.6 21. R qvs.4b “Dual” and PID control for feed composition disturbance 2 1 . Sampled-data control Discrete PID sampled-data control ccDual” sampled-data control “Dual” and set-point control 21.1 Partial signal flow diagram for system with reflux manipulated by reflux drum level Signal flow diagram for system of Figure 20.10 20.1 Tracking sampled-data control of X2 with set-point disturbance 1 .2 21. vs x. 4 ~ “Dual” and PID control for feed rate disturbance 2 1. v.4 Partly reduced signaVflow diagram of Figure 20.4a 466 467 469 470 472 473 476 485 486 487 488 496 498 499 503 504 505 507 yTvs. vs.

13 Tracking sampled-data control of X2 with feed composition disturbance 21.14 Comparison of conventional and tracking PL control 518 519 520 .21.12 Conventional control of X2 with feed composition disturbance 2 1.

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

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. it is not possible to hold exactly constant the compositions at both ends of the column. . that is. satisfactory operation of the column.” -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. such as that caused by rainstorms 3. composition may vary somewhat at both ends of the column. 2.” For some columns compositions are allowed to vary at one end. t For multicomponent columns subjected to feed composition changes. and sometimes both ends. 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. to satisfjr certain economic constraints. but not the nonkey components. For example: -The column shall not flood. Satisfaction of constraints For safe. we can conml two keys. to prevent serious weeping or dumping. With only two drawoffs.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. 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. --Column pressure drop should be high enough to maintain effective column operation. Harbed4 has called this requirement that of keeping the column in “balance.4 StrMeD J%. In most cases material balance will be controlled by so-called “averaging” liquid-level or pressure controls.t It is usually true that minimum operating cost is achieved when the products are controlled at minimum acceptable purities. --Column holdup and overhead and bottoms inventories should be maintained between maximum and minimum limits. This is so because the relationship between thermodynamic work of separation and purity is nonlinear. If feed variations are i the nonkey components. certain constraints must be observed. -Maintain the composition at the other end of the column as close as possible to a desired composition. the composition at one end n must change a little.

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

in Figures 1. As shown by Figures 1. Once the basic concept of material-balance control has been selected for a process. If the design engineer uses the concept of control in the directwn opposite t o w . Generally speaking the direction of material-balance control is determined by the demand stream. concepts. and a frequently encountered arrangement is shown on Figure 1. can lead to many interesting level-control and flow-ratio options.3 and 1. Control in the direction of flow is the most commonly used concept (although the least desirable). is the ratio of maximum required flow rare to minimum required flow rate. For the last tank in a series (final product storage). for simple tanks with level controllers. 1. In this instance the meaning is that. in which case the condensate receiver level sets top product flow.” as used here. or distillate flow. tanks may be smaller and plant fned investment and is used. In the first case. 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. the required change in the manipulated flows will be smaller in one case than in the other. other advantages of control in the direction opposite to flow are (1) less difficulty with stability problems.6 Stratemf Didhttim-Column Control w of design and plant personnel to achieve a mutual understanding of design objectives. and methods is one of the frequent causes of unsatisfactory plant operation. It is for this reason that the first step in designing column controls is to determine the material-balance control arrangement. and (2) reduced internal turndown requirements. one must apply the same concept to all process steps. it is more common to have an operator make the adjustment. in other columns base level sets steam or other heat-transfer media to the reboiler.1 and 1.4. while the level in the base of the column sets the bottom product flow. The working capital can be lower than if umtroZ in the directwn meaning of these expressions is illustrated. These are discussed in derail in Chapter 6 . in response to a given change in demand flow. Here level in the condensate receiver (also commonly called reflux drum or accumulator) sets the top product. In the second case. we can easily use simple automatic controls. the demand flow is always the shipments to a customer.2. 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).’ 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.5.2 ARRANGEMENTS FOR MATERIAL-BALANCE CONTROL As shown elsewhere. in the direction opposite to flow. When more than one tank is involved. . “Turndown.

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

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

o U Q iz&% .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. e 3 - E d m -?$ ' zE= 3 e.

10 Strategy for Dljtillatwn-Column Control FIGURE 1.5 Distillation column with material balance control in direction of f o lw .

Visualization of column operation. let us limit our attention to an ideal. in terms of reflux-to-feed and boilupto-feed ratios. 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 bottom-product-to-feed ratio will be fixed. . Rippin and Lamb’ have shown that. (2) the reflux-to-feed ratio and the bottom-product-to-feed ratio.1. was suggested by Uitti7 and has since been proposed in varying degrees bv many others. although the results will be applicable in a general way to multicomponent systems. Similarly. As will be shown in Chapter 2.3 Funahmentnls of Composition Control 11 1. particularly those that may be treated as quasib&ary or pseudobinary systems. Methods of handling other disturbances will be discussed later. This is somewhat restrictive. then we want the “operating lines” on a McCabeThiele diagram to remain constant when feed flow changes. particularly when the column is not making a sharp separation. if feed composition and feed thermal condition are constant. To simpl@ the analysis. basis for column composition control. reflux-to feed. 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. then the distillate-to-feed ratio will be 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. for small perturbations. boilup-to-feed. 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. Throughout the rest of this book. for example. since the bottom product flow is the difference between the stripping section liquid and the boilup. and bottom-product-to-feed ratios are held constant.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. In considering case (1).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. since the distillate flow is the difference between the reflux flow and the vapor flow. it will be used as the primar). 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. The constants KO Kf4 may be calculated approximately by the column designer.

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

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

and control valves have improved greatly in reliability. flows. * Digital computers and microprocessor controls. energy-recovery systems. Since then primary measuring elements. computing devices. smaller and fewer holdups. Startup/shutdown will be discussed in more detail in Chapter 9. pressures. usually a valve. For newer plants with higher throughputs.” and must implement process logic by switching in and out of ccautomatic. and one final control element. for a given algorithm. the traditional “instrumentation” approach to control is often seriously inadequate. and sometimes compositions. Each process operation has a number of independent or single loops for feedback control of temperatures. progress in single-loop design has been essentially at a standstill. one controller. Further.” are so sluggish that they are ineffective in dealing with typical ‘disturbances. and speed of response. the quality of control achieved by single-loop systems is not greatly affected by type of hardware. The controllers usually have neither antireset windup nor automatic tracking. whether one uses analog pneumatic or electronic gear. 1.* In addition. Consequently these characteristics less and less frequently pose limitations to the single-loop designer. however.6 CONTROL SYSTEM DESIGN PHILOSOPHY Current Design Practices In the preface we noted that we would try. it is not surprising that some loops never work in “automatic.” Others.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. or a digital computer. The term “single loop” means there is one measurement. let US look at typical controls in existing plants. process control was hardware limited. single output). As recently as 1950. but not all. with heavy emphasis on conformity to past practices. microprocessor controls. however. As a consequence the operators must perform startup operations with the control stations switched to “manual. For some years now. and elaborate materialrecycle systems. controllers. and there is little or no logic circuitry to tie the many loops together. operation closer to hard constraints. it makes little difference. in this book. although stable in “automatic. .” Since the original design procedures are usually qualitative and intuitive. sensitivity. This statement is true for both analog and some digital hardware. to present a multivariable control approach to distillation column control. Many. 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. liquid levels. 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. control stations. Before discussing this. of these loops are represented in the central control room (CCR) by control stations.

they yield results that are far inferior to those obtainable with well-damped feedback controls with simple feedforward and override control. depending on circumstances. it is increasingly common to use microprocessor controls instead of analog (see discussion under “Hardware Conventions and Considerations”). precision. we use an approach we call multivariable control. If base level control is normally achieved bv a PI controller that can be overridden by high or low base level proportion&-only controls.1. and reliabilinr that are completely beyond the capabilities of human operators. For example. To provide automatic control of this sort. From .* For plants being designed today (late 1983). we call that “variable structure. the steam valve for a distillation column reboiler. 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. in a given situation. but most of them are expressed in mathematical terms rather than in terms of process hctions. Multivariable control structures. If. 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. however. base composition is normally controlled by steam flow that can be taken over or overriden by high column AP. 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. this is a variable configuration. Hardware permits us to automate this kind of control with a speed.’ Many definitions of this term may be found in the literature. the best of several preprogrammed strategies (algorithms) for manipulating one or more control valves (or other final control elements). we make extensive use of “variable configuration” controls that are usually implemented by overrides.” Multivariable control may involve both variable configuration and variable structure controls. particularly when computers are involved. are functionally more like the modern “matrix” concept of management. with their extensive lateral and diagonal crossovers. for instance. 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.6 Control System Des@ Philosop@ 15 optimized tuning procedures for unaided feedback controllers have limited practical value for continuous processes.

The traditional objectives of feedback control system design are: 1. troublefree operation. Select process designs that eliminate or minimize interaction. To compensate for or to attenuate disturbances as much and as quickly as possible. In practice we furnish calibration data for controller parameters and computational devices for the majority of control loops prior to startup. Speed of Response. 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. 3. such as pressure control and flow control at the same point in a pipeline. Stability. 2. Our design procedures are accurate enough that only a modest amount of empirical controller tuning is required at startup. clearly are different from those just mentioned. . Similarly. For microprocessor computer controls. the shorter lines of communication and decentralized control functions permit more rapid and stable control. To get the fastest possible response to set point changes. 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. We calculate these from simulations or simple linear models. Avoid designs that are intrinsically interacting. Further. 2. we calculate scaling parameters for computation blocks (either in software or hardware). Design material-balance control loops to be at least a factor of 10 slower than related composition control loops. which means that controls must be added and must be kept in “automatic” for stable operation. For example. A typical chemical plant or refinery has hundreds of single loops with many interactions among them. For example. These must be accomplished with a reasonable degree of closedloop stability. and Interactions Most existing literature on automatic control is concerned with the stability and speed of response of single loops. Our preferred philosophy of controller tuning is discussed in a book’ and a paper. and more reliable.” There are five simple methods by which to make a system noninteracting: 1. The objectives of averaging level control. * Some processes are open-loop unstable.” In process control the objectives are often quite different. One of the two controllers must be detuned. for example. make the secondary or slave loop at least a factor of 10 faster than the master loop.16 Strategy fm Distillation-Column Control a control engineering standpoint. 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. it is usually undesirable to make rapid changes in manipulated variables since these may upset the process. in cascade systems.

we recommend dual stations. 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. 4.6 Control System Des@ Philosqphy 17 if a tubular reactor is fed at several points along its length from a common header. * Much of this section may seem superfluous to instrument and plant personnel. level. or microprocessor based. In the c‘loca”’ position. Hardware Conventions and Considerations* For control loops represented in the CCR (central control room). the station may be physically distinct. This may or may not be subject to restrictions imposed bv feedforward compensation. but also another fUnction-“remote-locaI. o however. overrides. such as top and bottom temperature controls on a distillation column. overrides. This is discussed in Chapter 20.” These may be analog pneumatic. valve position may be set manually or the controller set point may be set by the operator (c‘local-auto”). the secondary station not only has manual-automatic switching. In the last case. Most single-station designs with which we are familiar are very inflexible and complicated. 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.Although cascade functions are sometimes combined into one station (or CRT “faceplate”) for space and money-saving reasons.). the desired value (set point). eliminate interactions. the feedback controller is disconnected and there is a knob that enables the operator remotely to set the valve position. interact. etc. there are some very sophisticated mathematical methods for dealing with interaction^.” In the “remote” position. and so on. analog electronic. by permitting only one controller at a time to control a given valve. Our experience.” Each provides an indication of the process variable (flows. as. really). depending on the design philosophy for a particular project. for example. or may be represented on a CRT display as a “faceplate. and so forth. and the valve loading signal (controller output. column designers. Use override circuits. Use interaction compensators (decouplers). If cascade control is involved. the secondary controller set point comes from the output of the primary controller. and s for& is that this discussion may be very helpfid. which some vendors label “hand-automatic. If two control loops. with process design engineers. we can eliminate the interactions by installing two compensators. There is also a “manualautomatic” switch. . liquid level control cascaded to flow control. override circuits.^^.” In the “manual” mode. In addition.1. it is normal practice (as mentioned earlier) to furnish “control stations. temperature. they do not permit ready implementation of feedforward.^^'^^ Some are intended for noninteracting design while others seek a design that provides an optimum amount of interaction. Although not specifically intended for this purpose. like an analog station. 5.

we have a decided preference for microprocessor controls. however. Digital readouts for important variables are worthwhile because they permit seeing their magnitudes with sensitivity approaching that of the original analog measurements. can be reconfigured or restructured without wiring changes. many are featuring satisfactory antireset windup and override capabilities. and contain more control algorithms.000. It is our opinion that using such a computer to imitate unenhanced two. it is possible to find worthwhile applications for a supervisory digital computer with a good data historian. Microprocessor controls usually have a sampling time of a fraction of a second.* Piston actuators with double-acting. coercive stem forces when the valves are in flowing streams. Better control valves. however. Valve positioners should always be used (it is assumed in this book that they will be). As far as CCR hardware is concerned. With regard to these. where reset time and proportional gain are functions of some process variable or the controller error signal. It should be noted. and adaptive tuning. Engineers’ consoles. and the fact that they can be calibrated more precisely. Computer consoles were originally provided in the CCR for the convenience of the operators. They are technically more versatile and are less expensive (some versions) than analog. have a larger range of tuning parameters. Positioners compensate for this and maintain the valve’s inherent flow characteristic expressed as a function of controller output signal. Others have gain scheduling. Some worthwhile applications for computers will be discussed later. For most projects today. Separate consoles for maintenance personnel (the computer can be a powerful maintenance tool) are highly desirable. . Some of the last named achieve self-tuning via stochastic techniques or by pattern recognition. * Most control valves today are either single-seated global types or rotary types. their performance can generdy be approximated by analog control algorithms.18 Strategy fm Dirtillation-Column Control As of this writing (late 1983).5-2 percent. regardless of the type of basic controls selected (pneumatic. In adchion. facilitate technical studies. the two biggest hardware needs appear to be: 1. Although slightly slower than analog controls. perhaps at another location. As of late 1983. Better measurements. For both there are substantial. two-stage positioners are recommended. they provide more advanced logic capability. 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. Most analog readout devices. dead-time simulation. more progress has been made in the design of valve bodies and trim than in the design of actuators. Most typical analog measurements have a sensitivity ranging from one part per 1000 to one part per 10.and three-mode analog controls is poor practice. electronic. Other advantages include freedom from drift. 2. especially of compositions. or microprocessor). that some users prefer not to use positioners. These are position rather than velocity algorithms. are limited to 0.

4. Calculate al other overrides. This is due to the trend toward increasinglv tight column design. Column diameter and tray spacing are now kept to smaller values. 9 . Add protective controls and automatic startup/shutdown circuits. Calculate feedforward and interaction compensators.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. prepare a simplified flow sheet that defines the control concepts: a. Use simulation for some columns. Select the overall material-balance control scheme first. l 8. c. The number of trays is held down because smaller allowan& are made for uncertainty in tray efficiency. Select measurement spans and calculate control valve sizes. Quantitative information for the other steps is presented in Part 111. Calculate feedback-controller gain and reset settings and control-loop natural frequencies. and after careful review of the overall process flow sheet. 1. With holdups determined calculate column-composition transfer functions.8 COLUMN DESIGN PHILOSOPHY AND CONTROL SYSTEM DESIGN Experience on many projects shows that even small. preferably l proceeding from final product inventory to raw material inventory. 6 . 2. Part I1 of this book deals with the qualitative and heuristic aspects of steps 1 and 3. particularly those in critical service or with a new untested control system or process configuration. d. It may also save startup time. 5.1. Determine holdup volumes required for smooth material-balance control and for liquid-level override controls at each end of the column. e. 7. Prepare material-balance and composition-control signal flow diagrams. Check feed-tank material balance and mixing time constants for adequacy. Add feedforward and interaction compensators as required. 3. with the . Hardware vendors may now be selected and the measurement and control equipment may be ordered. Simulation of the column and its control system will be usefd in confirming control concepts and controller tuning parameters.8 Column Des@ Philosophy and Control System Des@ 19 1. b. 1. simple columns benefit from a modest application of feedforward compensation and overrides. After careful discussion with the process engineer and column designer. Select composition control schemes. Add miscellaneous temperature and pressure controls. Al individual equipment-piece material-balance controls must be consistent with this scheme.

1. Avoid sidestream drawoff designs. 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). Incorporated. our opinion that these controls should be used to some extent on almost every column. 5. that the customer insists on minimum application of feedback controls with no feedfonvard compensation or overrides. Design for normal operation at 60 percent of the rate for flooding. 8. Provide five extra trays or increase the number of trays by 10 percent. 6. 9. whichever is larger. we suggest the following: 1. and columns with heat recovery schemes. This incremental investment not only provides better normal control. it is fairly common practice to use excessive boilup and reflux to make sure of meeting or exceeding product specifications. What column design philosophy should be followed?Having had considerable adverse experience with columns with primitive controls. the opportunities for their improvement.* 2. 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. Frequently encountered problems include unstable or ineffective controls. existing columns. Provide surge tanks between columns with at least 30-60 minutes of holdup each. Provide larger reboilers (lower heat flux). 8000 hours per year). Control column Al’ by boilup. Provide increased tray spacing.20 Strategy fm Dljtillatwn-Column Control result that columns typically operate closer to flooding. The combined effect of these design policies is to make columns much touchier and harder to control. but it also helps to avoid inadvertent shutdowns. Provide 100 percent reserve capacity in heat-recovery schemes or avoid them altogether. In addition. and how to troubleshoot them when necessary. 3. Suppose. and f l d g or dumping. Provide larger condensers. it also reduces column capacity. 7. however. hs it is pertinent to consider briefly the controls of typical. use tempered water. It is. . or flow control steam.9 EXISTING COLUMNS-TYPICAL PRACTICES AND TROUBLESHOOTING Although ti book deals primarily with &stillation control in design projects. one may note that 100 Ibm/hr of steam is worth $3200 per year (basis $4. particularly sidestream drawoff columns. To save this amount of steam would probably be only a modest accomplishment for most columns. This not only wastes energy. To provide a perspective on energy savings. off-specification product or products.00/1000 pounds. 4. If water cooled. therefore.

the most obvious deficiency of most exisring columns is the lack of appropriate composition measurements. At this point with flows established. 6. review the control schemes in . ratio controls should be replaced by impulse feedforward compensation . and a minimum steam flow ltmiter to protect against dumping. If PI level controllers are’used. 8. Column feed rate should also be held between maximum and minimum limits. If composition is measured and is cascade controlled via reflux or boilup. For further improvement provide steam-to-feed ratio control. internal reflux-to-feed or distillate-to-feed ratio control. For pressurized or vacuum columns. make sure an adequate scheme is provided for maintaining an inert balance. This is done for convenience. Frequently composition control is attempted at only one end of the column. auto overrides or nonlinear controllers should be used. (see Chapter 12) if feed flow turndown is greater than 2: 1. in some cases. it will probably be possible to see some improvement in composition control. . Another shortcoming frequently observed is the use of fixed flow controls for steam. 5. and that hardware. Pneumatics. and. especially level and flow transmitters and control valves. 1.IO Cunventwns Followed in thlj Book 21 the case). in which case flow measurement should be linear. In view of the preceding comments about problem areas and likely opportunities for improvement of composition control and reduction in energy consumption. pressure. . Most commonly temperature in the upper or lower section of the column (or both) is used in lieu of true composition measurements. is in good working condition. smoothed. 2. at least for part of the time. Chapter 3. Make sure that column material-balance controls are properly designed and tuned. we will frequently illustrate control schemes with pneumatic components. Any such unaided flow control should be regarded with suspicion. follow the tuning procedures of Chapter 16.1 . With rare exceptions flow-control set points must be changed to accomplish either composition control or material balance control. It is usually desirable to cascade level control to flow control. Provide a linear flow measurement for steam flow control. reflux. 4 If condenser controls are a problem. or product drawoff. Also provide temperature and pressure compensation. or both. flow measurements to deduce compositions. the following guidelines are suggested: 1. If temperature by itself is not an adequate measure of composition. 7.I 0 CONVENTIONS FOLLOWED IN THIS BOOK Throughout the remainder of this book. Provide averaging level control of column feed. 3. and in some cases limited. high A P override on steam to protect against flooding. and sometimes at neither end. consider one of the schemes in Chapter 10 for using temperature.

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

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

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

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

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

Outputs from the box are the two product streams D and B with compositions x. required to make the separation is approximately the heat added to the reboiler. and x . is fed back to the top tray of the column. pcu/hr. Distillate product is removed from this drum. Heat is transferred out of the condenser at a rate g.. The energ\. The heat-transfer rate is qR energy units/time (e.2 Control variables for distillation column . mol fraction light component. This liquid reflux and the vapor boilup in the base of the column are necessanr to achieve the separation or “fractionation” of chemical components. mols/min). and reflux drum as a “black box” process..2).2.. Overall View f o a Control Perspective rm One can stand back and look at a distillation column with its associated reboiler. The vapor coming from the top of the column is liquified in another tubeand-shell heat exchanger called a condenser. Btu/hr. condenser. called ccreflux” ( L o . joules/minute. Liquid from the condenser drops into the reflux drum. with a distillation column is that the feed rate and feed composition must be FIGURE 2.) . The usual situation . In addition.g. etc. 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. and reflux are inputs into this box (see Figure 2.1 Introduction 27 rates D and B mols/min and compositions xD and X. some liquid. pcu/hr. heat. Feed.

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

2 2 Tray Hydrauliu . 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. and drops into a “downcomer. or if vapor-liquid contacting is poor. 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. Thus there is an increme in pressure from the top of the column to its base. It is able to do so because the liquid phase is denser than the vapor phase.3.” which provides liquid for the tray below through an opening at the base of the downcomer. See Figure 2. FlGURE 2. Liquid must flow against this positive pressure gradient.” It is important in control system design because it imposes very important constraints on the range of permissible liquid and vapor flow rates.3 Schematic of typical sieve tray . the separating ability of the column drops drastically. passes over a weir. 29 flows across each tray. providing more or less effective hole areas as vapor flow rate changes. Vapor flows from one tray up through the tray above it because the pressure is lower on the upper tray. Valve trays are built with a cap that fits over the hole in the tray and that can move up and down.

t If this occurs. etc. exerts a certain pressure at a given temperature. 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-liquid equilibrium (VLE) data and analysis are vital components of distillation design and operation. Pressure is measured at various temperatures. 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. 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. The temperature at which the pure component exerts a pressure of one atmosphere is called its “normal boiling point. vapor-liquid contacting is poor and fractionation suffers. . number and size of holes. there are maximum vapor and liquid rates. The control system must keep the column from flooding. outlet weir height.)..30 Fundamentals o Dktillation f This pressure difference depends on the vapor pressure drop through the tray (which varies with vapor velocity. Tray “flooding”* occurs when the liquid height in the downcomer equals or exceeds the height between trays (tray spacing). A measurement of column pressure drop can also be used to prevent flooding. On the other end of the scale. Vapor Pressure The liquid phase of any pure chemical component. This pressure is called the pure component ‘‘vapor pressure” P. 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. * For a further discussion of flooding.” Light components have low normal boiling points and heavy components have high normal boiling points. 2. Therefore. These concentration differences are analyzed and quantified using basic thermodynamic principles covering phase equilibrium. see pages 424-430 of reference 8.4). if vapor rates are reduced too much.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. speciesj . vapor density. etc. These hydraulic constraints can be handled in control system design by using maximum and minimum flow limiters on heat input and reflux. It is a physical property of each component.) and the average liquid height on the tray (which varies with liquid flow rate. This is usually due to excessive boilup (vapor rate) but sometimes may be caused by excessive reflux.

and B. = = = vapor pressure ofjth component in any pressure units (commonly mm Hg. psia. a nonlinear dependence of vapor pressure on temperature is obtained.3 Vapor-Liquid Equilibrrum Fundamentals 31 When the data are plotted on linear coordinates (see Figure 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.6.5).4 Vapor pressure and temperature measurement * A three-constant version of the Antoine equation is used in Chapter 10. atmospheres. + B.2. where = A. ESSURE GAUGE VAPOR LIQUID TEMPERATURE MEASUREMENT FIGURE 2. Vapor-pressure data often can be described by the Antoine equation*: In P./T (2. H a ) absolute temperature (degrees Kelvin or Rankine) constants over reasonable range of temperatures Therefore. Note that the constantsA.

pressure for pure component . using the symboly. Vapor composition is expressed as mol fraction light component. Then temperature and pressure are measured. Samples of the vapor phase and liquid phase are taken and analyzed. Liquid compositions are usually expressed as mol fiaaion of light component and the symbol x is used. VAPOR PRESSURE P (mm Hg) 760 "NORMAL BOILING POINT" FIGURE 2 S .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. When working with a distillation column. Temperature vs. Mixture composition is then changed and the procedure repeated.32 Funhmentals o Dirtillation f They can be easily calculated by knowing two vapor-pressure points (P. 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 ) . B = (Td(T2)In pz Tl .

one merely draws a horizontal line at the given temperature and reads off x and y values. as illustrated in Figure 2. T-xy Diagrams If the data are taken with pressure held constant (isobaric). FIGURE 2. P-xy Diagrams Data taken at constant temperature (isothermal) are plotted as two curves: pressure versus x and pressure versus y. either light or heavy. it is convenient to plot two curves on the same paper: temperature versus x and temperature versus y. temperature s .6 Typical method of plotting vapor pressure v . B. since these points correspond to pure components.” A. and at the pressure under which the data were obtained. To determine the composition of liquid and vapor phases in equdibrium with each other at a given temperature.8.2. Note that the ends of the curves must intersect at x = y = 0 and at x = y = 1.3 Vapor-Liquid Equilibrium Fundamentals 33 These results are conveniently presented in graphic form using several types of “phase diagrams.7 itlusrrates a typical T-xydiagram. Figure 2.

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

2.3 Vapm-Liquid Equilihum Fundamentals 35 TEMPERATURE CONSTANT AT T x and y (MOLE'FRACTION LIGHT COMPONENT) FIGURE 2. composition of binary mixture at constant temperature .8 Pressure vs.

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

and Note that we can talk about either bubble-point temperature or bubblepoint pressure. depending on which variable is fixed (isothermal or isobaric situations).2. 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. These calculations are called bubble-point.10 Bubble point and dew point at constant temperature . dew-point. VLE Calculations Instead of using graphical techniques. TEMPERATURE CONSTANT FIGURE 2.11).3 Vapor-Liquid Equilibrzum Fundamentals 37 The same concepts can be visualized using constant-pressure T-xy diagrams (Figure 2. The mixture is superheated vapor at temperatures above the dew point TDp subcooled liquid at temperatures below the bubble point TBp. and flash calculations. The same is true for dew-point temperature and dew-point pressure.

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.11 Bubble point and dew point at constant pressure . A somewhat simplified equation representing this condtion is: 3.38 Fundamentals o Distillation f A. Pr = mol fraction ofjth component in liquid = mol fraction of jth component in vapor = total system pressure PRESSURE CONSTANT x and y FIGURE 2. pT = xj pj Y j (2-4) where x. y .

. increase C T. Bubble-Point Calculations In all bubble-point calculations.) of thejth component is defined as the ratio of vapor composition ( y t ) to liquid composition (x. distribution coefficients or “K values” are customarily used.3 Vapm-Liquid Equilibrtum Fundamentals = = 39 P . We will assume ideal VLE behavior for the rest of this section for purposes of simplicity. If PP is less than P. C‘ 6. Calculate: x< P.P. y. 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.+ ( T ) = CX. The system is “ideal” and obeys Raoult‘s law (7. B. = number of components 4 Check to see if . = 1). Nonideality will be discussed in more detail in Section 2. The problem is to calculate the unknown temperature or pressure and the‘ composition of the vapor phase (3.3. there is little attraction or repulsion of neighboring molecules of different types.). In addition we must be given either the pressure or the temperature of the svstem. The calculation procedure is iterative: 1. If not. In the petroleum industry. we know the composition of the liquid (XI’S are all given). If the components are chemically quite s d a r . ( T ) j=1 where N. If P T is greater than PT. the K value is simply the vapor pressure divided by the total system pressure. The K value (K. is sufficiently close to PT. This is by far the most common type of calculation encountered in distillation work because column pressure is usually known. 3. 5. reguess T and go back to step 2. When convergence has been achieved. If the system is “ideal” ( y j = I). 2.2. Bubble-Point Temperuture Cakulutwn. Guess a temperature T . decrease T.). calculate vapor compositions. Calculate vapor pressures of all components at T .

25 120°C XIP. Bubble-Point Pressure Calculation.40 Fundamentals . the lightest 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. on the other hand.’s are all given) and either temperature m pressure. The vapor rising in the column gets richer and richer in light components at each stage.071 Notice the enriching of lighter component that occurs in the vapor in the above example.671 0. has a higher concentration in the vapor than in the liquid.25 = P. 3 1 2300 1000 380 pcalc -j 9 5 1365 920 350 2600 1140 440 pcalc T - - 1040 399 130 1549 0. C. The liquid moving down the column gets richer and richer in heavy components. In this case temperature T and liquidphase composition are known. Dew-Point Calculations In dew-point calculations. P. Dew-Point Temperature (PGiven).’s. at XI B 0. Solve iteratively €or the temperature that satisfies equation (2.35 X 0.35 Guess T 125°C 0-Xylene (X) x3 = 0.258 0.9). 0-Xylene.40 Toluene (T) x2 Guess T = 120°C = 0. Given: P T = 1520 IIUII Hg Benzene (B) x1 = 0. Nc N‘ .fDistillation Example. has a higher concentration in the liquid than in the vapor. we know the vapor composition (y. This illustrates precisely why a distillation column can be used to separate chemical components. Benzene. the heaviest component. at 125°C XIP.40 T 0.

Variables that are unknown are liquid and vapor compositions and liquid and vapor flow rates.11) (2. Material of known composition zJ is fed into a flash drum at a known rate of F mols/min. Isothermal Flash Calculations These calculations combine vapor-liquid equilibrium relationships with total mass and component balances.10) (2. See Figure 2.9) (T D.L (2.) N.3 Vapor-Liquid Equilibrizcm Fundamentah 41 Rearranpg: 1 pT= J=1 c WP.12 Isothermal flash . The equations describing the system are: F=L+V z2F = x. (2-9) Dew-Point Pvessu~e Given). Both the temperature and the pressure in the drum are given.12.2. Calculate pressure directly fiom equation (2.12) + ~/IV 3/1 = xJpJ/pT FIGURE 2.

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

The larger the relative volatility a. The volatility of component j relative to component k is defined as: (2.15) (2. Values of ajkclose to 1 imply that the separation will be very difficult. Figure 2.18) where x and y are mol fiactions of light component in the liquid and vapor phases respectively. Before starting any flash calculation.17) A large value of relative volatility a+ implies that components j and k can be easily separated in a distillation column. real (V/F) ratio.18) leads to the very useful y-x relationship that can be employed when a is constant in a binary system.14 sketches y versus x lines for various values of a. both Pop and PBpcan be calculated explicitly with no iteration involved. 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.2.3 Vap-Liqud Equilitrrium Fundament& 43 equation. For an ideal (Raoulfs . For binary systems relative volatility of light to heavy component is simply called a: (2. Rearrangement of equation (2. requiring a large number of trays and high energy consumption. but this is not the correct. Relative Volatility Relative volatility is a very convenient measure of the ease or difficulty of separation in distillation. 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. That is.16) Since T is known. the fatter is the equilibrium curve. (2.

a will be independent of temperature. yP= = P L X (1 . particularly when the components are chemically similar.x ) (2.14 Relative volatility on x-y diagram .20) (1 . Distillation columns are frequently designed assuming constant relative volatility because it greatly simplifies the vapor-liquid equilibrium calculations.44 FutzdmaentaLr o Didlatim f law) binary system. relative volatility is.19) . This is true for many components over a limited temperature range.y ) / ( l ._ (2.PL YIX .4 px If the temperature dependence of the vapor pressure of both components is the same. a= FIGURE 2. In other words.y)PT = Px(1 .constant if the vapor pressure lines are parallei in a L P versus 1/T n plot. a can be expressed very simply as the ratio of the vapor pressures of light and heavy components. Relative volatilities usually decrease somewhat with increasing temperature in most systems.

Equation (2. they exert a lower partial pressure than if they were ideal.225 0.45 0. the system may break into two-liquid phases with different compositions in each liquid phase.16). This can lead to formation of a “maximum-boiling“ moa-ope (Figure 2.16. A modest amount of repulsion can lead to the formation of a minimum boiling azeotrope (Figure 2.945 0. applying the basic definition [equation (2. there can be repulsion or attraction between dissimilar molecules. Negative deviations (attraction) can give a higher temperature boiling mixture than the boiling point of the heavier component.5 2.084 Nonideality In most distillation systems. An azeotrope exists when the liquid and vapor compositions are the same (x. 2. Activity coefficients are usually calculated from experimental data. This is called a “heterogeneous” azeotrope (Figure 2. Figures 2. When chemically dissimilar components are mixed together (for example. and given liquid compositions (xis) and relative volatilities (ais). Wilson.517 0.18 sketch typical phase diagrams for these..) are used to correlate activity coefficient data. Activity coefficients are less than unity (negative deviations). Azeotropes occur in a number of nonideal systems. Given the liquid compositions and relative volatilities.35 0. Positive deviations (repulsion) can give a lower temperature boiling mixture than the boiling point of the light component. calculate the vapor compositions: -5 0. .17). = There are several types of azeotropes.20 ffj xJ ffl YJ 3.17)] and rearranging lead to the following explicit relationship between any vapor composition y. where a is the volatility of thejth component relative to some arbitrary base component (usually chosen as the heaviest component. In this case the activity coefficients are greater than unity (called a “positive deviation” from Raoult‘s law).15 sketches typical activity coefficient data as a function of the light component composition x l .4) contains y.399 0. oil molecules and water molecules). (2.1 1 1.17. If the repulsion is very strong. Figure 2. CYjXj Yj = N. etc.200 0.2. they exert a higher partial pressure than if they were ideal.) Example. If the molecules repel each other.18). the liquidphase activity coefficient of thejth component. and 2.21) . x).3 Vap-LtquiCt Equdihum Fuhmentals 45 For multicomponent systems. Empirical equations (Van Laar. If the molecules attract each other. the predominant nonideality occurs in the liquid phase because of molecular interactions. Margules.

0 72 1.46 FunAamcntaLr o DirtiuatiOn f 4 3 7 1 2 1 I I 0 0.o Xl FIGURE 2-15 Typical activity coefficients as functions of light component composition .5 1.o Xl 2.5 ' 1 0 05 1.

16 Qpical homogeneous “maximum boiling” azeotrope .2.3 Vapor-Liquid Equdibnum Fundammtak 47 FIGURE 2.

3 o~fDistilhm FIGURE 2.48 FundamCnta.17 Homogeneous “minimum boiling” azeotrope .

2. The control engineer should be aware that the existence of azeotropes imposes restrictions on the operation and performance of a distillation column. These techniques are very useful in gaining an appreciation of the effects of various design and operating parameters.4 Gqhical Solution Techniques 49 A detailed discussion of azeotropes is beyond the scope of this brief introduction.18 Heterogeneous azeotropes .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 2.2.

in mols/minute.50 Funriamentals $Dirtillation These effects are usually apparent on the diagrams.. These are assumed to be constant REFLUX RATIO = R FIGURE 2. so one gets a picture of the process.19 Simple distillation column . Liquid and vapor rates below the feed in the stripping section are called Ls and V. System Figure 2.19 summarizes the system and nomenclature to be considered. Liquid and vapor rates in the section above the feed (“rectifjmg‘‘ section) are called LR and V.

20): L x n +1 Rearranging: = vsyn + B XB (2. F = D + B (2.27) (2.25) (2. The parameter q will be used to describe the thermal condition of the feed.22) Reflux ratio is widely used as an indication of the energy consumption. This constancy of molar flow rates is what we call “equimolal overflow.23) (2. Stripping Section B = F . In many systems this is a pretty good assumption. Thermosyphon.24) if feed conditions and product compositions are specified. + .2. This straight line can be plotted on an x y diagram (see Figure 2.D A light component balance around the nth tray in the stripping section yields (see Figure 2.28) This is called the operating-line equation. R = LR/D (2. A total condenser is used to produce liquid reflux and distillate product.26) B. kettle.24) F z ~ x ~ +I x ~ B = These relationships must be satisfied under steady-state conditions. A.4 G r a p W Solution Techniques 51 on all trays throughout each individual section.X B ) (2. internal. and forced-circulation reboilers are all usually partial reboilers.21). It has the form of a straight line: y = rmc b. The ratio of the internal reflux flow rate L R to the distillate flow rate D is called the reflux ratio R. The slope of the line is the ratio of liquid to vapor flow rates in the stripping section. D = F(zF - xB)/(xD . The reboiler is a “partial reboiler” (vapor is boiled off a liquid pool). The composition of this liquid pool is the same as the bottom product composition.” Primarily it assumes that the molar heats of vaporization of the components are about equal. Note that distillate and bottom product rates can be calculated from equations (2.23) and (2. Equations Overall Balances Mass and component balances can be written around the entire column system.

-t B XB Ls xi = Vs xi + B XB (2. = Vsy..29) (Ls . Substituting into equation (2.20 Material balance on stripping section .VS)xi = BxB LS %+I VS Yn B XB FIGURE 2. This is easily proved by letting the point of intersection be xi = yi.27) gives: Ls X .52 Fu&ntals of Dit&& The intersection of the operaring line with the 45" line (where x = y) occurs at x.

21 Operating line o stripping section f .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.

= B.21). 1 + D X D LR D (2. Ls . yn = -xn+l 7 LR X" +1 VR yn FIGURE 2.22 Material balance on rectifying section . Ls = V. 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).54 Fundumentah 0 f D i s t i l h k From a total mass balance around the system in Figure 2. B. Therefore.29) gives: + Bx. Rectifying Section A similar component balance around the upper part of the column above the nth tray in the rectifjmg section (see Figure 2.30) + -XD VR VR This is the equation of another straight line called the rectifying operating line. Substituting into equation (2. Figure 2. C.V.20.= B x B x i = XB Therefore.22) gives: VRyn = L R X n .

if we know x E . and Ls/ .4 CraphiEal Solution Techniques 55 It intersects the 45" line at X. Stepping off Trays Tray-to-tray calculations involve the solution of vapor-liquid equilibrium relationships and component balances. as shown in Figure 2.23 shows both operating lines together with the y-x vapor-liquid equilibrium curve. Our VLE relationship gives us y. 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 ) ) . let us start at the bottom of the column with known values of x. and has a slope equal to the ratio of the liquidto-vapor rates in the recufylng section. For example. V. a FIGURE 2.24.Graphically. Figure 2.2.23 x-y diagram showing both stripping and rectiqring operating lines .

We know y. We could plug into equation (2. Then this ni 1 Y YB X FIGURE 2. YE = Vsy. Alternatively. + Bx.31) x B = -x1 L S VS + ($) (2.24 McCabe-Thiele dlagram-stepping between VLE curve and operating lines to estimate number of trays requlred . A component balance around the reboiler gives: Lsx.56 Fu-talr ofDktd. and then read off the yB value.hti~t~ we simply move vertically up a line through x. until the VLE curve is intersected. (2.28)]applied to the reboiler stage.31) and solve analytically for xl. utl the operating line is intersected. 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.

)yi = F(l (LR - L. q is a fraction.4 Gaphical Solution Techniques 57 same procedure is applied again to the first tray. and the stepping is continued until x. q= It follows that 1 - L S .). B % B) Using equations (2.33) q is the fraction of feed that is vapor: (2.%i + (x.2. depending on the temperature. + .LR F (2. is reached. calculate the total number of trays NT and the feed tray N F ) using this graphical technique.V.24). Values of q greater than 1 indicate subcooled liquid feed.35) . As we will show in the next section. Feed Thermal Condition The feed to a distillation column can be liquid or vapor. and composition of the feed. x2 is determined from y1 by moving horizontally to the stripping operating line. q = 1. = + ZF F (2. If the feed is a vapor-liquid mixture. Use of equations (2.%D D Subtracting gives: (VR . If the feed is a saturated vapor at its dew point. (2. q = 0. Then the rectifying operating line is used. pressure. or both.33) and (2. Now let us look at the intersection points ( x f . yi) of the stripping and rectifying operating lines. The number of trays in the rectifying section can be determined in this manner.. the parameter q is defined as the fraction of the feed that is liquid. Values of q less than 0 indicate superheated vapor feed.24). To quanti@ the thermal condition of the feed. 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. having specified product compositions and operating line slopes. This determines ie the number of trays required in the stripping section.34) If the feed is a saturated liquid at its bubble point.34) gives: - q)J. Thus we can design a column (Le. the operating line slopes are both known if the feed thermal condition and reflux ratio have been specified.28) and (2. D (-@)Xi + .30) gives: VRy = g LR x. yl is determined from x1 (move vertically to the VLE curve).

lines must lie on the q line. If the reflux is a saturated liquid at its on. Feed: zF. while increasing trays increases investment costs. This will be discussed fUrther under "Limiting Conditions. Operating external reflux ratio: R = Lo/D * If external d u x is subcooled-mually the case-then where C.58 Fu-tals ofDid&!&m This is the equation of a straight line with slope ( . VLE curve in x-y coordinates 4. which can be easily drawn given zFand q. Figure 2. To = vapor temperature.36) Thus the rectiqing operating line can be drawn if xD and R are specified. Thus the intersection of the operating . Design Problem We are now ready to summarize the graphical design technique for determining the number of trays required to achieve desired product purities. (2. Then = molar latent heat. It is a straight line with a slope of R / ( R + l). The slope of the rectifjmg operating line (LJVR) can be expressed in terms of the internal reflux ratio LR/D = R. distillation design involves a classical engineering trade-off between the two design variables: reflux ratio and number of trays. Since increasing reflux ratio increases energy costs (V. Then the smpping operating line can be drawn if xB and q are specified.25 shows q ha for several values of q.q / ( 1 . XB 3. q 2. Desired product purities: xD. AM temperature of Lo. intersecting the 45" line at xD.4)). = molar spec& heat. given a reflux ratio. and T = R ." Given: 1. As we will show later. It is a straight line joining xBon the 45"line with the intersection of the rectlfylng operating line and the q line. the lower the reflux ratio specified. the more trays are required. It is called the q line and intersects the 45"line at 2.F. bubble p i t " LR = Lo. = Lo + D ) .

q ) ) and draw 4 line from zF point on 45" line. and ZF on 45" line. 3. Calculate B and D fi-om overall balances [equations (2. 2. 5. Calculate liquid and vapor flow rates in reg and stripping sections.4 Graphical Solution Techniques 59 Calculate: 1.2. Total number of trays: NT 2.26)]. Draw VLE curve. Feed tray location: NF Procedure: 1. 6. 4.25) and (2.q / ( 1 . (SATURATED) LIQUID Y X FIGURE 2. Draw 45" line. Locate xE.xD. Calculate slope of 4 line ( .25 q-line on x-y diagram .

trays on which the vapor and liquid streams leaving the trays are in perfect phase equilibrium. Don’t let this worry you. depending on the system.39) (2. There are a variety of possible rating problems. When this stepping procedure crosses the intersection of operating lines. if there are multiple feed points available on the column. Thus the feed tray NF has been determined. Lo.” as opposed to the “design problems. 10. The first step corresponds to the partial reboiler. implying a noninteger number of trays. The feed tray may also be fixed. 12. it may be varied. Calculate slopes of operating lines: rectifying = LR/VR.?any other feed tray would require a greater total number of trays). 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). In this case the total number of trays in the column N T is fixed.q ) F s = LR 7. The next step is tray 1.” in which NT is calculated. stripping = LS/VS. In this procedure we have assumed “perfect” or “theoretical” or “100 percent efficient” trays.38) (2. is at its bubble point: LR = Lo = (R)D (2.37) (2.40) VR = LR Ls +D + qF v = VR . This last step will not go through the xD point exactly. Actual industrial columns seldom achieve this ideal situation. 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. When the xD point is crossed. 13. Start from the XB point on the 45” line and step up the column from the stripping operating line to the VLE curve.(1 .e. Draw rectifjmg operating line from xD point on 45” line with slope LR/VR 9. Typical efficiencies run from 40 to 90 percent. These fixed-column problems are called “rating problems. this is the “optimum” feed tray (i. and so on.If the external reflux. 11. 8. 14. that is. or. this is the total number of trays NT. Switch to the rectifying operating line and continue stepping. Draw stripping operating line fiom XB point on 45”line to the intersection of the q line with the rectifying operating line. The two most commonly encountered are (with NT and N F fixed): . Check your calculations by seeing whether the slope of stripping operating line is Ls/Vs.

VLE equilibrium relationships. mal-and-error solution techniques. 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. Notice that in both of these problems. and specified variables) from the total number of system variables.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.26 McCabe-Thiele diagram for rating problem . and energy balances. 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.2. Y X FIGURE 2. two variables must be specified to define the system completely.26). This magic number of two occurs again and again in distillation (see Section 4). component. 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.

Economic optimization studies have led to the commonly used heuristic (rule of thumb) that the optimum actual reflux ratio is 1. Figure 2.1 to 1. Increasing reflux ratio requires fewer trays (less capital cost) but increases energy costs. It occurs when the operating lines just intersect on the VLE curve. Y TAKES AN INFINITE NUMBER OF TRAYS TO STEP PAST THIS “PINCH” POINT.2 times the minimum reflux ratio (see Figure 2.fDistih&n A.27 Column operation at minimum reflux ratio . 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.62 Funahmat& .27 shows the normal minimum reflux ratio situation. An infinite number of trays are required to step past the feed plate because of the “pinch” condition (the converging operating and VLE lines). The actual reflux ratio used must be higher than the minimum.28). X FIGURE 2.

=1 VR v s Thus a column with fewer than the minimum number of trays cannot achieve the desired separation.4 Graphiurl Solution Techniques 63 In some unusual VLE systems.Ls _-. No feed is introduced and no products are withdrawn.2. The L/V ratios in both sections of the column become unity and lie on the 45" line (Figure 2. B. the pinch between the VLE curve and an operating line can occur at some point other than the feed point. but heat is added in the reboiler and all the overhead vapor is condensed and returned to the column as liquid reflux. RID = m. 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. reflux ratio s . This situation actually takes place in a column when it is operated under "total reflux" conditions. COST It) REFLUX RATIO (R) FIGURE 2. Since D = 0.29). Also: LR .28 Costs v . even at very high reflux ratios.

Y X FIGURE 2. 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.29 Minimum number o trays required at total reflux f . Fenske equation can the be used to solve analy~callyfor the minimum number of uays (AIT)-.64 Fundumtak OfDistdkkm For a system with constant relative volatility aLH.

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

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

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

particularly if 316 SS tubes are used.1). If a smooth flow to the next step in the process is needed. 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. it is also better suited to applications in which it is desired to absorb the maximum amount of low boilers in the condensate. takeoff should be by “averaging” pressure control. this type of condenser cannot be operated partially “flooded. For the same reason. fiom column overhead composition. Horizontal shell-and-tube condenser with liquid coolant in the tubes and vapor on the shell side (Figure 3. With a longer condensing path.2). if possible. with less pressure drop) through condensers of this design. and distillate overflows from the vapor-liquid disengagement space beneath the condenser. the speed of venting inerts at startup time is limited. Vertical shell-and-tube condenser with liquid cool an^ on the shell side n and vapor entering the tubes at the top (Figure 3.1 has two vents. a reflux drum. Because all vapors must pass through the tubes.2 TYPES OF CONDENSERS In chemical and petroleum plants.70 Overhead Syst. In addition. we find at least five different kinds of condensers: 1. which also serves as a condensate receiver. Since the exit water is hot. As an alternative. The cooling water valve is normally at the exchanger exit to make sure the tubes are fdled at all times. Sometimes condensate is taken out through two drawoffs instead of one. -Same as foregoing except that distillate flow is set by reflux drum level control.” Again. This is probably the most popular type in petroleum refineries. column inerts are usually vented more easily (ix. vapor may be taken off on flow control cascaded fiom top composition control while column pressure is controlled by heat input.. In this book we usually refer to the condensate receiver-the vessel that receives condensate &om the condenser-by another name: reflux drum. thereby minimizing corrosion. the cooling water valve is located at the exit to ensure that the shell is flooded. If the column top product is a vapor. the valve may need anticavitation trim. it is much better suited to partially ‘‘flooded” operation.. This condenser commonly has at its lower end a vapor-liquid disengaging pot. this term is widely used in the petroleum industry. r r a y m a A -Flow control of reflux is cascaded. By comparison with the vertical design discussed below. . The design itlustrated in Figure 3. with averaging level control of distillate. 2.Although it is less exact. at startup time. 3. each with a valve if the exchanger is operated flooded (see further discussion in Chapter 15). Some designs bring the vapor in at one end and vent uncondensables at the other. should be employed.

-Air-moZed umdensers (Figure 3. -Horizontal tube bundle with wohnt i tubes (Figure 3. The vapor n comes up &om below and condensate drops into an annular space around the vapor nozzle. or externally through a gravity flow line with a control valve.5).2 TrparofCondensen 71 3. in-column head condenser (sometimes called “dephlegmator”). and comes down the tubes with the condensate. Here condensate is recirculated through VAPOR CONDENSATE FIGURE 3.3). This design comes in two variations: “reflux” design. Reflux may return internally via an overflow weir. where the vapor goes up the tubes and is countercurrent to the condensate f a h g down.3.4) -Spay ma’ensen (Figure 3.1 Horizontal condenser. -Ve?-td bunde with wohnt on shell sia!e. vapor in shell . Here we have a number of different designs. and a design with a chimney in the center such that the vapor rises in it. reverses direction. The latter has a “hat” over it to prevent condensate fiom dropping back down the column. Internal.

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

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

4 Air-cooled condenser .74 Overhead Spent A m a q p w n ~ FIGURE 3.

For the vemcal condenser. This will have a secondary advantage of reducing the probability of cavitation in control valves and pumps. If water-header pressure fluctuations are a problem. lo Section 6. be located in the liquid line just beneath the condenser for maximum speed of response. use dual. Spray condenser .3.8. split-range water valves. If summer-winter heat-load variations are sufKciently severe. 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. (See Figure 3.) 4 For the horizontal condenser. the temperature detector preferably should . Select the number of degrees of subcooling so that the sensible heat load will be at least 5 percent of the total heat load. 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. 3. 2. (See a s discussion in Chapter 11.3 Atmospheric Columns 75 The suggested approaches to a v o i h g these ddKculties are as follows: 1. The smaller valve should open first and will provide adequate winter cooling. use a cascade temperature water-flow control system.) FIGURE 3 5 .

6 Preferred overhead s s e for atmospheric column ytm .76 Overbead System Arraqgements FIGURE 3.

or under vacuum. This keeps the condenser dynamics constant and eliminates the problem of retuning the controller as the load changes (see Figure 3.9).10). to compensate for changes in condenser dynamics as condensate rate changes. An override from cooling-water exit temperature is also normally needed. 7.3. 6. 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..7 Alternative overhead system for atmospheric column .. As an alternative to Item 5.. Another. or perhaps adaptive gain and reset. a limitation to this technique: for protection . There is. say 3-5 psig. however. This is discussed more fully in the next section.. FIGURE 3.3 Atmaspheric Columns 77 5.-.....-. completely different approach is to run the column at a slight pressure.. The controller should have auto overrides (see Chapter 9).

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

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

even though each valve is sized to handle a maximum flow equal to five to ten times the average.11. the low-pressure source is usually a steam jet.10. and difficulties with. condensate temperature control. For a vacuum column with a small amount of inerts. the vent or bleed valves may be tied to the pressure controller to work in parallel with the condenser cooling-water valve. the arrangement of Figure 3. For many columns the vent flow functions primarily as a purge and is small enough that moderate changes do not affect column operation. For other cases. An example is shown in Figure 3. is better. it is common to minimize or eliminate split-range overlap to reduce consumption even further. Since the sum of the two acoustic resistances is always high. the other closes. A more complicated but more flexible arrangement. it may be necessary to use a cascade temperature-vent flow-control arrangement. 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. As far * The symbolism “ A 0 means air-to-open. The vent line is connected to a pressure-dividing network with two control valves connected so that as one opens. and that must handle severe transients during the startup period. where column feed rate varies signhcantly. If the downstream pressure fluctuates too much. an additonal large vent valve is sometimes installed in parallel. In this event the arrangement of Figure 3. or (2) throttling an air or inert-gas bleed if there is only a small amount of inerts. “AC” means air-to-close. such as that of Figure 3. It should be acknowledged that many engineers today prefer to control condensate temperature by manipulation of condenser cooling water. Section 10) permits both valves nearly to close when the controller output signal is at its midrange value. The objections to. It has worked well. pressure is then controlled by (1) throttling the vapor takeoff if there is a large amount of inerts. A split-range adjustment of the two positioners (see Chapter 11.12.80 Overhead System Arrunpnents 3. with a controlled bleed from the atmosphere (or source of inert gas). . For large columns that must be started up and shut down frequently.13. This facihtates getting the column on line at startup. normal flow of air or gas through the two valves in series is economically small.10 may require an impractically small vent valve. were stated earlier. and split-ranged with the small one.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. When an expensive inert gas such as N2 must be used.* is well suited to either vacuum or pressure columns when the amount of inerts fluctuates over a wide range. for example. via condenser cooling-water manipulation. on a column in a semicontinuous process that is shut down and started up every day or so. For a vacuum column. In such cases a manually set vent or bleed valve is ofien adequate and no direct control of condensate subcooling is necessary.

4 Vacuum an/i P m r e COltrmnr-Lq~id P m d m 81 FIGURE 3.3.10 Overhead s s e for vacuum or pressure column-large amount o inerts ytm f .

If cooling water is adjusted manually. Section 2). So-called "water savers" are cooling-water exit-temperature controls. pressure is controlled by manipulation of makeup and vent valves.7. This kind of control is often implemented as shown in Figure 3.11 Overhead system for vacuum column-small amount o inerts f . They have the advantage of minimizing cooling-water flow rate for any given heat load.82 Operhead System An-anpnenB as we can tell. the flow is either insufficient or excessive. 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. Their use also minimizes subcooling-and there are instances where this is desirable-but at the expense of variable condensate temperature.

3. gradual correction. if used. 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.4 Vacuum and Pressure Columns-Liquid Product 83 For a large amount of inerts.12 Alternative overhead system for pressure or vacuum column-small inerts amount o f . it is certainly feasible to control pressure by throttling the vent flow. We have previously indicated that pressure control. the engineer should keep in mind that these columns have a narrow range of operation. this procedure is recommended for columns with a vapor product (see next section) where condensate temperature is not conuolled. which provides slow. In designing controls for vacuum columns. should usually be of the "averaging" type. One last consideration should be noted here-that of dynamics. This fits in well FIGURE 3.

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

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

Several configurations have been employed. In one system the condenser is at a lower level than the receiver by 10 to 15 feet. we can calculate wR. we can calculate the heat transferred.the reflux flow rate in porncis per hour. qc. 3.6 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.15 Alternative overhead system for pressure column-vapor product .86 &erhead System Anmgemenk and flow rate. 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. This means FIGURE 3.

1. as suggested by Holland& and shown in Figure 3. or by injecting inerts into the incoming vapor partially to blanket the tubes. 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. mverse response”. At low heat-transfer loads. the controller opens either the distillate or reflux valve. has the condenser mounted above the reflux drum. These hot-vapor bypass systems are not recommended for systems with even small amounts of inerts.3.16B. the horizontal condenser of Figure 3. As suggested by Chin3. perhaps as a consequence of increased boilup. The pressure controller pinches the bypass valve to force more vapor into the condenser. Suppose. Maximum cooling-water rate is normally used. Another problem was observed by Mueller5 on a partial condenser. violent surging and hammering may ensue.17.6 M i c e h m Pressure-Cuntrol Techniques 87 that the condenser runs partially flooded. as shown in Figure 3. 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. lower level. other examples will be discussed later. as shown in Figure 3. The liquid in the refiux drum is subcooled. This results in a temporary increase in pressure since it takes time for the condenser level to drop.* This may be minimized by designing adequate clearance into the condenser. thereby lowering the liquid level. however. Common practice. which limits heattransfer capabilities. Consider. Eventually. Pressure drop of uncondensed vapor from inlet to the two exits caused a low liquid level in . Dynamic problems with such a system can be severe. the hot-gas line around the condenser has m valve in it. If there is insficient clearance between liquid level and the top of the shell. which permits a higher rate of condensation and causes the pressure to be restored. for example. A vertical reflux drum is recommended to reduce this interfacial area. so there is condensation of vapor at the liquid-gas interface in the drum. Flooded Condenser A pressure-control technique that is growing in popularity involves partial flooding of the condenser without a hot-gas bypass. for example. Another technique sometimes encountered involves throttling the vapor to the condenser. A mathematical analysis is presented in Chapter 15. condenser contents drop to a new. At low heat-transfer loads. thereby dropping the liquid level and increasing the heat-transfer area available for condensation. the liquid level will run high. If the pressure gets too high. that the column pressure has risen. It also lowers the operating pressure on the condensing side. This suffers from the drawback of requiring a large valve.16A. A valve on the liquid from the condenser floods the condenser to hold column pressure. The vapor enters at the center and uncondensed gas exits at the two ends. There are some practical problems that must be taken into account. The temporary “wrong-way” pressure response is commonly called <<. the liquid inventory in the shell was high. Another configuration.

88 fierbead Spem Awaqpun& FIGURE 3.16A Column pressure contml by hot gas bypass .

6 Mkcellaneow Presswe-Control Techniques 89 FIGURE 3.16B Column pressure control by hot gas bypass .3.

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

one uses the arrangement of direct return reflux and overflow distillate. Use of this technique leads to the arrangement of Figure 3.18. The reflux flow or flow-ratio controller will usually be fast enough that this will not be a problem for reflux flow. unless the drum has a large cross section. may have to be cascaded to distillate flow control. If the individual valve is shop calibrated. the condenser can be located overhead with only a modest increase in structural requirements over a ground-located condenser. Since the surge tank needs a level controller. there is no savings in instrumentation. a plot of reflux flow versus valve stem position will be a straight line. Here reflux drum level is controlled by throttling distillate flow. For all gravity-returnreflux systems. Overall a properly designed gravity-flowreflux system is sigdcantly cheaper than a pumped-back reflux system. a more expensive columnsupporting structure is required. A disadvantage. then the reflux does not come from the overhead surge drum and this vessel can be located at ground level. The support problem can be minimized. is that variations in level will cause momentary changes in both reflux and distillate flows. it backs liquid up into the receiver. Another method of controlling gravity return reflux is shown in Figure 3. lo however. and it is harder to remove the condenser tube bundle for maintenance. But one a s nseds a higher head cooling-water pump. which is especially important if fouling and corrosion are problems.7 Gravity-ReturnR t f w Venus Pumped-Back R@ux 91 is often easier to maintain. that is. Then. The level controller. . however. then valve stem position can be accurately related to reflux flow. Since this tank with its contents is often far heavier than the condenser.19. coolant-in-shell condenser. one may design a distillate overflow system that provides constant head for reflux.20. flow will vary only when the valve position is changed. the scheme of Figure 3. When this valve is pinched. For this application a valve with linear trim will have a h e a r installed characteristic if line drop is negligble. To avoid this problem. If. Here a reflux flow measurement is coupled through a controller to a distillate valve. for example. by building the condenser into the top of the column. For maximum effectiveness the liquid pool in the vapor-liquid disengagement space should have a large cross-sectional area. Reflux Flow or Flow-Ratio Control When reflux is flow or flow-ratio controlled. Perhaps the most common arrangement is that of Figure 3. it is also probably safer since there is no pump to fail. 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. with head across the reflux line fixed.21. on the other hand.3. which features a vapor-liquid disengagement space built into the lower section of a vertical-tube. and the overflow weir should permit a wide range of overflows with only a small change in head. but the equipment that needs to be installed at a high elevation is minimized. Consider. particularly if the overhead surge drum is also located at the top of the column. piping and instrumentation can be very simple. If an overhead condenser is used.

The Sutro weir has the advantage of being a linear weir. An elegant way of doing this is to cause reflux to overflow through a Sutro weir.6 Distillate Flow or Flow-Ratio Control For those columns with gravity return reflux. where distillate is on automatic flow control or column-composition control. a severe oscillation in overhead vapor flow to the condenser is sometimes encountered. FIGURE 3.18 Gravity flow reflux (flow controlled) and distillate (level controlled) . This is commonly called “reflux cycle” and has a typical period of several minutes. that is. 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.22. as shown in Figure 3. It has been observed primarily in columns where reflux flow is the difference between rate of condensation and distillate flow rate.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. is the condensing temperature. -Select the flow-metering orifice to hold liquid head in the reflux line at (. -Use a condenser with a recirculating coolant. T.3. -Increase column operating pressure. -Keep condensate subcooling to a minimum. -Use a large-diameter vapor line to reduce acoustic resistance and to increase acoustic capacitance.’ 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./dP. This increases vapor density and decreases dT. thereby fUrther improving stability.19 Liquid-vapor disengagement space built into condenser . CONDENSER L I QU I0 -VAPOR DISENGAGEMENT SPACE FIGURE 3. or use a short. -Use a horizontal condenser with vapor and a generous fiee volume on the shell side. &a at the bottom of the pot with the Sutro weir. there will be no reflux cycle. If subcooling is zero. vertical condenser with many tubes (vapor inside tubes).

20 Gravity-flow reflux system with ground-located surge tank for distillate .94 o~erbead Spem Arranpnents FIGURE 3.

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 .

96 Overbeead System Awqqement~ FIGURE 3.22 Control o gravity reflux flow rate by overflowingthrough Sutro weir and by throttling f distillate flow .

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

98 Overhead S j m m An-aqpncn& FIGURE 3. surge tank With Sutro weir.23 Gravity-flow reflux. T~ > 3-5 minutes .

4 Undesirable piping arrangements for returning reflux to column FIGURE 3 2 . They have demonstrated.5 Preferred piping arrangement for returning reflux to column . the piping may enter horizontally. 3 8 CONTROL TECHNIQUES WITH AIR-COOLED CONDENSERS . certain control problems. The preferred arrangement of Figure 3.25 avoids ti kind of flow instability.8 Gmtrol Tecbniqua with Air-Cmlcd condcnrm 99 This phenomenon is particularly troublesome with vacuum towers where some slight air leaks are unavoidable.3. or hs with a slight indination as shown. however. In recent years air-cooled heat exchangers have gown enormously popular. They are far more FIGURE 3 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. Small holdups favor good composition control. control of the column with which the holdups are associated. This permits sensitive. Speed of response is greater and condenser dynamics do not change with load changes. 4. 3. It eliminates problems with high-freezing-point condensate that might p l q the condenser if once-through coolant were used.4). and the condenser must be designed for a small pressure drop on the coolant side to minimize pump horsepower requirements. 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. 2. Variable-pitch fans. Various techniques have been devised to control the rate of heat transfer or to compensate for condensate temperature changes: 1. 3. 7. Condenser dynamics are radically improved over those achieved with once-through coolant. AND REQUIRED HOLDUP In this section. 2. Tempered coolant is employed for either or both of two reasons: 1. Flooded operation. But when the holdups are part of a feed system for another process step. comments or suggestions regarding required holdup will be primarily fiom the standpoint of getting good. In extreme cases tempered coolant is taken &om and returned to a supply tank that is temperature controlled. Internal reflux computers (see Chapter 11). The temperature rise per pass is kept small. rapid temperature control.100 Overhead System Arraqqemenk sensitive to atmospheric changes such as rainstorms or even changes in wind velocity &an are liquid-cooled exchangers. This is discussed in more detail in Chapter 5. Variable-speed fan drives. . or at least adequate. Adjustable louvers in the exchanger housing to control suction air flow.9. requirements may be much greater. 6 . Use of induced-draft rather than forced-draft exchanger designs. low-head pump therefore is required. 5. Condensate-temperature and column-pressure control are easier. Partial bypass of hot liquid from upper section of the exchanger and mixture with cold liquid leaving at the bottom. 310 LEVEL CONTROL OF CONDENSATE RECEIVER . A high-flow. Toplocated fans provide much better protection against rainstorms (see Figure 3.

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

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

27 Proportional-only condenser seal pot level control via reflux flow .3.10 L m l G m t d OfcOnaCnrate Rtctipcr and Required Holdup 103 FIGURE 3.

System performance then should be checked by frequency-response methods or computer simulation. and for pneumatic instruments various methods are available for minimizing lags and improving speed of response. by the use of special techniques. to design for rH or [rH]oR less than 1 minute. For electronic-analog or microprocessor controls. the limiting factor will be the speed of response of the valve-positioner valve-actuator combination. High-performance valve positioners probably will be required. Typically. Kc = 2. valvesizing procedures lead to: (3-4) where k. the various control schemes usually will require a rH 2 2 minutes. that is. for process reasons. A AHT and . If pneumatic instruments are involved.’ Experimental data for long pneumatic tubing runs are gven in reference 10. = 4. 7. the level nozzle spacing usually must be determined before control valves are sized. Holdup time is usually considered to be equal to volume divided by throughput. Process engineers often think of “holdup time” rather than a time constant. If.. This may require a rH much greater than 2 minutes. however. however. one should make sure that override time constants are at least 1 minute. is a multiplying factor typically in the range of 2-6. and A$& = A$L: = ~QFS AHT = Kc keQFs A rH (3. if additional volume is required for smoothing out feed to the next process step.5) In the discussion that follows. for a linear installed valve characteristic. horn equation (3. the available holdup must be very small. it is sometimes possible. [ T H ] o R 2 1 minute.3). For any given scheme.104 Overhead S s e A m n p n e n B ytm During the early stages of a design project. If we think in terms of the volume corresponding to the level transmitter span. it should be preferably in a separate vessel outside of the reflux path.) Then. for k. As noted earlier. (See discussion on valve sizing in Chapter 11. 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.

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

and reflux-flow control loops are fast compared with the level control loop. ft3/min and K R is the override relay gain and K. Level Control Via Reflux Flow For the system of Figure 3.12) This indicates that override action may be extremely rapid compared with that of normal level control.13) .< = 2. If K. For gravity flow reflux. it may upset the process. ‘H [‘HIOR = (3.7.11) If the flow-meter spans are in the same ratio as the average flows ( 5 : l ) . while a level below 25 percent must close it. x 12 ~ 3 2 Q ~ QmfD (3.25. since such designs are often plagued by a “reflux cycle” as mentioned in Section 3. = 0. Here it is assumed that both the distillate. 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 . This is not usually desirable. is the subcooling factor discussed O in the next section.27. .then: -[‘HIOR [‘HI - 32 x 5 = 160 (3. . For this reason we mostly choose KOR = 2 if possible.7. if possible. ft’/min = reflux flow-meter span. level control via reflux should be avoided. For level control via reflux. K.2 for an AC reflux valve and a level above 75 percent must open the distillate valve.10) Therefore.106 Overhead S s e Arraqements ytm and where Qmp Qmfl = distillate flow-meter span. the characteristic time constant is defined a little differently: (3. the entire condenser-reflux system should be designed according to the recommendation in Section 3. If this design cannot be avoided.

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

. Buckley. San Francisco. 61-66. B. 1964. and W.. Luyben. Wdey. Apr. June 1975. ulation of Vertical Subcooling f 9. Buckley. Quantitative Des@ o Condensers. INTECH. 1974. D. P.” INTECH. Apr. Shunta. Tyreus. 43“Designing Long-Line Pneumatic 48. S. S. Council. part 2.” INTECH. New York. 1969. .” Automatic Control Pneumatic Control Loops.108 Overhead System Arranpnem 8. Jan. part 1. Level. P. Control Systems. P. and W. 1976. 39-42. P. Buckley. S. 1983.. “Modelling and SimControl. Fehervari. Technzques of Pmce~s 11. ccNontinear Control of Liquid 10. L.. 12.. J. 33-40.

to permit entrained vapor bubbles to rise and separate from the liquid pool. It requires simultaneous consideration of fluid mechanics. and second.1 INTRODUCTION he design of a column base with its associated reboiler can be a comdex ” matter. Vortex breakers should be installed routinely. mass transfer. 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. can be largely offset by item 4. For example. This is required for two reasons: first. 109 . 4.4 t ~ Column-Base and Reboiler Arrangements 4. or the falling liquid drops will entrain too much vapor into the liquid pool. Vortexing (like the swirl at a bathtub or slnk drain) is undesirable since it promotes entrainment of vapor into the drawoff line. however. There should be a minimum liquid level depth above the nozzle to the drawoff line from the column base or vaporizer separator. This difficulty. 2. the following must be considered when a vertical-thermosyphon. This is typically one vapor nozzle diameter and normally will be specified by the column designer. to minimize the likelihood of vortex formation at the drawoff nozzle. they should have an orientation no more than 90” from the vaporreturn nozzle and should not be located under the last downcomer. In severe cases this may cause foaming in the column base and “gassing” of the reboiler. forced-circulation. heat transfer. Spacing between the vapor-return nozzle and the lowest tray should be large enough to minimize entrainment of liquid drops in the rising vapor. Minimum liquid level should not be too far below the vapor nozzle. 3.’ 5. or kettle-type reboiler is used: l. which may cause “gassing” of the reboiler or bottom product pump or circulating pump. and process control. If level-measurement nozzles are not protected by an internd damping chamber.

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

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

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

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

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

overrides. performance tests. and sometimes by another process variable.4. a liquid-level transmitter should be installed on the reboiler shell side. startup. 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. Instead condensate is removed through a loop seal whose top is vented to atmosphere. For troubleshooting.4. A special version of a flooded reboiler designed for low-boiling materials (requiring low-temperature steam) is shown in Figure 4.3 Flooded Tbemmypbon (Steam-Side) Reboilers 115 Control of the condensate valve is sometimes achieved by a steam-flow measurement. Response is more sluggish than when steam flow is throttled.3 Flooded reboiler . in some cases. sometimes by a condensate-flow measurement (more accurate and cheaper). The theory and design equations are presented in Chapter 15. The height FIGURE 4. No trap is required. and.

This arrangement has the equation: H L P BL = H s p B L + Ps + AI'..5.4 Flooded reboiler for low boiling point materials .116 Cohmn-Base and Reboiler Awanpnents of the loop is typically 5-10 feet.4 FORCED-CIRCULATION REBOILERS Forced-circulation reboilers are commonly horizontal as shown in Figure 4. To minimize pressure drop. feet Ps p shell pressure. but also occasionally are vertical. a restriction downstream of the reboiler is sized to prevent vaporization in the reboiler tubes. 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.ne B C B C (4-1) where HL Hs = loop-seal or standpipe height. Ibf/fi? absolute condensate density. This restriction may be in the vapor line to the column or right at the vapor nozzle outlet. 4. feet = condensate level in = = the shell. FIGURE 4.

for applications where a thermosyphon would be expected to foul.5 Column base with forced-circulation reboiler . 4.5 Fhkd-Bundle Kettle Rebuih 117 Forced-circulation reboilers are widely used for vacuum columns because of their lower pressure drop. The usual arrangement is that of the flooded-bundle type FIGURE 4.4. and for applications where thermally sensitive materials are being distilled. and avoid the temperature elevation encountered in the lower end of a thermosyphon reboiler. They eliminate the need for the circulating pump required with a forced-circulation reboiler.5 FLOODED-BUNDLE KETTLE REBOILERS Kettle-type reboilers are sometimes used with vacuum columns.

6. This is very similar to an internal reboiler with an isolating baffle or chamber. which provides a high signal the first time the reboiler is filled with liquid. as will be discussed in the next section. to suitable overrides. Shown here are two overrides (pneumatic devices illustrated).7. 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.6 Kettletype reboiler wlth internal weir . or. One has a latch-up circuit with a gain 25 relay. This can be connected to an interlock.” 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. Once latched up this circuit does not function again until it is unlatched by switching to “shutdown.118 Column-Base and Reb& Arrangemats shown on Figure 4. To protect the tube bundle in the event that boilup temporarily exceeds downflow. the steam valve is held closed u t l the level ni covers the tubes. a head measurement must be made on the tube bundle chamber. as shown on Figure 4. This requires FIGURE 4.

4. usually in the form of horizontal. 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. however.7 Protectbe circuits for tube bundle chamber in kettle-type reboiler . have been used to a modest extent. Offsetting these in part.6 INTERNAL REBOILERS Internal reboilers.6 Internal Reb& 119 field calibration since the presence of froth between the tubes produces an average density well below that of clear liquid. cylindrical bundles of U-tubes with the heating medium in the tubes. as are piping and reboiler shell.4. In comparison with external reboilers.

Since most level instruments really measure head.Here the tube bundle* is submerged in a pool of liquid. lbm/ft3 To get adequate holdup for level control.120 Column-Base and Reb& Arrangements adequate space for controlling the fioth level. Ibm/ft3 (4-2) The Hepp7 correlation may be used to estimate 4: 4 where = 1 . If. the level transmitter is calibrated such that “100 percent” level is 76 inches above L2. 4 = 0.8 is suggested. not level. it is more conservative to keep the bundle submerged. Note that L1 and L3 are the nozzles used for level control (see discussion in Chapter 11).5. this is no trivial p r ~ b l e m .5. the nozzle spacing of Figure 4. not clear liquid.L3 and Lz are used to determine that the bundle is covered by liquid. for example. Some bubbles. the column must be shut down to clean the tubes. 2. exist in the spaces between the tubes. Tube bundle with isolating chamber.62 Y GV (4.0.3) v = vapor velocity. of controlling the level of a variable-density f o h In addition. the top 24 inches it between Lz and L3 are clearance. of course. 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. we need twice the volume that would be required for clear liquid.8. ~ From the column or reboiler designer it is necessary to get an estimate of relative froth density: = Froth density. 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. Two basic arrangements for installing and operating internal reboilers are employed: 1. Tube bundle without isolating baffle or chamber. A indicated by the s literature. ft/sec pv = vapor density. . lbm/ft3 Clear liquid density. sometimes it is a two-pass bundle extending through the column with the intermediate head on the far side. Thus this type of internal reboiler probably should not be used in vacuum towers since the required base froth volume would be too great. The latter design is better than the former for flooded-rubeoperation. * Sometimes the bundle consists of U-tubes. we must take 4 into account. For the particular case where C#I = 0. Tube Bundle Without lsolating Baffle or Chamber This case is illustrated in Figure 4. To allow for variations in f o h density. There is also the added difficulty it. Above the bundle we normally find fioth.

6 Internal Reboilers 121 FIGURE 4.8 Column base with internal reboiler-no baffles o weirs r .4.

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

9 Column base with isolated internal reboiler .7 Steam Supply and Condenrate Remmal 123 FIGURE 4.4.

This figure also illustrates temperature and pressure compensation of the steam flow meter.2). Header pressure and temperature variations severely impair the accuracy of many uncompensated plant steam flow meters.124 Column-Base and Reboiler A r r a y w n t s be used instead. and less probability that flow regimes will switch during normal operation (see Section 4. FIGURE 4. Errors in steam-flow metering without such compensation are also discussed. There is more likelihood of critical pressure drop. This saves one valve and controller and the higher pressure drop means smaller valves. A suitable scheme for temperature and pressure compensation is gven in Chapter 11.10 Steam header configuration . Here each reboiler is supplied fiom a valve connected directly to the high-pressure header.

Sometimes enough steam leaks through to impair the accuracy of steam-consumption estimates based on steam flow metering. some reboilers. are now equipped with condensate seal pots.7 Steam S u . in effect.11 Improved steam supply and flow control system . Two practical problems are associated with the selection of condensate removal valves: (1) in sizing calculations allowance usually must be made for some flashing. 4 and Condenrate Removal 125 When header pressure variations are small. Steam condensate removal is most commonly achieved by traps-devices that. and (2) the pressure drop increases with flow rate since reboiler shell pressure increases with load. Many plants have experienced the need for high maintenance with these. particularly large ones. 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. temperature compensation is sometimes omitted since temperature changes more slowly than pressure. As a result.4. In addition. These pots then have conventional level controllers. are very simple level controllers. This means that the installed valve flow FIGURE 4.

intermediate buffer tanks to feed another process step. Avoid scheme 2. while a third pinches bottom product flow if column base composition has too high a concentration of low boilers. The characteristic time constant for this system is: where A B = column-base cross-sectional area. for example. f? i AH.12. . its required volume may be influenced by downstream requirements.12. as indicated by Figure 4. If flow smoothing is noncritical (as. = level transmitter span.126 Column-Base and Reboiler ArranpnmB characteristic should be between linear and square root. because of reboiler and columnl base “swell. One pinches the bottom product flow if a downstream level gets too high. or at least adequate. For maximum flow smoothing. however. if the bottomproduct flow is very small. A second pinches the feed flow if column base level becomes too high. As mentioned earlier. Base-Level Control via Bottom-Product Withdrawal In addition to the auto overrides of Figure 4. In the discussion that follows. Quantitative design is discussed in Chapter 16. good composition control is favored by small holdups. we will avoid the use of separate. we recommend a PI level controller cascaded to flow control. if at al possible. The PI level controller should have. control of the column. If possible. feet K* = level controller gain QmfB bottom-product flow-transmitter span. 4. Anticavitation trim may be required. fi3/min = For this system it is usually satisfactory to make TH 2 10 minutes. But if the column base serves as a feed vessel for another step. auto overrides to keep the liquid level between the level transmitter taps. It is sometimes required. or (3) feed flow. we commonly have three more overrides. it will be satisfactory to use a proportional-only level controller as shown in Figure 4.” mentioned earlier.8 REQUIRED HOLDUP FOR LEVEL CONTROL Base level is most commonly controlled by manipulating one of three variables: (1) bottom-product withdrawal. (2) steam or other heating medium flow. This is discussed in Chapter 5. the basic criterion for choosing column-base holdup will be good.13. if bottom-product flow goes to a large storage tank).

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

128 Coiumn-3me and Re&& Arrayemmk The dynamic response of the proportional-only.4a) de. of the column base or vaporizer separator AHT = level transmitter span. f ? . where TH A B is in minutes = cross-sectional area. x 12 x 42 o (4. averaging level control system may be defined by a first-order time constant: TH = A AHT B K.13 Proportional-only level control system for column base . of process fluid FIGURE 4. in feet.

8 Required Holdup jiw Level Control 129 K. 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. that is. Further. In addition.1) that a generally useful level transmitter span would be 38 inches of process fluid. = controller gain fi3/min = manipulated flow. Then if the valve positioner input span is 12 psi: A B 2 x 4 x AFS 38/12 = 7H (4-5) 2 . the discussion in Chapter 11. psig For a linear installed flow characteristic. maximum steam flow . For this design: where ( w ~ ) .= input span of steam flow transmitter. See. the level controller should be cascaded to a steam-flow controller with a linear flow transmitter (or orifice AP uansmitter. for a thermosyphon reboiler. @o)max ~ Q F S QFs = flow sheet or average flow. fi3/min There is some uncertainty in valve gain because of Merences in the valve- sizing philosophy followed by various instrument engineers. however. 5 3 Q ~ 7~ s This is the required column-base cross-sectional area. 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 . A mathematical analysis is given in Chapter 16.4. = signal to valve positioner. pounds per minute. one should make the volume AB X AHT at least ten times the volume inside the reboiler tubes. Base-Level Control via Steam-Flow Manipulation In this case rHshould be 20 minutes or more. and square-root extractor). we now lean to using a factor of 4 for design calculations early in a project. 6 . Earlier it was suggested (Section 4.

This may or may not be undesirable. As can be seen. This means that reboiler feed is richer in low boilers than is the column-base contents. which extends h o s t to the base. Minimum Holdup Design For those distillations involving thermally degradable materials. 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. care should be taken to ensure that the lower level nozzle is not below the midpoint of the tube length. 4. . (See Figure 4. 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. If a cooler is used.) If this arrangement is used.. the thermosyphon reboiler suction is taken fiom behind the outlet weir for the last downcomer.130 Column-Base and Reboiler Awa?.14.” Level nozzle spacing should be chosen as before for the same reasons.16. Note that for overall process control.’* and it is our understanding that there are various proprietary designs that have never been publicized. pcu/lb process fluid latent heat of condensation. When this valve is pinched. it should be noted that the surge tank may act as a condenser via the equalizing line.gemenB = = A .15 involves maintaining liquid level in a small-diameter pot or “peanut. pcu/lb = process fluid density. the bypass valve opens to send material to the column base. The preferred design of Figure 4.A typical control system arrangement for level control via steam-flow control is shown in Figure 4. If a thermosyphon reboiler is used. An occasionally encountered design for this purpose is illustrated in Figure 4.9 MISCELLANEOUS COLUMN-BASE DESIGNS A large number of column-base designs have appeared in the literature. the bottom product valve may be used with an override to maintain a low level in the column base. hP steam latent heat of condensation. It therefore is recommended that all such applications be subjected to a dynamic analysis such as presented in Chapter 16. Many of these include baffle arrangements intended to make the column base and reboiler function as an efficient stage of separation.17.

E W m E L C m U W 9 m J i 0 *I . d m a 3 $ G g E m s 2 d 8 w L 0 q w s 8 g - C 8 gg gz .9 Mhcehneous Column-Base Desgns 131 c .4.

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

16 Reboiler piping arrangement for preferential boiling of reflux from lowest downcomer .17.4. 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. 4. One way of connecting the two is to overflow from the column base into the bottom product receiver as shown in Figure 4. FlGURE 4. as a minimum. and (2) direct-fired reboilers. should have enough holdup to contain the entire column contents. an external surge tank must be used.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. Both are more common in the petroleum industry than in chemical plants.1 0 MISCELLANEOUS REBOlLER DESIGNS Two other types of reboilers are sometimes encountered: (1) reboilers with hot oil as a heating medium.

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 .

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

lC) is really the preferred overall design. In the following discussion. no pulsation damper is required. however. 5.1A. and (3) there is good mixing in the tank. then a pump is required. Hydraulic problems have been experienced in columns where the feed 137 .1 shows three commonly encountered feed flow schemes.2A or by adjustment of a variablespeed drive of a pump with a fixed stroke as shown by Figure 5.1B. When the upstream pressure is higher than the column 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. upstream or downstream pressures can vary signdicantly.2 FEED FLOW CONTROL Figure 5. as shown by Figure 5. feed composition. For minimum energy consumption operation. If. then a cascade level control/liquid flow control system such as that of Figure 5. it may serve to receive recycled column product streams.1C is required. then only a letdown valve is required. it should also send the feed to the proper feed tray. This (Figure 5. If the minimum stroke rate is at least three times the reciprocal of the feed-tray holdup time. And for startup/shutdown of the column being fed. it provides the most protection and offers the operator the maximum flexibility. we will assume that (1) process material-balance control is in the direction of flow. (2) feed-tank level control is of the averaging type.5 t Feed System Arrangements 5. Control may be by stroke adjustment on a constant-speed pump as shown by Figure 5. Some care should be taken in locating the feed control valve if flashing can occur. and sometimes feed enthalpy. For small feed rates.2B. If column pressure is greater than upstream pressure. as shown by Figure 5. The flow signal should be linear-one should use a linear flowmeter or a square-root extractor with an orifice and AI' transmitter. it may be considerably cheaper to use a positivedisplacement pump (piston type).

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

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

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

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

lbm/min FIGURE 5.142 Feed System Arrangements enthalpy cannot be measured directly. lbm/min feed rate. of bottom product entering economizer temperature.TB2) fps + WF I!~ . pcu/lbm "C temperature.5 Column feed enthalpy control with economizer and preheater . pcu/lbm feed specific heat. = TB. of bottom product leaving economizer feed iiil temperature.t f T1 WB + -cpb wF (TBl.J (5.P . degrees Celsius. degrees Kelvin ( = "C + 273) nta bottom-product flow rate.1) JrF enthalpy of feed entering column. it must be calculated as follows: JrF where = cpf = TB. degrees Celsius. = TI = wB = wF = .

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

144 Feed System Arrawemenk S. fi3/min.6 Column with multiple feed trays . Here P = average vessel holdup.6 FEED TANK SIZING There are a number of possible criteria for tank sizing. FIGURE 5. They include the following. fi3. F h v Smoothing. 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.

as a starting point. preferably with gain 2. surge tank volume between the level taps should correspond to a 711 2 1 hour. If it is decided that a separate surge tank is needed. = 2 rH x @ff). control of the succeeding process steps will be easier if the feed holdup is large. then. Compromises. pump. the use of intermediate tanks with supports. Generally speaking. If this cannot be achieved with the proposed vendor's hardware. 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.or bottom-product quality specification of the column to be fed is 0. 5.5. Top. A AHH.7 to 5. composition control of a column will be easier (faster) if the reflux drum and column base are kept small.9 illustrate several systems where two (or more) columns are operated in series or in parallel.7 shows a simple split column where two columns are employed instead of one tall column. On the other hand. If the engineer is confionted with one of the following situations. (very close to flooding) for a substantial fraction of on-stream time. one should consider using a dedicated small computer as a level controller. Column to be fed has a sidestream drawoff that is a major fraction of the feed. the recirculation rate should equal or exceed l o a s . but i general least 15-20 minutes. However. a separate surge tank should be used for column feed. if required 7 H or rM is much greater than 15-20 minutes.= 0. then a separate surge tank for column feed is recommended: 1.25 (see Chapter 16).. This scheme . If item 4 is a significant consideration. For a PI controller with auto overrides and Kch = 0. Columns in train have primitive controls (see Chapter 1). The reset time for unity damping ratio and 7H = 1hour would then be 4 hours. 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.25 7 H X @ff)m (vertical cyhdricd tank assumed). 5.7 FEED SYSTEMS FOR DOUBLE-COLUMN SYSTEMS Figures 5. The surge tank should also have good mixing. 2. therefore. Then A M . usually because of structural advantages. As a rough rule of thumb. 3. Figure 5.1% (mole or weight) impurities or less. electric switchgear. Otherwise one must go to a proportional-only level controller. are frequently necessary.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. 4 Column to be fed is expected to operate at or close to maximum capacity . and instruments may increase investment substantially. proportional-only control is best since output flow changes are never larger than input flow changes.

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 .

The lower codguration.5. The tank is connected in series with the main process flow. Note that this system uses about twice as much energy as a single column would require. Al the vapor is generated in the base of the second column. 2 normally increases energy consumption in process Unit No.9 illustrates feed systems commonly used for two (or more) columns operating in parallel. If the two columns are of unequal capacity. is used when there is some advantage in keeping the feed rate to one of the columns constant. The top of each column is maintained under vacuum. It is usually undesirable to put everything through the feed tank. 2 significantly. 5.8 shows a similar but distinctly different two-column system.10A is a simple technique for minimizing energy consumption. permits both feed rates to vary. 1 must be controlled.8 FEEDS WITH MAKEUP/PURGE TO TANKAGE Figure 5. at the same time.10 shows a system sometimes encountered when a lqge feed tank is used to balance load and demand at some point in the process. Cooling all the process stream and then feeding it to Process Unit No. Most of the flow from Process Unit No. Figure 5. Figure 5. depending on the optimum feed-tray location. Figure 5. Each column has its own reboiler and condenser. 2. This guarantees that a minimum amount of material is going to or coming from the tank. Vapor from the top of the second column is condensed and pumped into the base of the first column. 2. 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. . Liquid reflux for the second column is provided by the liquid from the base of the first column. it is desired to fx the flow rate into i Process Unit No. Feed can be introduced into either column. A bias signal is usually needed to balance the feed rates between the columns. The upper configuration. 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. The level in Process Unit No. since this typically operates at a lower temperature than Process Unit No. The control system shown in Figure 5. Feed can be fed to either column as dictated by composition. The vapor from the top of the second column is fed into the base of the first column. The other column then takes the swings in feed rate. 1 goes directly to Process Unit No. 1.9B.9A. 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.8 Feeds with Makeup/Puvge to Tankage 149 is also sometimes used when revamping an existing process to udlize existing l columns. Figure 5.

SINGLE SYSTEM Feed System Arrarpynwnts FIGURE 5.10 Makeup/purge feed systems .150 A.

9 FEED SYSTEMS IN SEQUENCES OF COLUMNS WITH AND WITHOUT RECYCLES Trains of distillation columns are often encountered. and E. the valve i the direct line connecting to two Units is pinched). 2. E. To prevent the level in Process Unit No. 53:963-969 (Dec. W..10A also requires overrides: 1. 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. 2 (if the n flow gets too high. Speicher. this problem is greatly deviated. To prevent too much material from entering Process Unit No. E. Cbem. and M. Luyben. D.Rt$mences 151 The system has some dynamic problems. L. W. REFERENCES 1. therefore. It has been shown by one of the authors. Eng. 2.. and M. should check all level controls in a sequence of process equipment. D. it may happen. that if one uses PI controllers enhanced with auto-overrides. Oglesby. Oglesby. 1961). The level loop is affected by the flow loop. Lupfer. Although this is uncommon. W. 3. If necessary. 2 from getting too low (a low level pinches the valve in the line to the feed tank).End. . J. in an unpublished study. 1962).10B. I d . IEC Fundamentals 10:147 (1971). This problem can be eliminated by the more sophisticated and costly system shown in Figure 5. Lupfer. The engineer. some of the tank sizes or controller parameters may have to be changed. A pressure controller on the "header" is used to balance supply and demand flow. Cant.. The system in Figure 5. 87-88 (Feb. 5. Material flows through these sequences of columns.

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

or if base level sets steam. Base composition may be controlled by trimming the steam/bottomproduct ratio. we have steam/bottom-product and distiUate/bottom-product ratio controls. level control must be as "tight" as possible for best composition control. For all ratio loops. Section 6. Our own studies f d to show any overwhelming advantages one way or the other. whether proportional only or PI. Bottom-Product Demand: Base-Level Control Via Feed Bottom-product demand requires. one should take care to account for reflux subcooling (see Chapter 16. in most cases. Since the bottom-product flow is the demand flow.2. 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. appropriate dynamic feedforward compensators should be provided. 6. Overhead composition may be controlled by trimming the disdlate/bottom-product ratio with perhaps a feedforward compensator connected into the overhead level control loop. . Then there will be two options for reflux drum level control. We can now visualize several possible designs. This section also gives a complete design.1.154 Level Control and Feedfmard 0pth. steam and reflux are ratioed to it. 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.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. Section 4). On the average. however. Section 2. Overhead composition may be controlled by trimming the reflux/bottomproduct ratio. In calculating reflux drum level controller settings. 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. Base composition may be controlled by trimming the steam/bottom-product ratio control. that base level be controlkd by feed rate. as shown in Figure 6. including overrides for bottom product and steam flows.u if reflux drum level control sets reflux. Reflux Drum Level Control Via Reflux In this case.

overhead level control via top product. base level via feed .1 Bottom product demand.6 2 Ma$eriul-Balana Control a Directwn Opposite to F h n 155 DEMAND FIGURE 6.

base level control via feed . overhead level control via reflux.2 Bottom product demand.156 LeueL Contml and Feedfmard 9th FIGURE 6.

Overhead composition may be controlled by trimming the reflu/distillate ratio. however. this column control scheme is not desirable. the calculation of base level controller settings follows that of Chapter 16. Overhead composition control may be accomplished by trimming the top-product/bottom-product ratio. Practically speaking. material-balance control in the direction of flow means that the column must take whatever feed is supplied. If steam flow is metered by an orifice. Section 6. either reflux drum or base level control must manipulate feed rate. it should be linearized with a square root extractor. in this case. Because of interactions between composition controls.3 Material-Balance Contra1 in Direction af Flow 157 Reflux Drum Level Control Via Boilup As shown in Figure 6. Top product composition may be controlled by trimming the steam/distillate ratio. Distillate (Top-Product) Demand: Base Level Control Via Feed Reflux Drum Level Control Via Reflux This case requires. base composition may be controlled by trimming the bottom-product/distillate ratio. Base composition may be controlled by trimming the reflux/bottom-product ratio. this scheme requires reflux/distillate and bottomproduct/distillate ratio controls. Determining settings for the reflux drum level controller is. Section 4.6. a buffer tank in the top product h e (notin the reflux line) is highly recommended.4. Since reflux drum holdups are usually small compared with base holdups. As indicated earlier. as shown by Figure 6. Base level controller settings may also be determined by the method of Chapter 16. 6 3 MATERIAL-BALANCE CONTROL IN DIRECTION OF FLOW . Reflux Drum Level Control Via Boilup As shown in Figure 6. The calculation of reflux dnun level controller settings follows the method of Chapter 16. although probably not impossible. Preferably one should make T H 2 5 minutes.3. level controller tuning will require a dynamic analysis of overall column material balance such as discussed in Chapter 14. this case requires reflux/bottom-product and distillate/bottom-product ratio controls. bottom composition may be controlled by trimming the bottom-product/distillate ratio. bottom-product/top-product and steam/top-product ratio controls. Note that if either distillate or bottom product (or side product) is a demand flow. A buffer tank in the top-product line is recommended. difficult unless a large reflux drum holdup is available. This is. Section 6. reflux drum level control via boilup is difficult unless a lot of holdup is available. subject .5.

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

4 Distillate demand. reflux drum level control via reflux. base level control via feed .6.3 Matffial-Balunce Conml i Direction o F h n f 159 FIGURE 6.

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

Top composition may be controlled by trimming the refldfeed ratio while base composition may be controlled by adjusting the bottom-product/feed ratio. If steam flow is measured with an orif?ce. In most cases base level control by boilup requires a dynamic analysis. Both level control systems may be calculated by the method of Chapter 16. 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. but base level control by boilup is very difficult. 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^.8) Overhead level control may be calculated simply by the method of Chapter 16.6) This is one of the most commonly encountered column-control schemes. reflux drum level is controlled by manipulating top product. care should be taken to include the subcooled reflux effect on overhead level control.^ If. bottom composition by trimming the steam/feed ratio. In many cases this scheme will require a buffer tank in the top-product line to the next piece of process equipment. Base composition may be controlled by trimming the steam/feed ratio. and perhaps supplementary plant tests. Base Level Control Via Bottom Product (Figure 6. can restrict feed flow. and base level is controlled by manipulating boilup (see Figure 6. Distillate/feed and steam/feed ratio controls should be provided. Reflux Drum Level Control Via Distillate.6. Base Level Control Via Boilup. Reflux Drum Level Control Via Distillate. however. Base Level Control Via Boilup (Figure 6. Reflux/feed and steam/feed ratio controls should be provided. Both level controls may be calculated by the method of Chapter 16. Section 7.7) This is also a commonly encountered control scheme.3 Matetid-BalanCe Control in Direction o Flow f 161 to the restriction that column overrides. It is normally used only when the average bottom-product flow is very s m d . a square root extractor should be used. Top composition may be controlled by tnmrmng the rdux/feed ratio. Reflux Drum Level Control Via Distillate. Top composition may be controlled by trimming the distiUate/feed ratio and with perhaps additional feedforward to the reflux drum level controller. particularly high base level. Reflux/feed and bottom-product/feed ratio controls should be provided. there will be a dynamics .9). unless the top product is going directly to storage. Section 3. Reflux Drum Level Control Via Reflux.

162 Level Control and Feedjimard Options FIGURE 6.6 Material balance control in direction of flow. reflux drum level control via distillate. base level control via bottom product .

6.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 reflux.7 Material balance control in direction offlow. base level control via bottom product .

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

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

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

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

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

2. 7. Let us assume that the feed consists of lumped low . This scheme has been used successfully on a nonequal molal overflow distillation. Top draw (distillate). this requires overrides or limiters for the side draw. empties into a small surge vessel. simple flow cone01 is sometimes used.fSidestream Drawof Columns Figure 7. a distillation column sometimes tends to collect intermediate boilers. instead of emptyrng into the tray below. Since the top and bottom products are small. Here the side draw is a liquid. the column-control scheme is very similar to that of a conventional two-product column with the addition of sidestream controls. and steam flow are shown ratioed to feed. For this column. compounds that are heavier than the light key but lighter than the heavy key. 7 3 SIDE-DRAW COLUMNS WITH SMALL SIDESTREAMS . base level is controlled by throttling side draw.and bottom-product flows (or their ratios to feed) are available for composition control. but the side draw is needed for column materialbalance control. this changes the rate of condensation.170 ControI . exact composition control of product streams requires as many streams as components. Base level is controlled via side draw while surgetank level control sets reflux back down the column. however. As mentioned in Section 1. which finally change reflux back down the column.9.3. composition of the sidestream is of primary importance. Here top. Let us consider how to provide composition control of the system in Figure 7. A practical problem is that of maintaining adequate reflux down the column. we can use an override on the side draw to ensure adequate reflux. column base holdup should be generously sized. if it is very small. By measuring the reflux flow being returned from the surge tank. but in practice. When the side draw is a vapor. As shown in Figure 7. As will be shown in Chapter 9. bottom draw (bottom product).1 shows a typical control scheme for a column with a vapor side draw. An alternative arrangement is shown in Figure 7. composition control will be poor. Changing the side draw changes vapor rate t the o top of the column. As an alternative one might control surge vessel level by side draw and base level by reflux from the surge vessel.4 COMPOSITION CONTROL OF SIDE-DRAW COLUMNS As indicated in Chapter 1. The downcomer. Ideally the side draw should be ratioed to the feed. In this case a small sidestream is required. otherwise the base level controller on occasion may cause so much side draw to be taken that reflux temporarily will be inadequate. and steam will be wasted. column capacity will be limited.1. If this is not done. The overhead level control loop that marlipdates reflux flow is nested within the base level control loop. and as many manipulated variables as components.

1 Basic control scheme for column with sidestream draw& .$&-Draw Columns 171 FIGURE 7.4 Composition Control of .7.

172 Control of Sidestream Drawof Columns FIGURE 7.2 Controls for liquid sidestream drawoff column .

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

increase the top-product/feed ratio. which are controlled as follows: 5 1. The top-product stream will consist mostly of A. See Figure 7. impulse feedforward compensation (see Chapter 12) for feed flow will do a better job than ratio controls. Figure 7. As shown in Figure 7. In view of the rising cost of energy. The fixed-gain relays are calibrated so that only two of the drawoff valves can be partidy opened simultaneously. See Figure 7.4. many columns can make good use of steam/feed ratio controls. It should be noted that many existing side-draw columns have neither composition controls nor ratio controls. increase the steam/feed ratio. See Figure 7. 3. Doukas and Luyben came up with the idea of changing the location of the sidestream drawoff tray. and if composition control is used.174 Contml o SirieStream Drawof Columns f boilers A.6A. This increases the reflux ratio.5 illustrates a method of minimizing the concentration of B in either the top product or bottom product.5 AN IMPROVED APPROACH TO COMPOSITION CONTROL OF SIDE-DRAW COLUMNS To ain one more degree of freedom for composition control.4. If the signal from either analyzer becomes too high.5. and the side product mostly of B with some A and C. steam flow is increased. 2. The following control strategy therefore is suggested: 1. 7. In some cases the purge flows are so small that the loss of product in them is deemed negligible. Thus as the signal &om the XS1 composition controller increases. product B. 2. 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 there is too much C in the side product. If there is too muchA in the side product. however. The concentration of the intermediate component in the top product (distillate) is controlled by manipulating the reflux ratio. this easily can be implemented by conventional analog hardware. If turndown of more than 2: 1 is anticipated. .6B. As shown in Figure 7. thereby decreasing the concentration of intermediate boiling component B at both ends of the column. the bottom-product stream mostly of C. and lumped high boilers C. If there is too much side product B in either the top or bottom product streams. there are four specifications for the three product streams (three components assumed). The concentration of the lightest component in the sidestream product is controlled by the location of the sidestream drawoff tray. increase the bottom-product/feed ratio. the valves for trays higher in the column open as valves for trays lower in the column close.

4 Scheme for control of sidestream composition .7.5 A n Impwed Approach to Composition Control o St&-Draw Columns f 175 FIGURE 7.

7 and 7. together with part of the FIGURE 7.5 Control of terminal composition .fSidestream Drawof Columns 3. controlled by heat input to the reboiler.176 Control . 4 The concentration of the intermediate boiler in the bottom product is .8. In both cases the functions of the prefiactionator were (1)to remove essentially all of the low boilers out of the top.6 PREFRACTIONATOR PLUS SIDESTREAM DRAWOFF COLUMN Doukas* studied two schemes involving a prefractionator as shown in Figures 7. The concentration of the heaviest component in the sidestream product is controlled by the sidestream drawoff rate. 7.

2s % c a)ni % E 25 a) P ) a 8 a)':%. E ws2gg m CSP5+.cl_o + m a n o.Em + . u-.g cu -2:s a l 323 = 00 % E: -3 9) i3a) I U 3 m C 3.E "3.5 cc a e+. a) *c mm'E: = a ) = f g.0 c 3 L 0 yp 00:sg m m + .7.s 5oyc iG-02 c g zc g % .O%g SSU** .6 Pr@&*w Plus Sidestream Drawof Column 177 x -EZ asp' 9 ) . $!!ES a ma) a) 2a)Z g m U EB& %=* 2 -09) E 0.

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

6 Pr@-actim~or Plus Sidestream Drawof Column 179 FIGURE 7.-scheme .8 I.7.

180 Control . Sidestream drawoff location is used to control the concentration of the lightest component in the sidestream product. N. Thus the heaviest and lightest components are detoured around the section of the column where the sidestream is withdrawn. “controlsfor Sidestream Drawoff Columns. ~ REFERENCES 1.” ISA J. N.8 provides manipulation of the distillate product flow from the prefractionator to control sidestream composition. Luyben. The intermediate product is then withdrawn fiom a tray or trays located between the two feed trays.” Ph.. together with the remainder of the intermediate boiler.” INE C H . The “D” scheme of Figure 7. L. L... Thesis. W. “Control of Sidestream Columns Separating Ternary Mixtures. 65(5):45trol Distillation Columns with Sidestream Drawoff. Doukas. and “pinches” are avoided.. “Composition Control of DistiJlation Columns with Sidestream Drawoffs Separating Ternary Mixtures. 7. S. 37- 51 (May 1969). The two product streams fiom the prefractionator are fed to two different trays of the sidestream drawoff column. Since the D scheme is probably easier to implement. Luyben.fShstream Drawof Columns intermediate boiler. D. Figure 7. 4.7 has been designated by Doukas and Luyben as the “L” scheme. Both of these schemes were shown to provide effective control for modest changes in feed composition.7 OTHER SCHEMES It is impractical to present here all possible schemes for controlling sidestream drawoff columns. and (2) to remove essentially all of the high boiler out of the bottom.” CEP. “10 Schemes to Con- 42 (July 1966). and W. 4 3 4 8 (June 1978). 3. it is recommended for most systems. Lehigh University. 2. 1978.. . I?. D o h . Buckley. A number of these are discussed briefly in a paper by L ~ y b e n .

a 70-tray column separating a mixture with a relative volatility of 1. Geyer and Kline' give. as an example.7 percent. Feed provided at the proper feed tray. conservation means designing and operating a column so that it makes the specified separation with the least amount of energy per pound of feed.8 t Minimizing Energy Requirements 8.6 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. Energy recovq-recovering and reusing the heat in the column product streams. Automatic control of composition of product streams. whether they be liquid or vapor.2 CONSERVATION For distillation. Consequently we will discuss conservation first.4 and with specifications of 98 percent low boilers overhead and 99. an increase of 8 percent in energy consumption results. 8. It can be shown that this results in a lower energy requirement per pound of feed than would feedmg on any 181 . The main emphasis of this chapter is on the latter approach. 2. Operators commonly overreflux conventional columns with a single top product and a single bottom product. but it is pointless to uy to recover energy unless we also try to conserve it. We have a number of techniques to accomplish this: 1. Extra heat is used to ensure the meeting or exceeding of specified product purities. Conseraatian-designing and operating a column so that it makes the specified separation with the least amount of energy per pound of feed.1 INTRODUCTION here are two basic approaches to minimizing &stillation column energy requirements: 1. 2.

whether they are liquid or vapor. . pressure steam is available that otherwise would be vented to the atmosphere. changes in production rate. The use of steam condensate receivers instead of traps reduces maintenance and steam losses. the necessary separations can be accomplished with lower boilup/feed and reflux/feed ratios. Dry distillation. A number of schemes have appeared in the literature. is usually small. Possible use of mechanical vacuum pumps. can often benefit from new. changes in feed composition.Requirements other tray. and the column may flood at lower boilup rates than it would when operating at higher pressures. recovering or reusing heat contained in the column product streams. This steam is usually cheaper than high-pressure steam.182 Minimtzzw E M . 4 Use of lowest pressure steam available. 8. it is sometimes economical to switch to steam-heated reboilers. one may use a partially flooded reboiler (see Chapters 4 and 15) and throttle condensate. 3. practically spealung. Condenser capacity may be limited. Therefore. 6. Energy recovery in a distillation column means. The two chief ones involve (1)"multiple effect" distillation. analogous to multiple effect evaporation. In many plants steam traps require considerablemaintenance and have sigmficant leakage.2Lower pressure usually means higher relative voIa&ty. Column operation at minimu pressure. The difference. Insulation. Where reboiler AT might be too small if the steam were throttled. regardless of the scheme. But. Use of steam condensate receivers. however. 8 3 DESIGN CONSIDERATIONS IN HEAT-RECOVERY SCHEMES . As feed composition or enthalpy deviates &om design values. increased insulation. Reserve capacities that may be required: -Extra heating capacity -Extra cooling capacity -Extra distillation capacity These are important for startups and shutdowns. and (2) vapor recompression.' In many plants excess low. Older columns. however. pressure and temperature compensation of flow measurements is highly desirable if steam is throttled instead of condensate. and changes in product specifications. 5. if * "Auxiliary"condensers and reboilers are those installed in parallel with "normal" condensers and reboilers for startup or peak load purposes. 7. For vacuum columns there is some opinion' that mechanical vacuum pumps offer energy savings over steam jets. designed before the energy crunch. auxiliary" reboilers and condensers should be avoided. the o p t i m u feed-tray location also changes. however. there are five design factors that must be considered: 1. Generally speaking. For columns now using live steam. Since low-pressure steam is seldom available at constant pressure or steam quality.

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

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 .

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 .8.

Another problem associated with this scheme is the selection and sizing of the feed valve to the first column. 8.4. The A P control. and very close control of supply and load pressures may be required. control may be both simpler and more flexible. we have found it advantageous to let pressure find its own level in the second column. If there is a serious discrepancy. that is. An interesting practical problem here is how to adjust the auxiliary reboiler on the first column. it finds its own pressure.4.3 has an interesting dynamics problem. the base composition of the load column is controlled by boilup in the supply column. Vapor flow to the first column from the condenser-reboiler will not be constant. This column will run at a low pressure at low feed rates and at a higher pressure at high feed rates. and low at high flow. for the cases studied. In this particular case. may have its set point adjusted by a composition controller for the lower section of the first column. The scheme of Figure 8. maximum column pressure occurs at maximum feed rate and boilup rate. one can see that valve pressure drop will be very high a t low flow. the first column gets only part of its heat from the second column. base level in the second column may experience serious upsets. The column that is the source does not need to be operated at a constant pressure-in the scheme shown. After examining some complex heat-balance schemes. It should be noted that for the schemes of both Figure 8.5 SINGLE SOURCE. In another version a following column in the train supplies heat to a preceding column as shown in Figure 8. Interactions between the two columns may be severe. we decided that the simplest approach was to use column AI'.3). Assuming that the feed comes from a centrifugal pump.186 Minimiziv Energy Requiremoats point fiom a primary controller. in turn. For columns we have studied to date. Again. particularly if the secondcolumn bottom-product flow is smd. The variation in valve pressure drop with flow will be much greater than that normally encountered in a pumped system. the one supplying heat. SINGLE LOAD When there is only one source and one load (see Figure 8. For the illustrative example. but the AI' control will provide a rapid method of ensuring aonstant boilup. there has been no problem with flooding at lower rates and pressures. the overhead composition of the supply column is controlled via reflux. In the absence of vapor flow control.3 and 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. the remainder comes from an auxiliary reboiler. . interactions may be severe.

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

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 .

but rather at an intermediate site.6 SPLIT FEED COLUMNS A third arrangement. runs at a higher pressure than the load column. 8. 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. water-cooled condenser may be necessary for startup (see Figure 8. 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 .An~auxiliary. is apt to be variable.7 COMBINED SENSIBLE AND LATENT HEAT RECOVERY In addition to the recovery of the latent heat of vapor streams. 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.~ feed between two columns that make the same separation (Figure 8.8.12 . Other papers on energy integration for distillation columns include those by OBrien” and by Rathore et al. The supply column. The feed enthalpy or temperature.~ Mosler7 discusses the control of a number of vapor recompression schemes.6). N d 9 presents investment equations and data. The incentive in many instances was to be able to use water-cooled condensers. The feed split is controlled to maintain a heat balance. therefore. 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. thus avoiding the expense of refrigeration. Since feed flow is typically set by level controllers or flow-ratio controllers.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. A review of compression equipment and methods of estimating operating costs has been presented by Beesley and Rhine~mith. steam-heated reboiler and/or auxiliary ~. Such schemes have been used in the chemical and petroleum industries for years.5). however. Another factor favoring vapor recompression is a small temperature difference between the top and bottom of the column.) 8. involves splitting ~.8 Energy R e m 7 iy Vapor Recompresswn 189 8. which is used in some s y ~ t e m s . its flow rate will not be constant. (See Chapters 5 and 11.

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

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

D s Dev. M. Bmley. Rhinamith. L. Mostafa. Powers. N. Shaner. “Heat . “Distillation with Optimum Vapor 4. Kline... A. (1983). 1976). CEP.12.”AICbE J. F. R.. Geyer. and H. 7. D. 1980). and W. 37-41.. Mosler. e. N. (July 1976). and W. 11. 6. 1974. Chiang. 950 (Sept. Recompression. H. Tampa. L. B. R. CEP. (Aug. Res. R. G. 1977. 5. 391-392 (Nov. Dtitillation ControL. and P.. 47-52 (May Des. and R.” CEP.. “Reducing Column Steam Consumption.. New York.. T. J.192 Minimking Energy Requirements REFERENCES Integrated Distillation Configura1.. Proc.” Cbem. Vanwormer. A. 1983). 9. CEP. lation. 1978). “Heat Pumps in DistilMcGraw-Hill. H. Eng. CEP. R.44-49 (July 1980). E. 10.. E. Luyben. 2. Fl. Fahmi.. 65-67 CEP. K. O’Brien.” CEP. Luyben. 8. Rush. G. 3. G.. 20(5): 490control. Rathore. AIChE Distillation Systems with Energy Workshop on Industrial Process Integration. F. F.. “Synthesis of tion Towers. D. 58-64 (July 1976).” paper submitted to IOEC 49-51 (May 1976). Null. tions. A.. Shinskey.. L. 59-66 (Sept. “Coneol of Sidestream and Energy Conservation Distdla. H. S. A. Tyreus.. and G.” Proceedangs. 1974).

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

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

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. Multiple-input low-signal selectors are now available from vendors. an alarm or interlock is activated.it also may be assembled with two devices or purchased as a combined device. 5 ) .1. but it is a technique that is worth keeping in mind. but a particular one that is inexpensive and works well involves the use of combinations of devices such as described in the following.3 Implementatirm . The outputs of the lo “override” controllers are a s connected to these selectors.3. As many input signals as desired may be accommodated by arranging selectors in series. or down to a predetermined minimum value (LL) as shown by Figure 9. The major application has been for auditing multiple flow measurements to a chemical reactor. One may also construct a median selector with two high selectors and two low selectors as shown in Figure 9. Their accuracy is poor. If pneumatic versions with adjustable gain are required. Summers Summers are devices that add or subtract. or to hold it a safe distance from a constraint. High Selectors (HS) or Low Selectors (LS) These devices select either the higher or the lower of two input signals. .2. one should avoid those with pressure-dividing networks. a low limiter is a signal source plus a high selector (Figure 9 . In practice one or more selectors are inserted between the output of a “normal” controller and its final control element.f Overrides 195 9.3 IMPLEMENTATION OF OVERRIDES Many ways of implementing overrides are possible.4). As process constraints are approached. Correspondingly. usually a valve. Functionally. It thus can be constructed with two separate devices or assembled into one housing.9. In either case the cutoff or limiting value is readily adjustable. a high limiter is a low selector plus a signal source such as a supply regulator (Figure 9. So far we have found no applications to distillation. If either the high or low measurement deviates too far from the median value. 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.


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 .

or in the case of vacuum or pressurized columns. the output of A1 is 3 psig. however. Condenser-Cooling-Water Overrides As mentioned in Chapter 3. As the condensate receiver level transmi&r output increases above 3 psig.26 where a PI level controller was used with auto overrides. if the reflux drum loses its level. This seldom works well.9.9.7 Overrides@ Column Overhead System 207 Low Condensate Receiver Level Override on Reflux As liquid level rises to the lowest position (3 psig). S1 starts the pump. it may also be closed by high liquid level in the next process step. to have overrides act directly on the valve as in Figure 9. and overall is probably preferable.2. This permits accurate calculation and prediction of override behavior. Without such protection moderate changes in reflux flow could either flood the receiver or run it dry. Similarly. 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. This design should be compared with that of Figure 3. this override starts to close the reflux valve. it may be the pressure controller. the output of reverse-acting override A2 with gain 4 starts to decrease from 15 psig and reaches 3 psig at 100 percent level. the pump shuts down. when the level reaches 25 percent (6 psig). This permits the reflux valve to open and the reflux controller to take over somewhere in the lower 25 percent of level. and high and low level overrides acted on the reflux flow controller set point. For some columns it is desirable to maintain a minimum reflux flow rate. If the level drops again. override A10 holds the cooling-water valve open as long as the condensate receiver level is low. This would require an additional override on the reflux valve. 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. High Condensate Receiver Level Override on Reflux If liquid level rises above 75 percent. Low-Low Condensate Receiver Level Override on Cooling Water To guarantee condenser cooling-water flow during startup. . These two overrides are particularly helpful for a column with a high reflux-to-distillateratio. This drives the reflux valve open.5 psig. some plants like to control condensate temperature by throttling cooling water as shown in Figure 3. Reflux pump operation is made automatic by switch S1. the output of reverseacting override A1 with gain 4 starts to decrease from 15 psig. It has been a more common practice in the past. We are using this scheme increasingly. The normal controller for this valve may be the cooling-water exit-temperature controller.

vertical-tube reboiler. causing the condensate temperature controller to cut back on cooling water. it is usually desirable to limit exit cooling-water temperature to a maximum of 50-60°C.10 will accomplish this.8 OVERRIDES FOR COLUMN-BASE SYSTEM The column-base system shown by Figure 9.208 Applicatiun o Protective Contmh to Dktilhttim Columns f if at all. particularly if 316 SS is used. so it is increasingly common to provide an override from coolingwater exit temperature as shown by Figure 9. the cooling-water exit temperature rises.10 High cooling water exit temperature override on condensate temperature control . 9. The override scheme of Figure 9. To minimize fouling and corrosion.10. The base liquid level transmitter is installed and calibrated FIGURE 9. If the condensate temperature becomes too low.11 features a thermosyphon.

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

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

It should also be noted that the bottom-product valve may be closed by an override from the next process step. Miscellaneous Column-Base Overrides The base system contains two more overrides. For this we provide an override (not shown) that is a minimum steam flow controller. A5 has an output of 3 psig. an automatic startup may be accomplished with an air switch connected to a three-way. But it is desirable to maintain enough boilup that the trays (sieve or valve) do not dump or weep. which turns the pump on whenever the base level transmitter increases above 6 psig. as soon as steam and feed are available. The bottom-product pump is operated automatically by switch S2.9 Aut& Startup and Shutdorpn 211 Minimum Steam Flow Controller As liquid level rises from 75 to 100 percent. (Some column designers prefer to provide enough space below the first tray that it can never be flooded. The column may be started up again simply by turning on the steam.) Such a shutdown may be made to accommodate production scheduling or may have been forced by process interruptions elsewhere. depending on whether the column base is full (“wet”) or empty (“dry”)when the column is shut down. A6 closes the steam valve if base temperature becomes too high. air-operated valve in the air line between the steam valve and its overrides. As will be seen. Wet Column Startup If a column is shut down by closing the steam and drawoff valves. As shown in Figure 9. this would also cut off steam.6.9 AUTOMATIC STARTUP AND SHUTDOWN Two kinds of startup and shutdown are commonly encountered. wet column startups and shutdowns are much faster. a column may or may not follow the same sequence of constraints from startup to startup.9. the entire column contents accumulate in the Iower section of the column. Typically the liquid level rises well above the vapor inlet from the reboiler and up over some of the lower trays. It should be noted that no formal program or sequence control is necessary. In most cases we have found it advisable to adjust override biases so that a high level pinches feed before steam. Depending on circumstances. During startup A4 holds the tails valve closed until base composition reaches a desired value. The “on” position corresponds to a through . Instead. The same basic override system permits automatic startups regardless of whether they are wet or dry. override A5 closes the feed valve. the column comes on line at the maximum speed permitted by the constraints. 9. At a level transmitter output of 15 psig. With s t e d f e e d ratio control.

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

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

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

Overrides may be provided such that if either valve goes beyond this ni value. xD. Entrainment Override With the previously discussed system. and AP. that the peak of the xD versus wR curve does not occur at fixed. A possible hookup is shown on Figure 9. . This would replace the AP override. shows a maximum value of xD at a particular reflux rate as shown by Figure 9. W R . it may be necessary to protect the utility supply during startups.9. it is not desirable for valves to go more than 95 percent open or less than 10 percent open. 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.15. the limitation on any one column is the maximum permissible column AI’ that was taken to be a known. The high-gain relay should have hysteresis or detent to prevent chatter. A particular arrangement is shown in Figure 9. for good control. For highly integrated plants. fixed value. This is not an override in the normal sense since no signal selectors are needed. This usually occurs at a column AP considerably higher than the nominal maximum permissible AP. it may be necessary during wet startups to avoid putting such a demand on the steam header as to reduce header pressure seriously. the feed valve is pinched slowly u t l the utility control valve’s position is back to a reasonable value. Implementation is shown in Figure 9. By measuring XD and WK and computing h D / d W R . one may use either the cooling-water valve position or steam-valve position for override purposes.13. reproducible d u e s Of XD.11 M ~ ~ U u n e o w Overrides 215 minimum value. Overrides for Maximum Capacity Where it is desired to be able to run a column or train at maximum capacity. xD. This requires a slow-acting override incorporating a propomonal-reset or floating controller. except on a momentary basis. A. as one increases feed rate slowly and before flooding starts. the high-gain relay can be set to open the bypass valve as the normal valve approaches the closed position. however. but must be connected to pump suction. In “bootstrapyy plants.16. certain additional overrides may be needed.versus reflux flow. Limited Utility Override If the condenser or reboiler limits before the column proper does. For some columns. a cooler should be installed in it to prevent overheating. It should be noted. for example. a plot of top composition.14. an inmeme in reflux flow will cause a decrease in overhead purity. Generally. however. at some high feed rate. If a flow measurement is not available. One automatic approach is to connect header pressure to overrides on the reboiler steam valves to close them partially as header pressure drops. wR. 8. If the bypass line cannot be connected back to a tank or receiver as indicated. To put it another way.This is so because the increase in entrainment o&ts the effect of increased reflux. a high-gain relay opens the normally closed bypass valve.

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

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. as shown in Figure 9.9. The summer bias. and if its loading signal is less than 4.8 psig. thereby permitting maximum column capacity. one of them will limit before the other does as feed rate increases.12 Destgn Considevatwns 217 C. maximum column-base pressure.18. Since any feedback control system must have some room within which to work. is very helpful to think of this zone as a fraction of the measured variable transmitter . The column and auxiliaries. Note that a high limiter holds signal A constant for cooling-water-valve loading signals greater than 4. 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 overrides must be so designed that the process does not normally approach the hard constraints too closely. the summer or “balance” controller calls for a slightly higher column pressure to increase condenser heat transfer. Automatic Balancing of Condenser and Reboiler Heat Loads If the condenser and reboiler do not have balanced heat-transfer capability. we can visualize. as. One way of balancing these two heat exchangers on pressurized or vacuum columns. provided the column itself does not limit first. 9. and the required range in pressure must not be so large as to cause adverse changes in relative volatility.2 psig. * It span. The manipulated variable always reaches its maximum (or minimum) value before the process variable exceeds its hard constraint. The soft constraint will correspond to the minimum (or maximum) value of the override output. With a proportional control loop designed for dead-beat response. for example. a zone between hard and soft constraints. C.12 DESIGN CONSIDERATIONS It is probably apparent that most overrides are really feedback control loops. for example. In the case of overrides with proportional-only action. The width* of this zone is determined by the override loop gain.2 psig and a low limiter holds B constant for steam-valve loading signals less than 13. then becomes the pressure controller normal set point. of course. They therefore are subject to stability considerations. If. which is limited by stability considerations. the condenser cooling-water valve opens wider than the steam valve. must have adequate static pressure ratings. there is a unique relationship between the value of the manipulated variable and the distance between the process variable and its hard constraint. In many cases they are also subject to truly hard constraints.17.

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

9.de~ationS 219 FIGURE 9.17 Control scheme for balancing condenser and reboiler heat loads .12 Des@ Consz.

Most of the feed is taken off as sidestream.21).20.18 Hard and soft constraints . Usually a gain 6 override is adequate. A proportional-reset controller is usually necessary here. Increasingly we are putting the limiters in the set-point channel instead of as shown. only one controller is necessary if the two transmitters are scaled to have the same gain. Following are the most important override^.^ 1. Low base temperature or low top temperature holds all drawoff valves closed (Figure 9.5 and reset time is usually 20-40 seconds. the column of Figure 9.1 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. 3. High column AI' or high base pressure closes the steam valve. but this really depends on column composition dynamics.19. see Figure 9. as an example. and a small amount of high boilers must be removed as bottom product. M overrides for the steam valve are shown in Figure 9.22. On-stream analyzers would permit a better job here. A high selector chooses between them. The function here is to hold the column on total reflux until overhead and base compositions are nearly correct. 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). FIGURE 9. A small amount of low boilers must be removed overhead. We will choose. This is a purification column in a solvent-recovery system. Since column dynamics are about the same for either variable.220 Appluation of Protectzve Controls to Distillation Columns 9. 2. Typical gain is 0.

19 Flow rate controls for composition control .13 Overrules fm Si&-Draw Columm 22 1 FIGURE 9.9.

this is accomplished by calculating the total vapor flow from the reboiler from the stuam flow. FIGURE 9. 5. The output from this controller goes to a low selector in the path between the base level controller and the sidestream drawoff valve. calculated vapor flow up the column becomes less than required.222 Appliuztaba of Pmtectave Conmh to DirtiuatiOn Columnr 4. the override controller closes the sidestream drawoff valve just enough to force the required additional vapor up the column. then. this is accomplished by subtracting the sidestream drawoff flow fiom the estimated internal reflux to calculate the net liquid flow down the column. 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. the override controller pinches back on the drawoff valve until the downflow becomes adequate.20 Feed flow system . For columns with a sidestream uapw drawoff. it is necessary to maintain a minimum vapor flow up the column d o p e the drawoff point. If. it is necessary to maintain a minimum liquid flow down the column b e h the drawoff point. and subtracting the sidestream vapor flow to get net vapor flow up the column. As shown by Figure 9. For columns with a sidestream liquid drawoff. As shown by Figure 9.24.23. If this flow is insufKcient.

9.21 L w temperature overrides for drawoff valves o .13 OPerriaRFfbr SA-Draw columns 223 FIGURE 9.

22 Steam valve overrides .224 Application of Protectzve Controls to Dkilhtwn Columns FIGURE 9.

13 Overrides fm Side-Draw Columns 225 FIGURE 9.9.23 Override for minimum vapor fo up column lw .

226 ApplitatMn Df Protective Conh.24 Override for minimum liquid f o down column lw .olr to DirtiUation Colamns FIGURE 9.

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

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

Today it is sometimes deliberately allowed to float to minim& energy consumption (see Chapter 8).230 Indirect Cumposition Measwements organized summary of the various ideas and criteria generated over the years.to high-purity columns. if the pressure is fixed. This can be very confusing to the operator . temperature probes should be installed in the active part of a tray. An early approach to compensating for pressure variations was to use two temperature measurements. but the control tray temperature will increase only very slightly because it is essentially pure high boiler already. 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.3 DIFFERENTIAL TEMPERATURE For a binary distillation. 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. Theoretically. 10. 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. For good speed of response. But pressure is often not very constant. If. However. For example. the temperature changes at the ends of the column are quite small in moderate. boilup does change significantly. Therefore. small changes in pressure or the presence of other lighter or heavier components can affect temperature much more than composition of the key component. However. One was then subtracted from the other. a single temperature measurement will provide a reliable p d e to composition. however. this works fairly well as long as boilup does not change much. The composite effect is that A T has a minimum at a particular boilup. Another consideration is nonlinearity. 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. Even if one fixes pressure at one end of a column. The relationship between tray temperature and the manipulated variable can be quite nonlinear if a tray near the end of the column is used. not in downcomers. An increase in heat input will drive more light components up the column. a decrease in heat input will drop light component down in the column and the control tray temperature can change very drastically. one usually near one end of the column and the other at an intermediate tray. 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. and if the two temperature measurements are separated by a substantial number of trays. it will vary at the other end as a function of boilup rate. 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. This nonlinear response presents difficultcontroller tuning problems when conventional linear controllers are used.

4 Dfetta Vapm Pressure zmnril 231 and may cause instability in the AT control loop. 3.4 DIFFERENTIAL VAPOR PRESSURE Another early effort to compensate for pressure variations in binary distillations resulted in the development of the differential vapor pressure cell. Efforts inevitably were made to stretch the AT technique to fit multicomponent distillations.” As shown in Figure 10. These two conditions will give similar 4T signals but vastly different product compositions. Bulb filling is a job for the factory. Some users. . It is connected to one side of a differentialpressure transmitter.1. Vermilion5presents a somewhat different application of dfferential temperature control. however. Some of the d16culties encountered and some guidelines for successll application are presented by web be^. For a 56-tray deisobutanizer. whch causes erroneous readings and sometimes control problems.6 10. There are. some practical problems: 1. The other side of the AI’ transmitter is connected directly to the same tray. pressure in the bulb will be the same as pressure on the tray.^ In another vintage paper. thus leading to measurement errors. Where applicable. and may either flood the column or shut it down. Liquid of the desired composition for 6 h g the bulb may not be chemically stable over a long period of time. The AT signal will also be small if the profile is so low in the column that it is filled with low boilers. This is easily visualized if one considers two extreme situations. this instrument is capable of great sensitivity. a bulb filled with liquid whose composition is the same as that desired on a particular tray is installed on that tray. have installed snubbers in the pressure impulse line. splitter column was reported by Bonilla. reusing an instrument originally bought for another job may not be easy. Here the intent apparently was not to compensate for pressure variations. Response to pressure changes is much faster than the response to composition changes. the AT signal will be small. but to control the temperature profile in the middle trays of the column. therefore. temperature was measured on trays 22 and 42 (measuring from top down). When the liquid on the tray has the same composition as the liquid in the bulb. This nonmonotonic relationship between AT and product composition occurs even when boilup changes do not affect Al’. Use of the AT permitted successful control of isobutane in the bottom product. 2. For that condition the AP transmitter is normally set to read midscale.10. and thereby control terminal composition. Successful application to a CJC. If the profile is so high in the column that the section between the two temperature points is filled with high boiler.’ The version made by Foxboro is known as the “DVP Cell. Deviations in the tray composition from that in the bulb are then reflected by AP transmitter output signals above or below midscale.

1) where PR = total pressure.1 DVP cell schematic . it can be used on some multicomponent systems. If the reference bulb has been properly filled with a liquid of known composition. To determine the required span of the DVP C l in inches of water corel responding to a desired composition span. atm. Tivy7 has shown that with a bit of patience and cleverness.XR) PHYHR (10. of low boiler at T R FIGURE 10. in bulb xR = mol fraction low boiler in bulb PL = vapor pressure. atm abs. then at a reference temperature T R pressure in the bulb is: PR = XRpLyLR + (1 .232 Indirect Composition Measur-B Although this instrument is primarily intended for binary systems. it is necessary to write and solve a few equations.

the process pressure is: PI.10.2) PT = total pressure in process. one might well ask.xR) PT = ~ L ~ L Y L (1 .~ = where L ~ L Y L + (1 . ‘Why do we have it?”The answer is that there is only one instrument to calibrate rather than two.3). atm xL = mol fraction low boiler in process PL = vapor pressure.3b) Since PR “washes out?’ of equation (10.PR and is normally calibrated in inches of water for such spans as 20-0-20 or 50-0-50. atm. a m .X L ) + PHYH (10.3) . of high boiler at TR yL = low boiler activity coefficient yH = high boiler activity coefficient Now the DVP Cell actually measures PT .5) . We can now solve for xL: (10.2): PT . The activity coefficients may be found from: (10. of low boiler at TR PH = vapor pressure.2).PR = ~ which reduces to: L ~ L Y L(1 .X L ) PHYH (10. and the need for very accurately calibrated temperature and pressure measurements is eliminated.3a) which is the same as equation (10. From equations (10. of high boiler at TR ym = low boiler activity coefficient in bulb ym = high boiler activity coefficient in bulb At the same temperature.X L ) + PHYH pHYHR] (10.4) (10.4 Dtfferential Vapor Pressure 233 PH = vapor pressure. a m . TR.[xRpLYLR + ( l .1) and (10.

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

.10. 1c e 9) 9) + E 0 0 3 r p 8 P.5 Presure-Compmated Temperature 23 5 c . c E E 3 u l m H E 2 c . 5 z1 sk 2.

. we want no change in Opc ifcomposition has not changed. AO. = 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..KpcAO. 8. . - P - = = transmitter input span transmitter output span (AO.. where = constant composition (10. we can see that if there is a pressure change AP from PFs that causes a temperature change AT. Kmp= pressure transmitter gain )ma P- - P- P .13) T.= K. . = temperature transmitter gain (10. In other words: AOpc = AO.. + Ob (10. .).8) where 6 = compensator output signal .AP K...9) or Kpc= '8. = pressure transmitter output signal .10) (10.11) (10.T- = = temperature transmitter input span transmitter output span (whether pneumatic or electronic is immaterial) (AO.12) AO.Kpc 0. = 0 (10..).236 Indirect CmposittiOn Measuremenn The compensator has the equation: epc = Om.. AT K. . Theory for Binary Distillation With the above equation in mind..

- ar/aT KP AP m cornposiuon constam -- K~~ ar/aT 1 “i 1 FIGURE 10. t and p .3 Composition vs.

3 and plotted C versus T (both read at constant PFs)on a new plot. we get: - (T + -1 273.or more generally when: .. 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. We therefore have taken Figure 10. The output of the compensator. . Some operators have found it confusing to have both temperature and pressure-compensated temperature. 3 .8 . For electronics an RC filter may be used. The operator's control station can then be calibrated directly in terms of composition.8 = 0 [see equation ( l O . . it is possible to take a considerably more accurate approach.ICpc. pi = e[Az+BJ(T+273. 1 6 ) ] Differentiating. In some cases the transmitted pressure signal is so noisy that a snubber may be required. 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. It is common practice to adjust the compensator bias so that it reads midscale when T = TFsand P = PFs./aT Some Practical Considerations 1 .16)] so n pT = i= 1 yixi c k + B i / ( T + 2 7 3 .e is a measure of the pressure-compensated temperature. If a computer is available. l ) ] Therefore: 2.238 Indirect Cornposh M e a s u r m a fiaction.

one may find Pi from equation (10.The activity coefficients may be found by equations (10. and (lO. 10. B j .18) The individual vapor pressures may be found by the Antoine equations.21).19) (10. In looking at equations (10.7).Y3P3 +zP2 . high boiler Equation (10. We may rewrite equation (10.5). 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 MULTICOMPONENT COMPOSITIONS COMPUTED FROM TEMPERATURE AND PRESSURE MEASUREMENTS If an on-line digital computer is available.22) y3P3 Piand y i .18). taken at the same tray in the column.16 It is desired to find x1 fiom measurements of PT and T. it is possible to use a more sophisticated computing scheme than otherwise would be practical. we can see that xL is really an implicit function.6 Multimnpiment Comparitimc f.S). (10. however. The activity coefficients .22).21) T + 273. one may find Knowing Ai. m Temperature and Pressure Measurements io 239 For a binary system. To treat this subject in a very general way is not feasible.4).20) P = -&yipi i= 1 (10. see equations ( 10.4) and (10.17) yL = activity coefficient of low boiler yH = activity coefficient.17) may be solved for xL: (10.10.(1 + +) (10.R 3 (10. This is easily done on any computer.= exp A. the basic equation is: PT = ~ where L ~ L Y L + (1 -~ L ) ~ H Y H (10. Then knowing x1 from equation (10. + ( = Y2 YlP1 p . By knowing T we can readily calculate PL and PH.6) and (10.20): x1 P. To show what is possible. but fiom there on we must find xL by trial and error.

23) Plots of A2T versus X. and are primarily usell for binary or almost-binary systems. splitter. Luyben’s method partially fixed the profile in either end of the column. . and also was nonmonotonic. Boyd‘s technique partially fixed the composition-versus-tray profile in the vicinity of the feed tray and on both sides of it. Both approaches effectively cancel out the influence of boilup rate and pressure on temperature measurements. Luyben’ made a computer study of a 25-tray deisobutanizer with temperature measurements on trays 5.02. He located two temperature measurements in the rectifying section of the column and two more in the stripping section. Therefore. but they did a show a maximum at about xB = 0. He then defined: A2T = ATR . 10. Since most deisobutanizers operate in a range of xB = 0. and 15 (numbering up from the bottom). (mol fraction E.0. In each case A2T was used to control top product. More generally. 10.03 . trial-and-error solution would be required.) showed that it was affected very little by disturbances.TlO) (10. Boyd’ took a somewhat different approach. Bonilla6 reported that Boyd’s double-differential-temperature technique did not work well for a CJC. The signal was found to be dependent on feed composition. this might well be useful for control. He defined: A2T = (TlO - 7-15) - (7-5 .10. The computed and measured data checked well. The thought has occurred to a number of people that perhaps additional temperature measurements could be used to calculate composition in multicomponent systems.240 Indirect Composition Measurements (yls) are sometimes fairly constant (if‘ the range of composition changes is small).24) ATR = temperature dfference in rectification section AT. each system was calibrated by gas chromatographs.ATs where (10. they both may be regarded as improvements on simple AT measurements. the y’s would be functions of composition. so an iterative. Two papers that illustrate two Werent approaches have appeared in the literature.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. = temperature difference in stripping section Plant tests on two columns showed great sensitivity.

As the profile moves up or down. In addition." This technique is useful for distillation columns in which a large temperature change occurs over a few trays (a sharp temperature profile).10. This technique has been successfully applied to many industrial columns. and is linear. ti average temperature changes gradually. The multiple-temperature-control scheme uses a number (four or five) of temperature sensors located around the normal temperature break. 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). Shah14 had good success with the Brosilow estimator as long as the column was operated w t i a linear region. For example. 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. when nonlinear ihn . An average temperature signal is computed from these multiple temperatures (this can be inexpensively implemented in conventional pneumatic or electronic analog hardware). However. distillate composition is estimated using an equation of the form: xQ = ajlTl + aj. This is found in columns where the components have widely differing boiling points (peanut butter and hydrogen would be an extreme example).T2 + aj3T3+ a j z s + aj5L. The Brosilow estimator employs a linear combination of selected tray temperatures and steam and reflux flow rates to estimate product compositions. They used the term "inferential control" for this type of composition estimation and control using secondary process measurements.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. handles multicomponent systems.25) where Fs = steam flow rate Lo = external reflux flow rate The a coefficients are constants that are determined experimentally or by calculations. 10.9 COMPOSITION ESTIMATORS Brosilow and co-worker~"-'~studied the use of several temperature and flow-rate measurements to estimate product composition. The estimator uses steady-state relationships. hs The average temperature signal is fed into a conventional feedback controller.9 Composition Estimm 241 10.

242 Indirect compmitimr Measwemm phenomena occur. 1975.. and W. 11. U. B. .. L. 5.. Griffin.. proposing a nonlinear composition estimator for binary systems.... distillate. and H.S.. Joseph. 14. W. Mosler. and steam flow rates and one or more tray temperatures). involving tray-to-tray material and energy balances. Webber. Luyben.627 (1973) and 3. Oil and G sJ. Brosilow.. W.. . IEC Fund. AIChE J. O.. Smith... A fundamental nonlinear mathematical madel of the column is used. 15. Bondla. 1955). 1976). that rdux and distrllate flow rates and a tray temperature in the rectlfjrlngsection are known. and measured variables (feed. Hyd. Rijnsdorp. 7.6/1(1979). the difference between two temperatures. 24: 485(1978). Sre. New York.855. 1948). 12. and the distillate-to-feed ratio to adjust manipulated variables. 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. Suppose.S. L. Shah. R. M. Weber. Parsons. (Nov. et al. eis Chem. 13. known process parameters (tray efficiencies and heat losses). 2. 6.” ISA Trans.. They proposed several control schemes using the sum of two or more temperatures. R. V. Rt$38(5): 187191 (May 1959). W. and A. R. 56. Symp. E.. a new guess of X.: et al. 2. Luyben. 3. D. no. I m . and D. the linear estimator performs poorly. Grote. 21. and C . Pet. En. h 127 (Mar. L. “Automatic Control of Fractionating ~ h m n s .. DynamM and contrd OfContinuOw Di&llathn Units” Elsevier.. If ni the calculated temperature on this tray does not agree with the measured temperature. 18: 238 (1972). 13(1) 8. U. K. REFERENCES 1. “The Use of Process Analyzers for Composition Control of Fractionators. W. CEP 71(6):55-60 (June 1975). 8: 739- (1979). 10. A. J. E. M. J.733. 744 (Nov.. 9. 18: 614 (19721. Hyahfux Pnx. Maarkveld. vapor-liquid equilibrium and physical property data. 4. 1961).. 29. L. 98a 100 (Aug. Joseph. 1976). Boyd. is made and the procedure is repeated. D. Shah14mended Bmilow‘s work.351 (Nov. B. P t ”e . Tivy. O. A. The work of Weber and Mo~ler’~ should also be noted.and one wants to calculate distillate composition x. Patents 3. B. E. W. that is. AIChE 1. poor estimates of product compositions are made. AIChE J. Shah verified the effectiveness of the nonlinear estimator by both simulation studies and experimental tests on a pilot-scale column. J. for example. . H. 24-0 (Nov. Patent 2. 1969). V e d o n ..725.074 (1974).. Webber. Rademaker. Luyben.

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

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

1 Measurements needed for internal reflux computation FIGURE 11.2 Pneumatic hardware configuration for internal reflux computation .2 Calculution o Dirtillation-ColumnInternal R$w f 245 FIGURE 11.11.

so equation (11. When the two temperatures are equal-no subcooling-Ksc will equal 1. Let us a s say that we want the output span of the summer to lo be 12 psig for 0 to [KscImax.T is maximum.3)[(1 12 - S)12 + S(C . . A . we get: (11. We now need to find the f and S factors. then: B = [KxImk x 12 + 3 = x12+3 [KSC lmax [KscIm When To .5) (11. (To .TR is some specified maximum value.C is maximum and R (11.I-. we want p = 15 psig.00 = [KSc]-. proportional to the internal reflux. Let us hrther assume that the output of the multiplier.4) = Since 0 ..Next we must have a procedure for determining B and KR.6) AlS0: (11.7) and solving for KR.TR)maxKmr Thus by combining equations (11. has the same span as that of the external reflux flow transmitter (assumed linear).C>m = ( T O .corresponds to 12 psi.6) and ( 11.).7) (A . and B corresponds to [Ksc]1.2) becomes: p=B (11.8) Multiplier The Foxboro 556-8 multiplier has the equation: A -3 = I ( B .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 assume that when To .3)] (11.[KsC]. A will equal C.

11) ( 11.00f Ksc = [KsclmaX Note that C .1 a) 1 6.]. we want: A . = x 0.11. l l a ) may now be solved to findf (11.00 psi If at the same time the temperature difference is maximum.13) C = 0: Equations (11.13) to find S: 6.S) 6.3 = 6.00 S + 1 = [KscImax [Kscl. x) a 2 and if the temperature difference is zero: A -3 = 6.15) (11.00 [ ( l .00 [Ksc]12 = 6.16) ( 11.17) (1 ._ x 6.S) x 12 + l2 S ] CKscImax (11.10) .12) and ( 11.14) This value off may be substituted into equations ( 11.3 = 6.5 (1 .00 = [K.12) -_ Note that forA - (1 .2 Ca&uLa&n o DljtiUatMn-ColumnIntwnal Refw f 247 When the external reflux is: (WR)max 2 B .S) x 12 12 = + 12S] (1 1.la)] (11.00 (11.10) and ( 1 l .S whence S = l +s (11.18) .00 x [Ksclmax [fromequation (ll.00 [KscImax 6.3 = 12 since: Next if external reflux is: ( W R.S) x 12 + 1’ 2 [Ksclrnax ] (11.

33psig =c Then: A .302 150 = 0.TR). S = 1.000 12 = .71 235 15 4 1. and fiom equation (11.33) 12 = 6.f As a check let: WR = 1.14).22 + 3.18).22 150 [fiomequation (11.000 pph To .302 = 9.22 = 13.5): B = .022 pph Therefore: WRl = -x 50.22 psig From equation ( 1 1Ab) : 12 x KR = 1- 0.248 MirceUaneow Measurements and Controls Exampple. Let us consider a case where the following conditions apply: External reflux flow-meter span = 0-50.= 100°C A = 235 pcu/lbm From these: [Ksc]max 1 + .00 1.00 = 12.00 (10.348 From equation (11. = 25. (To .x 100 = Oo71 235 = 1.302 = f From equation ( 1 1.2)l 1.l2 + 3.1136 + 12.348 x 40 l2 + 12.1'302 x 6.000 pph Temperature transmitter spans = 0-150°C tP = 0.725 28.71 pcu/lbm "C.3 = .TR = 40°C p = = 0.302.

/pll = ratio of gas density to that of air at 14.C. T = temperature. "C. inches W. 60°C orifice flow constant flow.21) .7 psia. and downstream if pc is taken downstream.1): wR1 = 25. "C For a particular value of Ab.3 Temperature and Pressure CmpensattiOn o Gas F h Meten f 249 Check via equation ( 11. psig. used in designing orifice pressure transmitter span. It should be measured upstream if orifice reference pressure pc is taken upstream.3 TEMPERATURE AND PRESSURE COMPENSATlON OF GAS FLOW METERS The basic gas-flow metering equation is: (11. "C. p = = orifice pressure drop. = temperature transmitter span.11. = orifice reference pressure. used in designing orifice orifice reference temperature. of flowing stream where p is measured p.021 pph 11.20) = k2 dAbo (11. psi = ps = T.000(1 = + 235 x 40) 28. pressure. and pc: (11.19) where Ab. measured at T. psia. scfm or pph SC = gas supercompressibility kl q Let us also define: = = pc T.

24) = + 273 T. equation (11.12 x (11. "C (Tm)- - (~m)Inirl = ATS temperature transmitter span.3. A more recent Foxboro bulletin..20) (1 1. and T # T. We have been unable to find in them any real advantage over the arrangement of Figure 11. features two alternate arrangements.27) where (Tm)m= = maximum measured temperature. + 273 (11.22).26) (11.23) provides a convenient way of determining true flow whenp # p .23) (T+ 273)(p.. so the following will apply to Figure 11.21) into equation (11. "C (Tm)-= minimum measured temperature. we obtain: = k2 Jm(11.3) 122 + S(C . Temperature Compensation The usual divider form of the 556 is: (A . we can divide equation (11.25) ( 11.3.For the same Aho but p to get: # pc and T # T.3 (see Foxboro TI45-3a of October 14.3 = . "C = .3) f and Foxboro defines for this application: (Tm)max B .23) with pneumatic instruments is given in Figure 11.7) As can be seen.22) Substituting equation (11. A common method of mechanizing equation (11. TI 45-3a of November 1969. 1966).+ 14.

3 Temperrnre and Pressure Cornpenration o G s Flow Metm f a 251 FIGURE 11.11.3 'Qpical compensated gas flow metering scheme .

25). transmitter signal. . we get: B .. e.29) 1 2 3.. and (11.33) where p .8 = = signal from Abotransmitter signal from temperature transmitter It can be seen that the denominator of equation (11. therefore. takes care of span suppression and converts both T and T..8 . This calculation. [(Tm)where + 2731 + ATs .3) (11. psig minimum value of transmitter span.29) is equal to T + 273.3) f 12 X [122 +S(C .3 = (T.31) (11.q-. + 273)(8.252 MhceUaneaa Measurements and Cont~oi3 A is the port receiving the Ah.8 On substituting equations (11.3)] (11. ( ) (Pm)= = maximum value of transmitter span.X (B .32) (11. receiving the temperature transmitter signal. Some engineers omit temperature compensation of steam flow on the grounds that it changes more slowly than pressure. (11.24). The wisdom of this is questionable.26). + 273) x 12 (Tm)max + 273 - (T. to absolute units as required. Pressure Compensation The multiplier form of the 556 is: A -3 = .30) and for this application Foxboro defines: (11. psig .27) into equation (11. and C is the port .

30): A . Substituting equation (11.3 Tempmature and Pressure CunzpmattiOn o Gas F h Meters f (pm)max 253 - @m)min = = 4 5 pressure transmitter span.37) This procedure. + 14. and (11. + 14. Substituting for B .36) 1 signal from pressure transmitter Clearing. enters. + where . we obtain: But: so: p A - = (p.. This sometimes requires juggling transmitter spans.71 + 14-7 (Pm)max + 14-7 ] (11. 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.29).7(8. (11.1I .7)(B . we get: A - 3 = (e.3) (T. (11. to absolute Units.7 + 14.3 from equation (11.3.7) (11. In speclfjrlng Foxboro 556-8's and 556-93.+. + 14. therefore.3 is now a measure of q2.34) + (pm)maxAps 14. 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. and C is the port where 8 the pressure transmitter signal . psi B is the port where the signal from the divider enters.27) into equation (11.3 = (Pm)max 12 [Ip.38) Since A . 8 = - 3) 1 (11. + 273)(p + 14. takes care of span suppression and converts both p andp.31). we .32). one must be careful to stay within available ranges f o r f a n d span S .7) (T + 273)(p.

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 = -5.2 percent error 1500 . for example.7) -1 (11.7) * = k .7) x 4 (11. the following: Steam flow = 1500 pph Design temperature = 397°F Design pressure = 240 psia Ahois assumed to be constant.may take the partial derivative of equation ( 11.aT - -AT a9 (11.39) Aho@ + 14.45) 100 Percent error = 2@ + 14. (Aq) = .41) 2(T + 273) 2(T + 273) -1 Percent error = Next: -loo x AT ( 11.19) with respect to temperature and pressure.42) (11.43) - 2@ + 14.40) (11.' x (PJlPU ISc aT - (-i) 1 x (T + 273)3/2 (11.46) Consider. J .

the heat exchanger of Figure 11. for example.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. reflux rate. We wish to know the increase or decrease of enthalpy FlGURE 11. or boilup rate.4 Ht.4 HEAT-FLOW COMPUTATIONS As mentioned i Chapters 3 and 4.4.4 Heat flow computer for heat transfer . Let us consider.3 percent error 1500 11.11. it is sometimes desirable to calculate n heat flow.

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

It is the subject of this section.2 ft lb mass/sec2 lb force liquid phase density. Regardless of the type of transmitter used-AP or displacer-the basic equation relating liquid level.51) APH = head.5): APH where =B -(pL L 5 c - p. feet pL = pp = vapor phase density. lb force/fi?.11. We will be concerned only with problems associated with determining level by means of head measurements. feet nozzle-to-nozzle spacing. Heat must be transferred. In the discussion that follows. we will assume that process design is adequate.5 Column-Base Lepel Measurement 257 most often encountered. and head is (see Figure 11.5 Head-level relationship in a v e s s e l . Ib mass/ft3 liquid-vapor interface elevation above bottom nozzle.)H + 84 pvA H N B C (11. density. ft/sec2 flL = gc = mass-force conversion factor = 32. and liquid and vapor must be properly separated. Ib mass/ft3 H AHN = = QL AP" = B ( p L ~ J H AH^ - 9c 9c ' I FIGURE 11. between two nozzles local gravity acceleration. liquid must be partially vaporized. These difficulties are at least partially process oriented in the sense that a column base with associated reboiler is often very complex.

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

6 Column base-reboiler manometer . in this case.55) L = length of piping from base to reboiler. It has the natural frequency of the “manometer” between the column base and the reboiler (Figure 11.75 cps. feet FIGURE 11.5 Column-Base Lmel Measurement 259 Noise amplitude is usually greater at low frequencies. Roughly this is: -fL = .6).11. Column-Base-Reboiler “Manometer” Another source of noise is sometimes encountered in columns with thermosyphon reboilers.\/Lcps where 1. Therefore. any filtering or damping that we design should provide sigdicant attenuation for frequencies as low as 0.3 (11.

260 Mhceh~w Mearurements and Controh As an example. let us consider an internal damping chamber for use with either AI' or displacer-type transmitters. in many cases "fiddling" with such valves has seriously delayed column startup or has caused premature shutdowns. Displacer-type Transmitter Resonance A third source of noise-ofien the most serious-is attributable to the basic design of most displacer-type transmitters. There are now several force-balance. The basic mathematics are given elsewhere (pages 187-191 of reference 2). The displacer and torque tube (Figure 11. mass-spring system that is usually highly underdamped. The use of adjustable hand valves is highly undesirable.3 = 0. or by snugly sized piping. which A I R SUPPLY FIGURE 11.7 Schematic diagram of displacement-type level transmitter .3 CPS The amplitude of this cycle may be arbitrarily attenuated by a properly sized restricting orifice in the line. for 20 feet of piping: f = M z 1. 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. 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.7) constitute an open-loop. displacer-type transmitters on the market. The incomplete information available indicates that these have a higher natural frequency and more damping than do the nonforce balance type.

11. this chamber has two key features: (1) a hood or cap to keep out liquid from the last downcomer or the reboiler return nozzle. a pot with 12 in3 and a Taylor snubber (58S104) will give 1 O : l attenuation at about 2 cps.5-0. It may also be used with gas or steam flow-meter installations with self-draining impulse lines.5 Column-Base Level Measurement 261 functions as a first-order lag. and (2) a l-inch-diameter hole in line with the lower level nozzle to facilitate rod out.6 cps. FIGURE 11.8. Useful for level measurements. Many combinations of volume pots and restrictors are possible.8 Internal damping chamber . Here a restriction and volume pot are installed in each impulse line to ensure that the inputs to the transmitter are dynamically equalized. 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.0-1. If there is a concern about solids collecting in the chamber. 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.9. this design is almost mandatory for specific gravity measurements via AP for slurries. As shown by Figure 11. The damping will then be 1 O : l at 1. The combination of orifice and chamber gives about 1 O : l damping (attenuation) at about 0. as an example.2 cps. a second l-inch hole may be cut in the lower section of the chamber. one may provide damping as shown in Figure 11.

need additional damping. particularly in column bases. A more sophisticated approach is to determine experimentally the noise-frequency spectrum and to design notch filters (i. band rejection filters) for the dominant frequencies. FIGURE 11. 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. and thus minimizes pilot saturation and output errors. The damping affects both input signals and the feedback. This instrument has the desirable feature of nonadjustable damping and so is “fiddleproof. first-order.262 Miccelhneow Measuremena and Controls The simple.9 External damping for AP level measurement .e. Its dynamics approximate those of a first-order lag with T = 0.7 second. In fact. if too large a time constant is chosen. filter approach may not always be adequate.” Experience to date (10-15 years and several hundred applications) has been very favorable on flow-control applications. it may interfere with control.. Apparently many level applications.

^ 2. fixed damping provided where required. One could deliberately use the upper connecting line as a condenser. It is pertinent at this point to mention certain potential problems with displacer-type transmitters: 1. The chamber and upper line should be insulated and heated. the natural fiequency of the “manometer” may be low enough to interfere with level control. and overflow the remainder back to the column. if it is desired to have a liquid purge and damping.5 Column-Base Level Measurwnt 263 Many AI’ transmitters on the market have adjustable dampmg in the feedback. This may take place fast enough to cause a sipficantly higher level in the chamber than in the column base. in the range 1-3 cps.11. provided the time constants are selected with some care. This leaves adjustment up to the instrument mechanic. If excessively long connecting lines are provided between the vessel and boot. As discussed later. who seldom has a rational basis for setting it. This feature therefore is not recommended. 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. the guard valves should be left wide open in normal operation and need be closed only for maintenance work. This is discussed in a very interesting paper by sander^. Downflow By far the most common case. the variable damping feature should be left wide open and external properly designed.10. But an excessive amount of damping can have an adverse effect on control. Level Measurement Errors Due to Velocity Effect As shown in Figure 11. send a fixed amount of condensate via a metering orifice or capillary to the measurement chamber. the purge rate must be fixed and the liquid level elevation in the chamber known. Where such instruments are bought on a project. Damping for Externally Mounted Displacer Transmitters As mentioned earlier. 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. level is sometimes measured with the nozzles located on vessels or pipelines of different diameters. If the liquid velocity V in . For extremely noisy installations. most &placer-type transmitters have a sharp resonance “ot”. 3. in difficult cases a small gas purge to the upper line is helpll. This provides a fiddleproof design. both internal and external damping may be used. changing the damping while the process is in operation may have a serious effect on control-loop stability. 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). Two cases are of interest: downflow and upflow. vapor condensations may occur in the upper line.

at times. then the error he in inches of process liquid is: he = -0.9 fi/sec.10 Velocity error in head measurement . FIGURE 11. upflow In this case there is no velocity error. a given column may have a different bottom specific gravity as a result of changes in bottom-product specifications. = .1 inch. 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. In other words.264 Mkcellaneous Measuremma and Controk the bottom-product h e is in feet per second. Also. the transmitter output signal wiU indicate a level that is 1 inch lower than the true level. h.28 Vz When V = 1.

a variable density froth exists above the tube bundle. consider the following: AHT = level transmitter span. note that equation (11. but the one shown in Figure 11.52) can be rearranged: H= APH (BLIBC 1PL ( 11. For an understanding of this scheme.56) In accordance with this equation.8 calibrated = 5 = 1.11 Specific gravity compensation o head measurement of liquid level f . the level transmitter signal is divided by the specific gravity transmitter signal to give a corrected level signal. As an example. if the column has an internal reboiler. feet of process fluid = 4.3 There are many possible ways of compensating for variable specific gravity.05 = value at which level transmitter is FIGURE 11.11was developed for a specific application.5 Column-Base Level Measurement 265 Further. a = = U& 1.0 feet maximum specific gravity .3 0.11.

Low-velocity laminar flow is usually adequate for sealing. may contribute to dynamic measurement errors. purges are intended to accomplish either or both of two functions: 1. Now let us look at some sources of level measurement error due to purge systems.24 f =-z S = 1. This permits f l ul transmitter output if the process specific gravity is somewhat less than design. 4-10 mA. S. Purge System Errors When a AT' transmitter is used for measuring base level.9 . This situation usually arises because of thermal degradation. etc. etc.3 Z = .). For these applications gas purges are usually used. and often both impulse lines. we obtain: 1. but to inventory (pounds. Now for level control purposes we want a transmitter whose output signal is proportional not to liquid height. Flushing or sweeping. it is common to locate it above the upper level nozzle and to purge at least the lower impulse line.615 0. Gas purges ofien are not adequate. But for liquid or gas. To accomplish this some vendors can furnish displacers with an eltiptical top-to-bottom shape. cubic feet.3 .8 1.385 . 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. . Alternatively the output of a standard transmitter may be compensated by a microprocessor or computer. 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. gallons. for example.3 1.3 wheref.If. a Foxboro 556 relay is used as a divider. turbulent flow through the exit diptube is recommended.0.3 1. it is desirable to calibrate the level transmitter so that a full-scale output (3-15 psig.) is obtained for sipficantly less than full nozzle spacing. Generally speaking. and Z are terms used by Foxboro for adjustable relay parameters._ 1. 2. but as we shall see later. which is simply the isolation of the transmitter f?om a clean (no solids.0.= 0. In the absence of specific gravity compensation. which may be required where solids in the process liquid tend to plug the impulse line.5 = 0. See also recommendations in Chapter 4. Sealing. This usually requires a vastly higher purge flow rate than is common practice. nonplugging) process liquid.

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

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. This design has a number of sources of error. and may even load the vacuum jets. usually 70°F and 14.13 Insufficient purge. will ride up and down with column-basepressure. Installation Errors Typical gas-flow purges are arranged as shown in Figure 11. and most common calibration procedures do not allow for it. the rotameter does not indicate true flow since it reads correctly at only one temperature and pressure. Upstream gas pressure is high enough that critical drop exists across the needle valve. The purge is then connected to the transmitter impulse line close to the transmitter.14. This is rarely measured or calculated. Second. level transmitter erroneous response to rapid rise in pressure . FIGURE 11. 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. The rotameter pressure. First.268 Midlaneow Measurements and Controls may be excessive pressure drop in the impulse lines. 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. however.7 psia. which is followed by a rotameter.

Here the fixed capillary restrictor is used in conjunction with a purge-flow regulator. the purge joins the impulse line close to the process rather than close to the transmitter. known flow for a particular application.5 Column-Base Level Measurement 269 Third. particularly for specific gravity applications. is shown in Figure 11. Consider. for example.15. the restrictor usually chosen is a needle valve whose annular clearance changes with ambient temperature. the required C. a needle valve with an orifice 1/8 inch in diameter. this system does not require critical flow.0002 inch. For 1. An even better system.16. supply at 50 psig. Further.004. capillary restrictor that has been designed to provide a specific.11. and critical flow. is about 0. A far better gas-flow purge system is shown in Figure 11. Here the rotameter is upstream of a fixed. For most applications the horizontal design is adequate and is cheaper. This is probably a good point at which to comment on horizontal versus angled nozzles.14 Typical gas flow purge system . The plug-seat clearance is only about 0. particularly when plug and seat are made of different materials. In addition to being more reproducible.0-scfh air flow. The angled design is believed by some engineers to offer advantages FIGURE 11.

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

if any damping is used. Liquid purges are also sometimes used with displacer-typetransmitters. In addition. it reduces the available level transmitter span.5 Column-Base Level Measurement 271 when solids are present. The level in the housing was then 9. As an example.17 Angled nozzle with dip tube . constant flow. the calibration should allow for the extra head due to the purge flow. either an external orifice or an internal chamber. When used with displacer-type instruments. particularly if a bottom connection to the measurement chamber is used. FIGURE 11. We chose 3.0 inches higher than it would have been with no purge. a purge-flow regulator is mandatory. 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. The orifice was calculated to have a diameter of 0.8. Liquid Purges Liquid purges are sometimes used with AP transmitters for difKdt applications where there is otherwise a tendency to plug the impulse lines.17. Such installations should be designed to avoid any pockets in which solids can accumulate. This is usually omitted with horizontal nozzles. With angled nozzles it is common practice to use a diptube as shown in Figure 11. For accurate. and the calibration procedure should allow for impulse-line pressure drop.0 gpm of a liquid with a specific gravity of 0.55 inch.11.

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

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

2. l ~ General Approach A general approach to choosing body size. Some of the newer designs. 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. developed by an FIGURE 11. and valve type has rm been developed that is based primarily on equations for C. or are noisier in gas service. temperature. There are today many new types of valves.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.(flow coefficients). There are many disk and other rotary valves with much larger C. There are newer methods of calculating the normal or operating C. ti size.20 Level measurement with flush diaphragm AP transmmer and 1:l repeater .. and greatly reduced weight. however. as compared with globe valves.14 In t y n to make these decisions. lower price. pressure. have a greater tendency to cavitate in liquid service.&' 3.. that are much more accurate than the old FCI (Fluid Controls Institute) equations. and fire safety.

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

In other cases we have installed computing relays in the control loops to generate desired curves. We have calculator and computer programs to facilitate these calculations. But whether one valve is used or two. such as flow. calculated for flowsheet maximum flow is: 1. . the mathematics is simple and the results have been tabulated.. If this cannot be accomplished with one valve. No greater than 0. No greater than 0. For some commonly encountered systems. and so on-and see what inherent flow characteristic comes closest to providing the desired installed characteristic.276 Miccellaneow Measurements and Controls The best inherent flow characteristic for a given application cannot really be chosen until the desired installed flow characteristic. A suggested procedure is to ensure that the operating C. is known. for globe valves with one-size-reduced trim. Maximum Flow and Turndown A control valve must be able to pass a specified maximum flow and should preferably.7 [C. This is 91 percent lift for a 50: 1 equal-percentage valve." But the older FCI procedure offers no flags to indicate when the new procedure is required.5 [C. 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). The criterion we prefer is that of choosing the installed characteristic that minimizes or eliminates the need for retuning the controller as throughput changes. or nearly so (splitrange arrangement). The slope of this curve is the valve gain that must be known for quantitative design of control loops.]. Next one must look at a particular system-pump curve. for control purposes one must be able to throttle more than flowsheet maximum flow and less than flowsheet minimum flow. At this point it is appropriate to note that the increased complexity of the new C. and pressure-control systems. for valves with full-size trim for which reduced trim is not available. Here 6. But how much more? There is no simple answer to these questions. flow-ratio. How does one determine maximum required flow?Under transient conditions a valve must be able to pass more than flowsheet maximum.]. 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. In some cases we can help matters with a cam-operatedpositioner. liquid-level. These curves may also be generated by the "calculated variable" function of computers and of some distributed control systems. And soon dedicated microprocessors probably will be available. Q or w versus 6. calculation procedure has led some engineers to refuse to use it except "when needed. but not necessarily. 2. This is 82 percent lift for a 50: 1 equal-percentage valve or 65-70' open for a typical 90' butterfly valve. is the signal to the valve positioner. pass the minimum required flow without getting too close to the closed position. piping and equipment pressure drops.

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. 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. If piping and equipment pressure drops are large in comparison with valve-pressure drops.25 and (&) 5 0. note that in a chain of process equipment. (aax)(a. 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. This is so because the valve has run out of pressure drop. valve AI’ is almost never a factor in quality of control. .and downstream pressures. low-holdup chemical reactors constitute the only applications we have found to date that require high valve pressure drops. it is desirable that valve lift be no more than 95 percent for and no less than 10 percent for (&). two or more split-ranged valves may be required.)Fs (aax) Turndown and Inherent Flow Characteristics as Related to Process Applications A. At the high-flow. however. A few high-speed. It has been shown mathematicallv that higher valve AP’s do provide better control. it is suggested that for valve-sizing purposes we let P 1. and around distillation columns. Startups and shutdowns. small valve AP’s mean less tendency to flashing or cavitation. for a given application. In view of the difficulties of accurately predicting up.2 but this effect is so small for most applications that it may be neglected. System with Pumped Liquid For pumped systems rapidly increasing energy costs have aroused considerable interest in designing for very low valve-pressure drops. high-lift end of the plot. The installed flow characteristic also becomes verv Merent fiom the lnherent flow characteristic. This pushes us toward line-sized valves. A suggested typical process turndown figure for valve sizing and selection is therefore five to one. For flow to.75 (&)Fs. from. 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. calculated valve-flow turndown is less than process turndown. often require very low flows for awhile. valves should be so sized that maximum manipulated flows will be greater than maximum disturbance flows. As an alternative to the above. On the positive side. all curves. regardless of valve inherent characteristic. Finally. however.) at 95 percent lift to that at 10 percent lift.11. If. it is probably desirable to provide at least 5 psid for the valve at flowsheet maximum flow rate. Valve-flow turndown is defined here as the ratio of flow (not C. level off at a maximum flow. valve turndown suffers (becomes smaller). A linear valve tends toward an installed square-root flow characteristic.6 Control Valves 277 Determining minimum required flow is even more difficult. however. But regardless of what flow criterion is used. the tak will overflow if inlet flow goes over 100 gpm. for example.

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

however. .11.4. The high-side impulse line is usually connected to the vapor space just under the bottom tray. the line may have slugs of liquid that can cause a partial vacuum to develop. which are usually near ground level. and usually enters the column at an angle to facilitate draining. 1 I COLUMN AP MEASUREMENT . to facilitate maintenance.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). To predict the dynamic behavior quantitatively in order to design a temperature control system quantitatively. It should be insulated and.7 Column AP is most commonly measured by installing a AI’ transmitter above the top of the column with self-draining impulse lines. Where this is not done. and probably should not be used for AP’s much under 50-75 inches of water. The instrument or control engineer is interested in the study of the dynamic behavior of thermowell/primary-elementcombinations for these reasons: 1. however. Opinions vary as to the optimum high-side impulse-line size. Vapor line AP. dual. except perhaps for new facilities being built at a new site. (See discussion of purges in Section 11. Particularly if it is not sufficiently vertical. I . it is located above the condenser or condensers. resistance thermometer. may introduce considerable error. If a thermocouple is used.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. which provides rigidity and mechanical protection. the impulse line may act as a condenser. The primary element is usually a thermocouple. negative AP. to purge the line. in some cases. Sometimes. This technique is not highly accurate.) It is also possible to use repeaters as discussed for level measurement in Section 11. heated with steam or electricity to minimize condensation. splitranged butterfly valves are recommended. We prefer to compromise at 1 1/2 inches and. or gas-filledexpansion bulb. Some engineers prefer at least a 2-inch-diameter line. it is most apt to be in the form of a “pencil” or sheathed assembly. if possible. In view of this problem and the need to avoid tight shutoff in the winter. while others will settle for 1 inch.4 on base-level measurement. The AP transmitter may then indicate a sizable. This high-side impulse line is ofien the source of errors.

21.and 7 6 . For the base case of 152 fi/sec.4 8.2 8. and either the step response to ambient temperature changes or frequency response.2 92. Let us consider a l/S-inch OD pencil-type thermocouple in a well 0.4 107 49 Step-response curves are presented on Figure 11.3 7.22. let us examine the effect of using different annular fills: AIR OIL MERCURY 92.1 . 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. Consider next the effectof changing clearance. Calculator programs calculate the two time constants..2 7.005 inch r.020 inch 0.2 7.280 Mhcellaneow Measuremnts and Controls 2. and for air in the annular space we obtain: Annular Clearance 0. The service is steam and the base case velocity is 152 filsec.405-inch OD by 0.040 inch 0.34 Step responses are given in Figure 11. 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. a single time constant of 1. Neglecting the thermal capacitance of oil and mercury as we did may introduce some error here.0 €t/sec r. sec 91. If the thermocouple had been used bare.6 22. sec 92.6 93.6 8.8 5. For the same outside thennowell diameter.2 7.205-inch ID. To decide on the design of optimum thermowell/primary-element combinations. In view of the large annular clearance.. sec 7 6 .4 14.6 52. 7. 3.9 seconds would have been obtained for V = 152 fi/sec.6 16. there may be some error here in assuming a purely conductive heat-transfer mechanism for the annular fill. for V = 152 fi/sec. let us look at several cases.7 0. To support these interests. sec 7 6 . To select optimum installation practices. Example I : Gm Flow. Illustrative Examples-Forced Convection To illustrate the application of the model mentioned above. The details are not reproduced here but are presented elsewhere. for the same thermocouple.’ The model is an improved version of one discussed in Chapters 21 and 22 of reference 2.

11.21 Effect of velocity on step response .8 Temperature-Measurement DynamuZr 281 FIGURE 11.

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

sec 25.7 6.24. Annular Clearance 0. Applications For large-scale processes. the major time constant changed very little u t l velocity ni became very small.0005 inch are plotted in Figure 11. liquid-liquid heat exchangers slower yet. Next let us consider flow of an organic liquid. sec = 1.25 0. sec 7 6 . v r. oil.3 1. Next let us consider the effect of different annular fills: air.1 31..53 1. a single time constant of 0.51 2.19 For ol the value of thermal conductivity used was 0.525-inch OD by 0. Example 2: Liquid F h . Let us look first at the effect of different liquid-flow velocities..6 3.e s r n n Dynamk eprzeMaueut 283 As demonstrated.0 25.8 T m e a w . If the thermocouple had been used bare in the first case (10 ft/sec).6 10. . 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. sec 7 6 . which are determined mostly by the service.52 second would have been obtained. and mercury.055 inch 0.079 i e.260-inch ID..5 In this particular case. Again.3 1.3 1.005 inch reduces the major time constant by almost a factor of 6. condensers a little less so. 00 v= 1. Thermowell has a 0.3 61.005 inch r.01 k/xc ft/= 25. AIR OIL MERCURY r. even though the mass of the thermowell is increased. sec 7 6 . sec 228 1. reducing the annular clearance from 0. In reality the size of such equipment has little to do with its dynamics. Comparative step responses are shown in Figure 11. Comparative step responses are shown in Figure 11. step responses @"C/ft) 0. The pencil-type thermocouple has a 0.3 4. in this example velocity was maintained at 10 ft/sec and annular fill was air.0 v = 0. one often hears the arguments that fast temperature measurement is not important because large didlation columns.5 16. reboilers are very fast.040 to 0.23.1 fi/xc v = 0. Consider next the effect of varying thermowell internal diameter while holding the external diameter constant. For instance.6 fi/sec 26.25.26. and so on are slow.250-inch OD.3 Step responses are given in Figure 11.11.8 1. all for 10-ft/sec velocity. heat exchangers.

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

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

286 M? k- Memuremenk and Controb FIGURE 11.25 Effect o annular flll on step response. v = f WseC .

8 Tempermre-Measurmt Qymunkv 287 FIGURE 112 6 Effect of annular clearance on step response .11.

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

12 We have not found a convenient way to compensate for this with analog equipment without interfering with antireset windup. 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. however. A simple dual-valve installation is shown in Figure 11. 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. it may mean controlling two or more valves fiom two or more controllers. I O CONTROL-VALVE SPLIT RANGING “Split ranging” is a term applied to techmques for controlling two or more valves from one controller. it has been shown81. has great flexibility. Normal level control of the reflux drum is via the distillate valve. In effect this makes the large valve signal range 9.0-psig input signal causes the output to be 9. Cascading temperature or other variables to flow-ratio control should be avoided unless ratio turndown and flow turndown are less than 2 :1. Here we have a large valve and a small one in parallel. Although the divider technique is sometimes objected to on the 2$ that the flow-ratio loop has nonlinear loop gain.0 psig. however.: both techniques are characterized by nonlinear gain. as mentioned earlier. consider Figure 11. If overrides are involved.0-psig signal produces an output of 9. and plant maintenance people do not like to have to keep track of nonstandard valves. See item 2. This technique.12 2. the large one remains closed until the small one is fully open.11. The gain 2 relay for the large valve is biased such that a 12. the gain 2 relay is biased such that a 6.0 psig. the relays may be so biased that the large valve starts to open slightly before the small valve is wide open.0-9.27.0 psig. This applies whether one uses the divider (closed-loop) or multiplier (open-loop) flow-ratio scheme. This means that if . As another example. In effect this bias makes the small valve signal range 3.0 psig. 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.10 Control-Vdve Split Ranging 289 multiplied by a gain factor (sometimes remotely set) to calculate the set point for a manipulated flow controller. For example. 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. For the small valve.0-15. With linear flow measurements.0-psig output. In pneumatics the amplifiers are commonly fixed-gain relays with adjustable bias. The former lends itself well to our preferred antireset windup scheme for cascade control where the latter does not. the divider causes less nonlinearity than if square-law measurements are used.28. The h7el controller is a gain 2 relay biased such that a 9.0-psig input signal causes a 9. I 1. The small one opens first. Such an arrangement is used to get a large flow turndown. This technique is not very flexible.

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

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

01. Buckley.. P.. “Liquid Level Mea“Problems in Slurry Service Instrusurement in Distillation Columns. “Dynamics of TemJan. N. 1955.0-6. To protect against low level. 12( 1):45-55 (1973). 9. Fifth National Chempid 69 (Feb.Eng.” presented at Texas of Chemical Operation. 6(6):34-39 (June and Se-. we provide another -4 relay so biased that for a level transmitter signal of 4. Flow Equations. S. “Optimum Control ISA Formulas.. S. ANSI/ISA S75..0 psig. 1958.. This gives the air-to-close reflux valve an effective signal range of 12. S. Buckley.0 psig. “Control Valve 15. “Analysis of System 1964. L. Buckley. Houston. Englewood 1979. Research Triangle Park.f Process 4th ed. P.. Buckley.0-15. P. (May 1981). 4. Jan... 59(2):66ceedinds. Berger. 10. 13. and R. S.. Driskell. 1963).292 Mticelluneow Measurements and Cmttrok the level drops to 25 percent.C.. Schmoyer.. 7. the relay output is 9. 2.. Sanders.” ISA J.” presented . 3. Advanced Fluid Meat Texas A&M University. Control Valve Sizing “Computer Control of Distillation Reflux.” Chem. S. C. Cliffs. N. “Design of Pneumatic Flow Controls.. P. C. Del. Schnelle. Driskell. Prod. A&M Symposium. 1983. W h g t o n . 33-48 Valves for Pumped Systems..5 psig from the level transmitter. B. Buckley. McGraw-Hill. New York. Symposium. R. perature Measurement.”Hydrocarb. ISA Conference. 1959). ISA. “Selection of Optimum boilers-Liquid Level MeasureFinal Element Characteristics.’’ ISA Paper NO. the distillate valve is wide open. 1964. P. 269-281 1951. J. 6. 1963. S. Binder.” INTECH. the distillate valve is closed. Buckley.” presented at Texas A&M Symposium.’Handbook. Jan.0 psig in terms of the level transmitter. 5.J. 14.0 psig in terms of liquid level. L. “Internal Column Re. the relay output is 9. E. chaniq Prentice-Hall. REFERENCES 1. W. New York. Tex. K... Dynamics Improves Level Control 12. if the level rises to 75 percent. PeWys Chemical Enginem.. Proc. This relay makes the effective signal range of the reflux valve 3. To protect against high level. Hepp.”Proment.” presented at National ISA Trans.5 psig. P. we provide a gain -4 relay so biased that at 13..11.” mentation.. (July 1974). 1982. R. Techniques . 55-6-3. Lupfer. 1976. 8. P. Suppose now that the level rises above 75 percent or drops below 25 percent. D. Wiley.” 1984. D. P. E. “Control Valve Sizing with 16. ‘What is a ‘Fire-Safe’ Valve?. S. Arant. and D. Control.

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

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

-Systems where controlled variables cannot be measured directly but must be estimated fi-om other measurements (see Chapter 10. .. money. and since it is becoming increasingly common to have on-line computers. it is likely that these techniques will be used more in process control in the future. are multivariable. and then adjust the controller parameters so that the process follows a specified trajectory. to more sophisticated knowledge. In the following categories.24Touchstone Mehra ~ constraints. there is not enough time. 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. 2 0 -Systems where time-optimal control is economically attractive.2 KINDS OF INFORMATION AVAILABLE In almost all design projects.2~ name a few. The latter requires increasing understanding of control theory and more calculations.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. therefore. 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.. which permits the quickest decisions. The reader is referred to et al. liquid-level. or personnel (especially quahfied personnel) to do as good a job as we would like.12.27and M ~ r a r for ~ i ~ details. -Interacting systems with two inputs and two outputs where straight decoupling (noninteracting design) is not satisfactory. 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. Batch reactors and batch distillation are examples. This approach is suited to processes that are &cult to model parametrically. for further explanations. The reader is referred to Bristol. Since there are real economic incentives for more extensive use of multivariable control systems.2 K i d OfInfmmatiOn Available 297 parameters are determined to achieve some prescribed criterion of performance such as minimum variance. -Systems with time-varying parameters requiring adaptive tuning.2’ &zrom. and pressure controls. references to work of Brosilow et al. and have H o o ~ ~ s . Some usell insights are provided in a recent book by McAvof’ and a paper by Ray. ~ et al.u~”7’9 12. and to depart from them only when necessary. great pressure to make the necessary design decisions as quickly and inexpensively as possible. to Predictive control describes control algorithms that use an internal discrete convolution to predict the process output. This logically leads to consideration of self-tuning regulators such as hscussed by h r o m . There is.20 and Sa~try.). we progress from the most basic kind of infbrmation.

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

straightforward fashion. It also easily accommodates M y automatic startup and shutdown-only a switch and a lag unit need be added. with over 35 years’ experience in d e s i p g control systems by quantitative methods.3 FUNCTIONAL LAYOUT OF CONTROL LOOPS The overall functional layout of a control loop is extremely important. In a recent book.1. This structure satisfies all of the design requirements stated earlier. and with optional overrides. K. Few plant loops have all of these features. or both. 12. such as temperature measurement.. not from the controller output). are seriously deficient in one or more of the requirements listed previously. is shown in Figure 12.3 Functional Layout ofControl Loops 299 For some applications. predictor.6 A single loop with external reset feedback (ix. and feedforward compensation (or decouplers). The design must accommodate such diverse requirements as: -Minimum operator attention for startup and shutdown. feedback from the valve loading signal. Note that the controller gain. we have HP-41C programs that calculate either step or frequency response. TR. As of 1981 most vendor-designed digital systems were less flexible and less versatile than the best analog systems. but the point is that if a PI control algorithm with external reset feedback is used.. whether analog or digital. Rather it is a function of algorithms (mathematical functions) and their topological relationship to one another in a control-loop layout. one must resort to the textbook approach: Start from the beginning.12. and use frequency response or simulation or one of the newer methods to design the control loop. Today. we resort only occasionally to this method.may be nonlinear or may be adaptively . Hougen5 presented an excellent discussion of techniques and equipment. Many commercial controls. If all else fails. write the equations for all elements in the loop. experimentally. “Book’ Design of Control Loops When the above-mentioned sources are not adequate. 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 . but there were indications that this might be changing. they all fit together in a simple. and the reset time. it may be necessary to determine process or instrument dynamics.

c . U L vi ti i a . W' !sg E= LLm . e E s U 0 c . g z ai > m > z i c . 2 5 z s E s T: E z L 5 T 2.300 Approaches to Quantitative Desbn 5 e U E 8 2 m a .

and a is a constant. which might not always be the case. the .em(s)] + OR($) (12. The theory and mathematics involved in combmng feedforward compensation. and feedforward functions. r-.is usually desirable.1 box shown in the measured variable channel would be moved to the set-point channel. At this point we wish to point out that (1) the feedfonvard function. Care should be taken to avoid putting derivative inside the reset loop. is being increasingly accepted by control engineers.commonly termed “impulse feedforward. Derivative is shown here as a separate unit. equal to the reset time constant.1. It is not valid for nonlinear functions. usually in the range 6-20. Note that rR is the reset time constant.2 with feedforward compensation. It may be noted that Laplace transform symbolism has been used in this and the next two illustrations to represent reset.1) and (12.3) Considering next the system of Figure 12. controllers (PI and PID). The basic PI controller equations are: eC(s) = Kc[e&). Following is a brief review of the mathematics of a PI controller with external reset feedback and impulse feedforward compensation. rR. rD is the derivative time constant. derivative. for “increase-increase” action. and either difference equations or z-transforms are better suited for some digital applications.3 Functional Layout of Cuntrol Loops 301 set fiom external signals.1) and (12. it is nearly so when high sampling rates are used. additional functions may be required for more complex control systems. A single feedforward function is shown but. and anti reset-windup have been covered in detail elsewhere7 and are reviewed briefly below. overrides. and (2) making the feedforward time constant. The configuration shown is for “increase-decrease” controller action. K$--s/(r-s + l). let us write the equation for the summer output: . as will be indicated later. This feedforward function is also shown ahead of overrides and limiters.” is a convenient way of feeding forward without interfering with reset when external reset feedback is used. = e&) and we can combine equations (12. however. The basic structure of Figure 12.2) to get the conventional PI controller equation: (12. but is sometimes combined with PI in analog devices. This approach to feedforward also provides a convenient way to connect interaction compensators (decouplers) into control loops.2) If there are no overrides and no feedforward signal.12. Although not totally valid for digital computers or microprocessors.

302 Appodes to Qmntithve Des@ FIGURE 12.7) If we make 7f 7R.8) . (12.- 7Rs +1 7Rs 9 7 R s + 7Rs E (’) Kff 7j5 7R + 1 e.(s> +1 (12.6) which reduces to: e. this reduces further to: Om(s)= K c - +1 E 7R.F (4+ Kff!f(S) (12.1)..e&>l = E (5) we may combine equations (12.2).4). and (12. ( s ) = Kc E eJ-u(s) Kf735 (4+ 3 8f(s)735 + 1 + (12.(s) = = K.5) .4) If we assume that there are no overrides: &(s) = el44 Letting: (12.2 Single-loop with feedforward compensation M S ) = ec(s> + ef(z. (12.5) to get: [e& @ . @!E & + 1 (12.

This is not cascade control since only one controller is involved. If. and (3) constraints. 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. but changes are. For example. (2) product quality.(s) also shows a step change. 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. then O. An example would be column-top temperature control cascaded to reflux flow control. Our preferred scheme uses the secondary variable for external reset feedback to the primary controller. 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). If we look at these loops from the standpoint of classical.12. one might want to maintain a constant holdup time in a chemical reactor and use throughput flow to calculate desired liquid level. Occasionally one finds a single loop with external set.4 Adjustment o Controller Parameten (Controller Tuning) f 303 With this design steady-state signals are not fed forward.3) has at least two controllers: a primary or master one and a secondary or slave one. whose set point is normally set by the primary controller. Further work needs to be done to compare their performance with that achieved by PI or PID controllers with feedforward compensation. A true cascade system (see Figure 12. single-loop theory. Perhaps. that is. however. . 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. This usually occurs when one wishes to control a variable whose set point must be calculated. 12. the most common example is open-loop flow-ratio control. Cascade control probably has been used many times when feedforward compensation would provide superior control. But we increasingly cascade to flow controls that are usually so much faster than the primary loops that there are no stability problems. A ‘kild” flow is multiplied by a constant to calculate the set point of the controller of the manipulated flow. for example: Of(S) = S which is a step function.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.

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

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

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

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

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

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

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

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

Vapor flow will be assumed to occur without lags. A discussion of reboiler dynamics will be deferred to Chapter 15. or (3) not to change at all. 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. * This chapter is based on reference 5. first-order lags. and then derive an approximate model for a combination of trays.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. an increase in vapor flow may cause internal reflux: (1)to increase temporarily. which is followed by a long-term decrease in low boiler concentration. and heat-storage effects will be assumed to be negligible. 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. The first of these is commonly termed “inverse response” because it causes a momentary increase in low boilers in the column base. We now know. (2) to decrease temporarily.’ 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. In reality. that.1 3 i Tray DynamicsMaterial Balance* 13. although for many columns the first-order approximation is adequate. it is a chain of noninteracting second-order lags. as we shall see. depending on tray design and operating conditions. It has also been assumed that internal reflux is not affected by vapor flow changes. however. 3 13 .

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

13.1 Distillation tray schematic for flows and liquid evaluations .2 Tray Hya'radh 315 FIGURE 13.

jSmHk(s) .5) .0 .-[PDcKx(~) .P n + l ( J > l Next.3 16 Tray Dynumia-Mated Balance Equation (13. we may write: (13.10) Now since wx is determined by APx.3) - [Pn(s) .BL .9) 1& .4) where hd = liquid head in inches of liquid-flow pressure drop through the downcomer downcomer flow. gpm minimum downcomer flow area.--i 2 hP.11) .7) or and (13.(s) . 0 5 0 3 6 ~ &X BL PDC (13.6) B C (13.2) may be Laplace transformed to: AP.c (13. ftz &= Ah = This equation may be rewritten: (13.03 Q& hd = (100A h ) 2 (13. from page 508 of Van Winkle:* 0.KRPTR(J)I 8 6 (13.

= volume. of aerated liquid on active area of tray (downcomer volume excluded) active tray area. = H. f (area where ? bubbling occurs) height.13) or in Laplace transform notation: .13. = = HA = aerated liquid height. feet. Ibm. is the active tray holdup. Then: ~ .1) A..15) where W .14) Volume of Liquid on Tray The volume of liquid on a tray is related to the inventory and the density: (13.. .1. fi3.16) where av. of outlet weir (constant) inlet (see Figure 13. H. at tray = H. over outlet weir height. we assume a linear gradient across the tray: where V. By Laplace transforming we obtain: ( 13.15)] .. feet. + H. feet. ( ~ ) = H25) ( 13.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. (13.2 Tray Hydraulics 317 Aerated liquid Holdup and Gradient on Tray With reference to Figure 13. - 1 Pm [From equation (13. aw..


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

328 Dirtillation-ColumnMaterial-Balam Control FIGURE 14.1 Distillation column material balance .

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

8) in Laplace transform notation: (14.7) Note that: wF = feed.. (u+/q.7): (14. If the feed is subcooled liquid.11) whence .9) or (14. q = 1 and V.10) wf(s) = k2wf+1(2)+ k3WF(I) + kdq(f) Usually k2 = 1. q is greater than 1 and V is less than V.6) becomes: nt (14. Ibmlmin From equation (14.8) Since (wf+J++..). ' It is also true that: V.6) (14.Lf = -Fq In terms of weight u i s equation (14. = L f + V.Note that if the feed is a liquid at its boiling point. and k4 = w F . = V. we can rewrite equation and (14. In going back to equation (14. (wdLf)are constants.4) and expressing it in weight units. Ibmlmin wf = liquid flow from the feed tray. k3 = q. From the definition of q above we can write: Lf+l . although the last may not be true if the feed composition is radically different from feed-tray composition. + F + Lf+. we obtain: (14.5) (14.

17) where NR = number of trays above the feed tray a1 = N R T ~ Note that: (14. Ibm/min (14.14) where G&) is the cumulative effect of the individual tray hydraulic lags. each with a hydraulic lag.18) .q.14. Stripping-Section Liquid-Flow Dynamics The transfer function between w 1 and wfis: (14.k d s ) (14.15) where Ns = number of trays from the column base to the feed tray Enriching-Section Liquid-Flow Dynamics (14.1 Mathematrcal MoakL-Open Loop 331 or Wr-l(s) = ~ w A J )+ k7~14s). and k8 = wF. T~ (no inverse response assumed): (14.13) Usually b6 = 1. although the last two may not hold if feed composition is radically different from feed-tray composition.16) where wm = liquid flow (internal reflux) from top tray. k7 = 1 .

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

It exercises its most serious effect in those columns where base level is controlled via steam flow.Without proper design the level controller can become very confused. We will assume that each level controller is cascaded to the appropriate flow controller. and top. Overhead Level Control The necessary additional equation (no subcooling) is: = ~T(.2 CONTROL IN THE DIRECTION OF FLOW Let us look at a material-balance control scheme that is in the direction of flow. By providing the proper additional connections.2. For purposes of this chapter. The third factor is the simplified. and let column base level set bottom-product flow. It is characteristic of valve tray columns and some sieve tray columns operating at low boilup rates.c)KmbrKdrrGdt(S) x 1 KD w (14. Let feed rate be set by averaging level control of the feed tank. the level controller increases the steam flow. Limitations of Preceding Analysis There are three factors that limit the accuracy of the precedmg analysis. entrainment may be severe enough to invalidate the simple materialbalance model we have developed.14. let condensate receiver level set top-product flow. This is discussed in detail in Chapter 16. steady-state treatment of the feed tray. the feed flow. 14. we can design any desired type of material-balance control.and bottom-product flows are all inputs. Feed tray dynamics will be dealt with more rigorously in Chapter 18. psi lbm/min (WD)- Note: linear flow meter . steam flow. As can be seen. we do not believe this introduces a serious error. The second limiting factor is entrainment. external reflux. If the level becomes too high.2 Control in the Direction of FIap 333 The preceding equations now can be combined into the signal flow dagram of Figure 14.22) where Ke = - distillate flow-meter gain. But this causes a momentary increase in base level due to the extra liquid coming down the column (also due to thennosyphon reboiler “swell”). Normally we assume that the only way we get liquid overhead is by condensing vapor. But at high boilup rates. The first of these relates to the phenomenon of inverse response discussed in Chapter 13.

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

For averaging level control. this introduces little error. for a 3-15-psig output.2 G m ~in the Direction o Flow l f ( I V ~ = top-product ) ~ ~ 335 flow-meter span.3 S g a flow diagram-condensate receiver inl .25) FIGURE 14.14.T)Kdf (14. psi/psi = G. in feet. lbm/min Kdrr controller gain. psi/fi KdT Note that we have ignored the dynamics of level measurement and of the flow control loop.3.24) (AHT T ) where ( A H T ) T is the level transmitter span. (14.IVR($) ~chrGchr(. We can now define a characteristic time constant: K& = [THIT = PTAT(AHT)T &~(wD x) a .23) ATPTK~ = &d%l7-(5) (14.(s) = = controller dynamic gain receiver level transmitter gain. We can now prepare the partial signal flow diagram of Figure 14. From this we can see by inspection that: WD(4 - 1 1 + IVd5).23a) For pneumatic instruments: 12 psi (14.

336 DiniUation-Colurnn Matffial-Balanu Cmttvl If a proportional-only control system is used. Ibm/min = controller gain. .30) where (AHT). This is large enough to ensure that instrument and pneumatic transmission-line dynamics i will not be sigdcant..29) 1 + KC~BGC~B (s)Km/l~ fPl(4 . Note that for best flow smoothing or to another process step. one may need a larger (rH)*.fiom which we can see that: wB(s) 1 (14. If we fx K. = KdB psi Ibm/min 12/(~~)~=. psi/fi = We can now prepare the partial signal flow diagram of Figure 14.29a) (14. then we must achieve the desired [ 7 H ] T by proper choice of A T or (AHT)= both.25) may define: we = (14.27) Base Level Control The necessary additional equation is: (14.28) where KnsB= bottom-product flow transmitter gain.23) becomes: (14. (fpB)- = flow-meter span..fpR(s) where rR is the level controller reset time in minutes. it is recommended that KAT = 2 be chosen. as indicated in Chapter 16: tpg(s) pc(c(s) = TRs 1 . If a proportional-only controller is used. say 2-5 minutes (120-300 seconds). rR[TH]Ts2 +1 + rl?.4.s -k (14.fp&) PdBKngrB KHBGHB(s) By analogy with equation (14. as Usually it is desirable to have [rHIT small as convenient. for best control of the associated column.26) whde for proportional-reset level control it becomes. equation (14. even an additional or buffer or surge tank. psi/psi KmhB base-level transmitter gain.is the base-level transmitter span in feet for 3-15 psig output.

then the proper time constant ) ~ is achieved by choice of ABor (AHT)B. As before.31) WB(S) - TR$ +1 tPl(S) . It is usually desirable for best control of the associated column to make ( T ~ = 10-15 minutes. Note that for best flow smoothing to another process step.5 where top-product flow is the demand flow.(S) while for proportional-reset control it is: WB(4 1 [THIBJ Wl(4 +1 (14. equation ( 14. level controllers are cascaded to flow controls with linear flow meters. and if KdE is specified. both.32) where T~ is the level controller reset time in minutes. tank. one may need a large ( T ~ ) or even an additional surge or buffer ~ .3 CONTROL IN DIRECTION OPPOSITE TO FLOW As an example let us choose the case of Figure 6.5. FIGURE 14. this statement would still be true. base level sets feed flow.fpp(S) TR[TH]BS’ + TRS + 1 (14.tp. and both reflux and bottom-product flow are ratioed to top-product flow. Note that the two level controls are independent and noninteracting.3 Control in Dire& Opposite to Flow 337 For proportional-only control. 14.14.29) becomes: . If we were to add reflux/feed and steam/feed ratio controls. For proportional-only control. or KchB= 2 is recommended.4 Signal flow diagram-column base . condensate receiver level sets steam flow. We can now prepare the overall closed-loop material-balance dagram of Figure 14.

5 Signal fo diagram-material balance control in direction of f o lw lw .338 DtitiUatMn-Column MaterialB&m Control FIGURE 14.

it is redrawn into the form of Figure 14.7. = psi lbm/min 12/(%)max lbm/min ( B J ~=) feed flow-meter span.qj = steam-flow transmitter gain. the AP transmitter should be followed by a square root extractor.. Ibm/min 12/(w5). Also. If an orifice flow meter is used.14. Base Level Adjusts Feed Flow (14.35) Physical techniques for accomplishing this are discussed in reference 2. Bottom-Product Flow Ratioed to Distillate Flow (14.36) Closed-Loop Signal-Flow Diagram The closed-loop signal-flow diagram of Figure 14. where (w5)- = steam flow meter span. since we have a cascade system. This is a much more complex diagram than that of .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. ~ ~ Reflux Flow Ratioed to Distillate Flow Let: Q(5) = ~RDGRD(-T)fPD(5) (14.34) where = feed flow-meter gain. the steam flow transmitter should have a linear relationship between flow and transmitter output. lbm/min Nore that we assume the flow control loop to be very fast compared with other dynamics.6 may now be prepared. To show the relationship between wD and wF more clearly.33) PSI K?fj where K.

6 Signal flow diagram-material balance control in direction opposite to flow .B a l a w Control FIGURE 14.340 Disdla&a-Column M a t d .

3 Control in Direction Opposite to F h 341 9 r -- L 8 g Lr. 0 C a 0 e 2% -m g gs 3 k gs .14.

Vapor Sidestream As an example let us consider a column such as that illustrated in Figure 7.w The steam flow is set by ratio to the feed flow: fP&) = fPv(4 (14.37) K+ = sidestream flow-meter gain. not composition control.Figure 14. a small bottom-product purge. and steam. This point of view is at variance with that sometimes expressed elsewhere in the literature. Base level adjusts side draw and reflux drum level sets reflux flow. Feed is assumed to enter at its boiling point.38) (14.4 MATERIAL-BALANCE CONTROL IN SIDESTREAM DRAWOFF COLUMNS Let us consider two cases: (1) vapor sidestream. The functions Gm(s) and GBD(s) have to be chosen with care will because of potential difficulties with stability. Ibm/min where (14.This column has a small top-product purge.. and a side product that is most of the feed. Three flows are ratioed to feed: top product. Also. control functions hs must be correctly preselected. and control-loop parameters calculated ahead of plant operation.fP.)max psi Ibm/min -. these two functions must be chosen with primary regard for material-balance control. (tp. and (2) liquid sidestream.12 - and .1. 14.39) and (14. it is probably apparent that conventional “tuninjf‘ procedures are essentially useless for a system of t i complexity.40) . bottom product.(s) = KubB -K h B G c h B ( w . B ( J ) K+ = sidestream flow.5. Finally. About the only new relationship we need is that which defines vapor flow up the column above the point of side draw: tp.

Then we define: (14. This does not mean that level control via the small flow is either impossible or impractical. .43) (14.42) The pertinent equations may now be assembled into the form of Figure 1 .5 Top and Bottom Level Control Com&u&ms 343 The two remaining ratio controls are defined as follows: 1 WB(S) = K#$FKR3GR3(5)- (14. we can prepare the signal flow 49 diagram of Figure 1 . Another expert recommends the second. the reflux flow may be ten or more times greater than the topproduct flow. There are many columns operating today with condensate receiver level controlling reflux and base level controlling bottom-product flow. There are other columns in which condensate receiver level adjusts top-product flow while base level manipulates steam flow. at least one of the two levels must control a drawoff flow. usually steam.45) With the same control scheme in mind. 48 Liquid Sidestream In this case we will assume that the liquid side draw is taken from a point above the feed tray.14. for some reason. In a superfiactionator. It does . have both levels adjust drawoff flows? It seems to us that it is largely a matter of convenience. It is probably apparent that we cannot follow both recommendations. 14. One well-known author argues strongly for the former. WF(4 (14. Inventory in the receiver may be regulated a little more readily by manipulation of the large flow than of the small one. assuming that we cannot. How do we choose between them.41) K?@ 1 and %S () - K$F&&&)- rc.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. Controversy also exists as to whether it is better to have column-base level control bottom-product flow or the reboiler heating medium. for example.. .

8 Material balance signal fo diagram-vapor sidestream drawoff lw .344 Dutdhthn-Column Ma&riul-Balana Control FIGURE 14.

9 Material balance signal flow diagram-liquid sidestream drawoff .14.5 Top and Bottom Level Control Combinatwns 345 FIGURE 14.

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

we are still lefi with the job of solving partial differential equations. It involves the following simplifjring assumptions: 1. one should probably resort to simulation. and (2) the vertical design with vapor in the tubes and coolant on the shell side. 2. Subcooling is negligible. For important applications. 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. of their solutions are those by Hempe12 and by G ~ u l dA ~ largely empirical. has been proposed by Thal-Lar~en. A condenser with the cooling. 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. Heat storage in the heat exchanger metal is negligible. An approximate analysis for a condenser that has a single pass on the coolant side is presented in Chapter 24 of reference 1. 3. Perhaps the easiest to read descriptions .1 LIQUID-COOLED CONDENSERS WITH NO CONDENSATE HOLDUP s mentioned in Chapter 3. Although these reduce the complexity somewhat. Mean temperature difference is arithmetic. and will not be repeated here.1 5 a Condenser and Reboiler Dynamics 15. much simpler model. 347 . we will not pursue their dynamic equations further. where subcooling may be of concern.~ most liquid-cooled Since condensers are fairly fast with time constants in the range of 10-60 seconds.

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

7) (15. WA average bulk temperature. back pressure downstream of the vent valve. As the liquid level in the shell varies. Also: ( 15.8) (15. so do heat-transfer area and rate of condensation.10) (15. We will make several simplifylng assumptions: .9) where w. 15.15.12) The first-order dynamics of this type of condenser make it much easier to control than the condenser with once-through coolant. O K .2 Fhded Co&men-Open-Loop Dynamics 349 (15.1).1 (see also Figure 3. of coolant coolant holdup. and rate of removal of liquid fiom the condenser shell.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. Ibm The subscript OL means open loop. We would like to find out how condensing pressure and rate of condensation are affected by rate of vapor flow into the condenser.

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

of condenser flow. lbf/fi? pressure. pcu/lbm. lbf/ftz. f? i condensing temperature.19) (15. ft3/sec ~ u. some in the time domain and some in the s domain: (15. Rv = = = ATC Q = . downstream of vent valve vent valve resistance.15.18) (15. = T = wc = wo = s= = = ww A. of condensing process vapor liquid outflow. lbm/sec vapor outflow.14) (15. lbm/sec submerged heat-transfer area. pcu/sec fi? "C condensing heat-transfer area. = Tc = TA = Pc = PR = w .15) (15.21) where v= qc = = system vapor volume. Ibf sec/fi5 pounds of liquid in shell between bottom of lowest tube and top of highest tube total heat-transfer area. ft3 rate of heat transfer. pcu/sec condenser heat-transfer coefficient.13) (15. "C or O K vapor space pressure. lbm/sec vapor inflow. A. p "C or O K Ibm/sec process vapor condensed latent heat of vaporization.17) (15. ft2. "C or O K average coolant temperature.2 F h h d conrlensers-Op~-Loop -b 351 We may now write the following equations.20) (15.

5-3 minutes. we obtain: (15. (15. P.22) qc(s) = UcZc AT(@ + UchTAc(s) (15.4.3 and the final reduction in Figure 15. we combine equations ( 15.+Kco Fc a - s+ 1 U A SIAp CC TA(s)+ &PR(s) s+a R V + w&) where 1 (15. From this we may write by inspection: RvlPv ( s PC(S) = [: a + a) - -Rv a -2 + 1 V + d . The above then reduces to: KC F where (5 TFCS +1 + a) . various inputs.13) to through (15.352 Chdenser and Reboiler Dynamia Upon Laplace transforming equations (15. The first reduction of this is given in Figure 15.16). to Various Inputs To find the response.24) Response.2.25) The first term on the right-hand side of equation (15. Pc(s). and (15.(s). one of which has a typical time constant of 1. the denominator has a large damping ratio so that the quadratic may be factored into two terms.14).23) (15.13).25) may be written: KC F Tis2 +2 ( 1 + a) 5 ~ ~ 1s + For all such condensers we have studied. while the other is only a few seconds.24) into the signal flow diagram of Figure 15.

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. -c 3 0 e 2 LL $e ....

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 .

.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 . 8 33 2s LLr.

Approaches Infinity For the case where Rv becomes very large (very little vent gas flow).26) + where Kkc and = ucxc aTc avpc+ --r Ecp P v a c 1lPv Response.6.+ w&)] uAd’TA(s) s+a The first term on the right may be written: Kkc (s + a) S(&S 1) (15.2 can be redrawn to show wC(s)as an output as shown in Figure 15. the last equation reduces to: PC(S) = W P V ) (s + 4 s + 1) uczc aTc avpc+ -& Pv a r c + . This can then be reduced to the form of Figure 15.(s).356 Gmdenscr and R e W LIpunia Case Where R.5. Various Inputs to The signal flow diagram of Figure 15. fiom which we can write by inspection: . w.

15. The analysis should take into account the temperature of the entering liquid and the sensible heat effect of the liquid mass.R . .5 VIR $ ucAc arc a PIPc + P P h 1 + iJ arc (15. The various equations may then be written as follows. Approaches Infinity Again.3 Reboilrrs-Open-Loop Dynamh 357 where L($Rvr a AP + 1) 1 + ( Y .27a) Case Where R.28) 15. 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. as R.3 REBOILERS-OPEN-LOOP DYNAMICS We wish to make an analysis of column bases with associated reboilers where there is signrficant liquid holdup. + Kco Fc $2 + P P C a a s + l X i c ( g R v s + 1) 1 (15.7. the preceding equation becomes simpler: X wc(5) = L . Such an analysis also applies to vaporizers with associated separators or knockout drums.

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 .

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 .15.

34) (15.hrp f B .30) Material Balance or j [wi(t) - wBU(t) - wE(t)] dt = WB(t) .7 v Schematic representation of column base and reboiler holdup .36) (15.32) [vi($) .31) (15.(s) = -PCS(5) 8PCS (15.35) (15.360 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) .33) (15.) qT(5) = FIGURE 15.T ~ 4 s ) l (Note that heat storage of the reboiler metal is neglected. ) 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 (15.37) 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.wBu(s) - WB (J)] 1 Steam-Side Dynamics The following are taken fiom Chapter 25 of reference 1: aC TS Tc.

where V.. The column-base pressure dynamics may be represented by: (15. upstream of control valve = valve stem position The terms aQv/aP. f? T U = boiling temperature of process fluid. lbm/min column-base pressure. ft3/sec qT = heat transfer. Ibf/fi?. = reboiler shell volume. "C P&) = reboiler shell pressure. lbf/fi? From the preceding equations we can prepare the signal flow diagram of Figure 15..9 where three new functions are defined: 1 (15.15. ft2 sec i A R = heat-transfer area. "C B = rate of steam condensation. This can be partially reduced as shown on Figure 15.3 Reboilnr-Open-Loop LlynamS 361 In these equations: Tu@)= steam condensing temperature. lbf/fi? Q = steam flow rate. fi3/sec = steam latent heat of condensation.. looking up from the base. pcu/lbm = steam density Ibm/ft3 at Pa and Ta = acoustic capacitance of reboiler shell. and aQv/dXv may be evaluated by the methods of Chapter 15 of reference 1.40) xPsc As2 1 (15. lbm/ft3 rate of boilup. aQv/aP.38) column acoustic impedance. fi5/lbf = VJT' .8. ft3 = steam supply pressure. lbf sec/fi5 density of vapor boilup.41) . pcu/sec p/"C UR = reboiler heat-transfer coefficient.

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

lbm/fi3 Since Ti varies slowly or not at all. We may then prepare the signal flow diagram of Figure 15.3 Reboden-Open-Lmp Dynamia 363 Base Level Control Cascaded to Steam flow Control We assume here that averaging level control is desired. FIGURE 15.9 Partial reduction of figure 15. fi? A B PL = liquid density.10.8 .15. 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. T4s) = 0 .

(b ( ap.. =0 Flow control loop gain and dynamics are determined entirely by the instrument characteristics.42) ap. then: 1 1 .W) since v - = 1 (15. QV aQ arc.364 condenrer and Reboiler Dynmzia Noncritical Versus Critical Steam Flow If steam flow is critical. 4 aTm (15.10 Signal flow diagram for base level control cascaded to steam flow control . Note that: 1 - CR1 W) = 1+-- 1 1 Pa CR5 A.43) FIGURE 15.

aQ.44) This leads us to the final signal flow dagram of Figure 15. Gf (4 B(s) ax./dPa = 0. the system should always be operated in one flow r e p e or the other. the natural frequency of the level control loop will be much lower than that of the flow control loop. 1 1 K+G*(s> K+ (15. 1 1 .11. In most cases steam flow is noncritical. we can prepare the signal flow diagram of Figure 15.9 and by assuming A(s) = 0. 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. Note that: K V GV(4 Kcf aQ. To avoid problems with flow controller tuning. permitting higher loop gain.365 If steam flow is noncritical. As we will see in Chapters 16 and 17. it usually will be possible to use much simpler reboiler models than that discussed in this section.W) has lead characteristics and a static gain of less than unity. Practically speaking ~ ( s ) . 444. and A($) reduce to constants. a Q V / e S Signal F o Diagram Simplification lw Since we have called for averaging level control. faster flow-control response.. and much higher flow controller gain. one may substitute l/K+for the flow control loop. X 5 - Base Level Control by Direct Manipulation of Steam Valve From Figure 15.12. Further Mathematical Simplification It has already been indicated that for averaging level control cascaded to steam flow control. .. Note that if steam flow is critical.

That area covered by liquid permits only sensible heat transfer h m the condensate. The signal flow diagram of Figure 15. As discussed in Chapter 4.5 for a flooded condenser may be used as a starting point. If the steam supply pressure is constant. it is a vertical thermosyphon reboiler. Usually.4 PARTIALLY FLOODED REBOILERS The partially flooded reboiler is similar in many ways to the partially flooded condenser. it is controlled by throttling the steam un condensate.366 CDnrinrer and Rebode-Dynmnia 15. t i is a small heat load compared hs with that of the condensing steam and is treated as neghgble. although not always. In this case pressure dynamics are essentially negligible. FIGURE 15. the steam condensing temperature is also constant. Section 2.11 Final signal flow diagram for base level control cascaded to s e m flow control ta . which in t r varies the condensate level in the shell and thereby the heat-transfer area for condensation.

lbm/min = steam latent heat. ft2 = condensate on shell side. lbm/min rate of steam condensate withdrawal.13 may be prepared next. As shown by the reduced form of Figure 15. f ? for steam condensate .ART ~WSHI= total heat-transfer area of reboiler.MR UR ATa sf wI wo(s) (15.12 Signal f o diagram for base level control by lw direct manipulation of steam valve . P i mn = average exposed heat-transfer area. . pcu/lbm p/"C = heat-transfer coefficient. lbm = = FIGURE 15. aS wH ART WH S rate of steam condensation.14: WC(S) -- - 1 A* s + l .4 PartiaUy F h U Rebodem 367 The signal flow diagram of Figure 15.15.45) where w c w o Ast UR x7 i aA.

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

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 .g r 3 UJZi Mal U 5s La SZ .15.

Solving equation (15.47) wc( t )is a perturbation variable (deviation from steady state). From Figure 15. "C S Design experience with flooded reboilers is limited but indicates that typical time constants are of the order of 2-5 minutes. 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 decays to zero as time goes to infinity. .Condensate level. The former are usually much less expensive.46) Commercial lead-lag compensators commonly have values of a between 6 and 30.49). Ibm boilup. would eventually increase (thereby decreasing&) and eventually wcwould have to equal w. however.Tsu. and therefore in wc. we would intuitively expect a step increase in AT to cause an initial increase in heat transfer. Ibm/min laten.Condmer and Reboiler Dynmtia weight of condensate that will fill shell side of reboiler.49) In equation (15. of boiling liquid. It is interesting to look at the response of wc to a change in AT. As shown by equation (15. heat.14: (15.47) for a step increase in AT shows this to be true: so that: (15. Some provide a fixed a and some have adjustable a..47) If w. pcu/lbm TT . were held constant.

feet Pc. WBW -41. ABW .57) (15.TB. feet L Hs = condensate level in the shell. As shown by Figure 4. We can now write the following equations: PcS(s) = Z(5) ~ 4 5 .5 PARTIALLY FLOODED REBOILERS FOR LOW-BOILING MATERIALS In Chapter 4. lbf/f? abs p L = condensate density. ap.15.just a little above atmospheric pressure.1): HLPL BL = Hs pLBL + P. B C B C where + APLine (44 H = loop seal or standpipe height. = steam pressure in shell.R a =- -P. (15.1).56) (15. is and that HL pLBL is constant.55) (15.4.(5> P L (BL/Bc) (15. Section 2. From equation (4.54) q ~ ( t= UR R ( ~ ) ) A AT(t) AT = T.53) ar H ART [Hrlrnax &(5) = a R~Hr(5) U . .50) (15. The static force-balance relationship is given by equation (4.51) = ---&) aT. if we assume that the line loss Mhne negligible. lbm/ft3 The vented loop seal maintains P.52) ( 15. (15. then: 8.wc(5) ) 5PY (VRJ (15.5 Partially Flooakd Rebodem fm Low-Boiling Materials 371 15. we discussed a variation in flooded reboiler design for low-boiling materials.58) wc = @ A .we throttle the steam instead of the condensate. H45) = .

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 .

ft3 Tu = condensing temperature. if there were a step change in Tsv. Hempel. Gould. 2.60) (15.59) where (15. New York.. Buckley. there would be an immediate spike in wc.62) where (15. write: (15.. A. Ibm/min wc = rate of steam condensation. REFERENCES 1. which may then be reduced to the form of Figure 15. A. Ibm/ft3 Other terms are as defined earlier in this chapter. 1964.15. which would slowly decay to zero.373 where ws = steam flow. Chemiurl PromControl. .55) into the s domain. Ibm/ft3 V = average free volume in shell above liquid level.64) As in the previous section. TramASME. 3.244 (1961). L.16. "C p L = condensate density. Techniques o Pmcea f Control. S.61) We may now prepare the signal flow diagram of Figure 15. P..63) Also: (15. Wiley. From the latter we see that: (15. To get equation (15. Ibm/min p I = steam density in shell.

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

’ As stated in Chapter 1. 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. where wR is downstream closedloop resonant fi-equency. For most of these applications. For the latter we make rH3 10/wR. surge or feed tanks.1 INTRODUCTION or distillation columns level control refers. 3. column bases. Maintaining inventory or accumulation between an upper and a lower limit (not at a fixed value). radians/minute. 2. but rather of achieving the smoothest possible transitions in manipulated flows in response to disturbances. we need only a few equations: 375 . to overhead condenser receivers (reflux accumulators). the chief function is not that of holding level constant. Before getting into specific applications of level control on distillation columns. This is “averaging” level control.1where the level is controlled by outflow and there is no sipficant level self-regulation effect. in most cases. and sometimes steam condensate receivers. or (2) the tank is to be sized and the level control system designed to achieve flow smoothing adequate for downstream composition controls.2 LEVEL CONTROL OF SIMPLE VESSELS For a simple vessel such as &own in Figure 16. Balancing inflows against outflows at a point in a process. the b c t i o n s of averaging level control are: 1.1 6 f Liquid Level Control 16. let us review briefly the theory of averaging level control on simple vessels. Providing for smooth and gradual changes in manipulated flows to avoid upsetting process equipment. 16.

H = level transmitter input span. = controller output signal fi3/min a / d 8 . fi? (vertical. where = = inflow. psi (see &scussions in Sections 3.2) (16.1 Level control of simple vessel . fi3/min outflow.Q4-t) = H($) Ltquid Lmel Control As &PI&) (16.3) (16.4) Q ~ J=)4 2 0 64s) de. or flow control loop gain.(s) = K&G&(s) x ed(4 (16. cylindrical vessel A assumed) H = liquid level. = valve gain.1) = LbH(s) e. feet of process fluid Kd G&) = controller transfer h c t i o n 0.9 and 4. feet k b = level transmitter output signal L b = 12 psi/AHT for pneumatics M .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. fi3/min = vessel cross-sectional area.376 Q4s) .

2. Equations (16. and if the valve has an installed linear flow characteristic with a wide-open capacity approximately equal to four times flowsheet flow. flow of flow-meter span.4) may be combined into the signal flow diagram of Figure 16. Then equation (16. GS. ft3/min maximum The analysis employed here is further simplified in that the effects of variable valve-pressure drop are omitted. from which we may write by inspection: (16. as are transmitter dynamics. A AHT..16.TH (16. 3-15 psig).9) QFS This last equation is very useful for finding the desired holdup. provided a value of rH is specified.5) becomes: (16.7) and equation (16.1) through (16. for example. then: A 4 0 [ h & A KJchdB.2 Level Control of Simple = Vesseh 377 where m]..6) becomes: If the input span of the valve positioner is the same as the transmitter output span (as. [TI 4Q3 - A M T Kch . Usually we choose Kch = 2 and bias the .5) and (16.6) Proportional-Only Control For this case KchGch(s) becomes simply Kch.

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 .

” Before exploring the theory.2 Level Conh.16. . Since the unenhanced PI controller with Kch< 1 does not ensure that the tank will not run dry or oveiflow.9c). however.01 @Sample Vesseh 379 proportional controller so that the tank level is midscale on the level transmitter. In recent years. This allows for some variation in liquid specific gravity.9a) A AHT = v~= THK*( 4 % ~ ) = STWQFS (16. 2. 3.9a) or (16. “Nozzles” here refers to &ose used for connecting the level-measuring device to the vessel. This leaves the top 25 percent of the transmitter span for overrides. Consequently we usually specify KA = 2. With floating pressure columns and with the trend toward small control-valve pressure drops for energy conservation. this is virtually mandatory to counteract the effect of control valve up. inexpensive. If tank size is specified. past practice has been to have high and low alarms or high and low interlocks. 3. that is. it must be followed by a square-root extractor. and almost foolproof (at least when implemented with fixed-gain relays). The flow measurement must be linear.9b) For level control cascaded to flow control: A AHT = V T = THKch (Qo)rnax (16. it has limitations: 1. m T / 2 . It provides. use of Kch= 1 is risky. the flow smoothing is limited by rHbecause one cannot safely use Kch< 1. If tank size is not specified but is to be calculated from equation (16. the PI controller enhanced or augmented with auto overrides has provided an almost foolproof way of keeping liquid within the vessel. 2. if an orifice flow meter is used. Then: (16. let us make some additional design assumptions: 1. much more flow smoothing for a given size vessel than will a proportional-only controller. This is required for proper functioning of the auto overrides.VTis proportional to K d . the valve may not be closed when the level is at zero. Note that the level transmitter span is usually less than the nozzleto-node spacing.9~) Proportional-Reset Control Although the proportional-only control system is simple. under most circumstances. Although in theory there exists a number of controllers that permit one to use Kd < 1. a large tank will be required since for a given rH. Because mechanics ofien do not calibrate valves precisely. the PI controller has been most popular. Nozzle-to-nozzle spacing is so chosen that process operation will be satisfactory with the level at any location between the nozzles. Normal set point for the PI level controller is midscale of level transmitter span.and downstream pressure variations. Level control is cascaded to flow control.

but a level indicator is desirable. The former also implies that whenever the flow control station is not switched to “remote auto. Substituting this into equation (16.11) (16.3./QC by 1/K4.10) ?-as+ 1 (16. but the former permits more accurate.e. 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.” This virtually eliminates “bumping” when the flow controller switches fi-om either “manual” or “local auto” to “remote auto. This may or may not be desirable. As a result of a number of studies (unpublished). we must resort to digital simulation. quantitative design. we get: K-. But regardless of override location (other than auto Overrides). 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.380 Liquid Level Control A schematic for PI level control on a simple tank is given in Figure 16.” the overrides are out of service.” This assumption permits us to use Laplace transforms and frequency response. The latter arrangement (overrides in the signal path to the control valve) has been far more common. level controller has very fast reset) whenever the flow control station is not switched to “remote auto. 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. If it is desired to predict system behavior when forced by disturbances large enough to cause an override to take over.12) .” A primary control station is not necessary. (16.5) and replacing aQ.. 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. The transfer function for a PI controller is: where rR is the reset time in minutes. we provide a switching design that causes the level controller reset to be bypassed (i.

The damping ratio. 2. The damping ratio preferably should be at least unity.2 Level Control ofsimple Vessels 381 Similarly. say that it is umdzthnahy stable.15) (16. In desigmng a level control system with a proportional-reset controller several practical considerations must be kept in mind: 1. ~ . The loop therefore becomes slower and less stable at the same time. approaches zero and the control loop becomes very resonant. This is commonly called a “gain cycle” since it is caused by excess gain. 5. then as one decreases Kchtwo things happen: 1. usually they are designed to be much faster.12) and (16.13) has some interesting characteristics not widely appreciated. TQ becomes small.16. from equation (16. transmitter and valve dynamics become sigdicant. as shown by equations (16. and the loop eventually becomes unstable. l / Flow and level regulation in that frequency range will be very poor. while holding TR constant.6): (16. A low damping ratio. TQ becomes very large. on the other hand.14a) Let us also define: (16. one increases Kd. 2. Adjacent or related process controls must be designed with closed-loop natural frequencies much different from that of the level control.15) we can see that if TR is fixed.13). E. approaching instability.causes severe peaking in the frequency response in the vicinity of the closed-loop natural frequency.14) and (16. This resonance is sometimes called a “reset cycle” since it would not exist if the controller did not have automatic reset.15a) From equations (16. Since the loop approaches instability for both very large and very we small values of Kch. It is a quadratic whose damping ratio is : or TR = 4 t2TH (16.13) Now the denominator of equations (16.12) and (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. Augmented PI Controllers A plain PI controller. 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 . useful in some circumstances. the transfer functions are identical if one reduces the PL level controller gain. It is not a standard commercial item but usually can be assembled with various standard devices. For most other applications. It has the feature. Kd.38 for 5 1. even if tuned for 5 = 1.17) (16.14 for5 1.0 or use a proportional-only controller. we have found two approaches useful: . one cannot readily obtain the necessary rR with a particular commercial controller. is It frequently happens that for a desired and specified ICd. The major instrument manufacturers can usually furnish modification kits or modified controllers with a larger rR.14a) that: TR = 4 TH 7R/2 (16. cannot guarantee that level will be held within the vessel. To protect upper and lower permissible level limits.382 Liquid Level control 3. 5 = 1 should suffice and places less of a burden on available controller settings. if the system we have been considering is subjected to a step change in inflow.18) TQ = 2 TH = and TQ = 4 TH = TR/4 (16. 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. it requires auto overrides just as the PI controller does. of not needing antireset windup. For and 5 = 1 it can be seen from equation (16.048 for5 = = = 0. For example. by a factor of 2. For such applications we should choose 6 = 2.0 The fact that the outflow swings more than the inflow can create serious problems if the process is running close to capacity.16) (16. ai.4 1. For Kd < 1. for a damping ratio of unity.0 2. we obtain the following: (AQo)m AQi = = = 1.

and b.2 As shown by Figure 16. = constant.0 psig. then from Table 15. 2. our studies show that it rarely permits excessive deviations of level. then: a. if its input goes down to 3. This is so because manipulated flow changes more rapidly so that level moves away from the top or bottom of the tank more quickly.9 for a 50:l equal-percentage value (kEp = In a where Referring to equation (16.4 whenQ. we see that if a controller is set up correctly with 5 = 1 a t e = Gs. t Q / d 8 . with an equal-percentage installed flow characteristic.0 psig when the input (level transmitter signal) is 9. the high-level gain 2 auto override is so biased that its output is 3. When the level becomes too h h or too g low. Then its output is 15. = 2Gs Thus we see that. Nonlinear PI controllers for electronic^. If the valve has an equal-percentage installed characteristic. reference 1: @o = e12 o kEP Since kEp a = 50): = 3. and then only by a small amount. = (eFs/2). Effect of Installed Valve F o Characteristic lw If the PI level control system (or proportional-only) is not a cascade levelflow system. is the flow through the valve in its wide-open position. As the level deviates signhcantly from set p i n & Kd increases and TR decreases. Auto overrides for pneumatics. 5 = 1.0 psig. the relative stability is decreased at low flow and increased at high flow. its output is 15 psig. Although in theory this is not quite as foolproof as auto overrides.^ A preferred version has long reset time and a s d ICd in the vicinity of the set point. .0 psig (zero level).3.0 psig. Then control-loop dynamics would be independent of flow rate and where (ep). that is.16.2 Level Control ofsimple Vmek 383 1.0 psig when its input is 15 psig. 4 = 0.or low-selector. a propomonal-only controller (usually a fixed-gain relay) takes over through a high.1. Correspondingly the low-level gain 2 auto override is so biased that when the input is 9.14). then it is desirable to have a control valve with a linear installed flow characteristic. For this case it is best to find controller settings for 5 = 1 at the minimum expected flow.7 whenQ. its output is also 3. This design does not provide flow smoothing quite as good as that of the PI plus override scheme.

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

2. Let [&I. and to determine [K&]d=. [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. the auto override will drive the system back into the l n a . To find out which.16. If choose [TRIdesign = [TRImax and find . = 0. In this regime it is not absolutely necessary that control be stable. On the basis of experience.’ 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..stable r e p e . The slope of this curve at average rate For a cascade level-flow system. of that controller. with permission from production supervision. the control-valve installed-flow characteristic should also be linear. For some applications a much larger value will be desirable. the following procedure is suggested: 1.2 Level Control o Simple Vessels f 385 If pump and valve curves are available.gn or [TR]d=ign. the e n p e e r will be constrained by either [&. we can usually calculate the installed flow characteristic.25 (see reference 10). This is larger than the minimum available gain of any commercial controller with whch we are familiar. the system temporarily will be under control of the proportional-only auto override.Imin or [TR]. 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. and if the hydraulic resistance and static heads of the process equipment are known. we can experimentally make a plot of valve loading signal versus is aQJaYlC. flow. Alternatively. we suggest that it not be less than one minute.

386 Liquid Level Control 3. Ibm/min cp = reflux specific heat. For proportional-only control. one should make TQ I2 minutes. discusses some of the practical details. pcu/lbm "C To = vapor condensing temperature. "C A = vapor latent heat of condensation. for a PI controller. The amount of vapor condensed in the column by cold reflux is: where wR = external reflux flow.4 LEVEL CONTROL OF OVERHEAD CONDENSER RECEIVER VIA REFLUX MANIPULATION For this application it is necessary to take reflux subcooling into account. pcu/lbm. "C T R = external reflux temperature. Section 9. 16. Chapter 3. It is also more convenient to use weight units.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. If choose and find 4. If and 16. one should make TH 1 2 minutes. of process vapor .

4 Level control of Overhead Gmhmer Receiver Pia Repm Manipulatum r 1 387 Then (16. We may now prepare the signal flow diagram of Figure 16.23) + l PLAT Let us now define: (16.1(s) - [ 1 + (To .TR) X 1 I 1 (16. Ibm/min = external reflux density.16.4 for propomonalonly level control. By malung TH 2 5 minutes.22) where wtwc WD = = pL AT vapor t o top tray. Ibm/min. 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. The larger TH or TQ is.24) Then (16. we can usually ignore condenser lag.21) (16.20) (16. By inspection we can see that: fP&) %(S) 1 wt. also rate of condensation = top-product flow. lbm/ft3 = cross-sectional area of tank.25) where . the less signhcant will be condenser lag. Ibm/min vaporfim top tray to condenser.

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

For most applications one should make rH 2 10 minutes.28) and (16. and cp = 0. 16. Some idea . Ibm/min. If the feed enters at its boiling point.26) 16.6 Column-Bme Level Control via Feed F h Manipulatirm 389 Note that rH contains Ksc and is inversely proportional to it.2 must be extended to take into account the lags between the column feed point and the column base. lbm/min wl = .5. flow due to reflux or feed change wV = boilup. Some of the practical details are discussed in Chapter 4. from lowest downcomer into column base.16. It has been shown elsewhere (Chapter 12 of reference 1) that a number of equal lags may be approximated by dead time: 1 = . 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).29) where flow.the magnitude of Ksc may be obtained by assuming as an example To of ' = 100°C. A = 100 pcu/lb.5 pcu/lb"C. Typical values are in the range of 3-8 seconds. Then Ksc = 1.6 COLUMN-BASE LEVEL CONTROL VIA FEED FLOW MANIPULATION For this case the simple analysis of Section 16. Ibm/min w B = bottom product. T Subcooling therefore can have a signhcant effect on TH.27) (rm5 + 1)" Such an approximation simplifies considerably either hand or computer calculations. then wl(5) = e-nTmswF(5) (16.-n +m> (16.5 COLUMN-BASE LEVEL CONTROL VIA BOITOM-PRODUCT MANIPULATION This is usually a fairly straightforward system to design since it fits the analysis of Section 16.2. For PI level control we can readily show that: (16.

Recently we have done some work (unpublished) that suggests that incorporation of a Smith predictors.25 =aoR (16. may be represented by the signal flow diagram of Figure 16.25/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 16.33) (0.32a) where u. The levels in the overhead condensate receiver and in the column . rad/& Equation (16.32) and U . level control cascaded to feed flow control. this measurement is ambiguous to some extent.[w&) + tPg(31 PL AB5 Let us next define: a = tarm (16. we will simply state that. For PI control choose: 1 = 20a T R Z 5 x (16.2. By inspection we can see that: wds) .1 KIK.32) defines a minimum value of rH.390 Liquid Level control The equations for the remainder of the system now follow the analysis of Section 16. 0.6. nb J = K L (16. dimensionless (wR) = closed-loop resonant frequency. by means of Bode and Nichols plots.9 would help greatly. 16.for greater stability and filtering one may use larger values of rH. = closed-loop resonant frequency. The complete system.7 COLUMN-BASE LEVEL CONTROL CASCADED TO STEAM FLOW CONTROL Ordinarily it is satisfactory to use liquid level as a measure of inventory.31) Without going through the mathematics. we found that for well-damped response (Mp = 2 db): a =-H= PLABKWF.5. In the case of distillation columns. however.

16. m a c = I E b: g e m L c . 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 .7 Column-Bme Level G m C m d d to Steam F h Conml m l 391 c 0 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 .

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

Ibm/min wv = vapor boilup.38) .8we can see that for stability with proportional-only level control we must have: (16. Overall Control System If we are not limited in column-base holdup and can design for reasonably well-damped control. A temporary decrease in liquid flow down the column.This may be partially reduced to the form of Figure 16. We may then prepare the signal flow diagram of Figure 16.394 Liquid Level Control where w1 wm wB liquid fi-om last downcomer due to reflux or feed. the mathematical relationship was shown to be: (16.8. 3. an i w e a s e in vapor flow up the column may cause the following: 1. For either inverse or &rect response. 2. then we can treat reboiler dynamics as negligible. This says that steam flow responds to the flow controller set point immediately. This is so because the reflux flow momentarily exceeds the boilup. Ibm/min = = = Inverse Response As pointed out in Chapter 13.Note that Ke = steam flow-meter gain = 12/(wfl)-. there is inverse response for base level as well as for base composition. which we term “direct)’ response. and that boilup follows steam flow without lag. This we call “neutral” response. No change in liquid downflow. Note that when K * / T ~is both positive and greater than unity. 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. A temporary zwease in liquid flow down the column.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.7. which is called “inverse” response. Ibm/min bottom product flow. Ibm/min liquid downflow due to inverse or direct response. A.

39) that P ( t ) will have a slight offset from set point.(t)l At + ~ (. .(t .t 1) (16.n T ~ ) .8 becomes: An examination of the mathematics suggests that for stability we should make A B ."8 It is analogous to the Smith predictor for dead-time compensation. The impulse function shown in Figure 16.' The control loop containing the predictor is shown in Figure 16. we can design a compensator for inverse response called an "inverse response predictor. allowing us to tune the controller as though inverse response were not present./& > p L B. 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 ~ ~ . The time-domain equation for the predictor is: ~ ( t )K~ = p.KV1> kVl Kmh and 7 R A.9 after the predictor eliminates this. when holdup is adequately large.e.39) where It can be shown by equation (16.7 Column-Base Level Control Cascaded to Steam Fknv Control 395 The term b92 is normally both negative and small. A common approach to design. 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. . kVl is often large at low heat loads.H TTR) . If a PI controller is used instead of a proportional-only controller. The predictor contains a model of the inverse response and cancels its effect from the measured variable. If the control loop has to be tightly tuned because holdup is small.16. the denominator of the upper-right-hand box of Figure 16. 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.9. we have inverse response and the effect is unstablizing. Inverse Response If Km is positive. But as pointed out in Chapter 4.

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 % .

7 Column-Base Level Control Cascaded to Steam Flow Control 397 2 r 2 3 c m IC 0 C as3 m11 -2 e. gg Ln. 0 .16.

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

8 COLUMN-BASE LEVEL CONTROL VIA CONDENSATE THROTTLING FROM A FLOODED REBOILER (CASCADE LEVEL-FLOW CONTROL) As a final example.e-nTrnJ (16.)max steam condensate flow. We will ignore possible column inverse response but will take swell into account. .1 . We need in addition the following relationships: (16.+ = = wo w. The basic. open-loop signal flow diagram for the flooded reboiler is given in Figure 15. let us look at the control of column-base level by throttling condensate flow fiom the reboiler.41) (16. lbm/min liquid volume between level taps.8 Column-Base Level Control via Cmhmate Tbrottltng 399 If K m / ~ m u i y or close to it. seen fiom the base. 16.40) The system now behaves as though it has dead time. Ibm/min fi3 liquid swell coefficient. Ibm/min boilup. will be assumed to be resistive only.8 is n t becomes simplified: K m (1 T m . wu VB pL = = = = = wl = kVl = ~~1 = steam condensate flow-meter gain 12/(tt7. ft3 liquid density. lbm/min boilup column acoustic resistance.13. the inverse response term in Figure 16. 1bf&/fi5.16.44) where K. lbm/fi3 liquid downflow fiom column.Ibm/min rate of steam condensation.42) (16. Column acoustic impedance. It is assumed that column impedance is fast enough to reduce to a constant.

11 may now be accomplished: kv1 1 --= PLS kvl P L J - 1 (16.46) PL5 Next.47) From equation (15.49) so (16. lbf/fi? process vapor density. it is neglected in this analysis. the relationship between heat load and tube vapor volume is a straight line.50) (16. Ibm/fi3 For a small number of flooded reboilers examined to date.400 PB Liqutd Level Control = pBu = column-base pressure. Some consolidation of terms of Figure 16.10 may now be prepared.51) (16.11.45) (16.27a): (16. let: (16. that is. It may be reduced readily to the form of Figure 16. the kvl term is essentially constant from low to high heat loads.52) . The compression effect due to process-side liquid-elevation changes has not been checked but is believed to be small.48) (16. The preliminary signal flow diagram of Figure 16.

62) .11 and equations (16. where: (16.61) (2 kR KL = + 1) k 2 d 71 m (16.8 Column-Base Level Control via Cma!ensate T r t l n botig 401 From Figure 16.54) or or (16.58) For 5 = 1: 4K~72 1 = Now let - 2KL71 + K Z < (16.56) It can be seen fiom equation (16.48) and (16.57) or (16.56) that for stability we must make KL 71 6 1. the damping ratio should be at least one.59) (16. But this is not enough for a practical design.60) (16.16.52) we can see that the characteristic equation for this system is: For proportional-only control this reduces to: (16.

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 && . al 3 8 m w I 2 E +.

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

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

more liquid flashes.. If the vapor flow must be throttled to each load.and downstream pressures help good flow control. and perhaps hrnishes heat to other loads. usuallv can be of the "averaging" type. In this section we will see how vaporization rate. And the temperature difference across the column is not too large. Column A' is small compared with absolute pressure. 1 7 . 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. The former might result in flooding and the latter in dumping. is A simplified analysis can be made if: 1. the reverse happens-the rate of boiling is decreased and pressure rises at a slower rate. a rapid change in pressure can cause flashing or cessation of boiling in the column. 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. as mentioned earlier. we should use these models with caution. w. If pressure starts to rise. constant up. are usually adequate. For well-damped pressure control.. and for some of them it is undesirable. the mathematical models in this chapter. Latent heats at the top and bottom of the column are nearly the same. 3.1 7 C I Pressure and AP Control 17.5-2 cpm. which tends to reduce the rate at which pressure falls.1 INTRODUCTION olumn Dressure control. Most of the tight column pressure controls we have studied have closed-loop resonant fi-equencies in the range of 0. which mostly treat condenser and reboiler dynamics as negligible. For tight pressure contol. l 2. If pressure starts to drop. affected by heat storage. 405 . Most columns do not need "tight" pressure control. For the upper value one should make at least a rough check of condenser and reboiler dynamics.

pcu/lb UR = reboiler heat-transfer coefficient. and if we assume that steam will be flow or flow-ratio controlled. we may need to account for reboiler dynamics to calculate qT. Referring to Figure 15. This.wc(s) .3. Tcp.8. may be reduced to the form of Figure 17. actual fi3/sec . we may combine the above equations as shown in the preliminary signal flow diagram of Figure 17.2) Reboiler with Steam Flow or Flow-Ratio Controlled If we treat reboiler dynamics as negligible. = reboiler hot-side acoustic capacitance. fi? Pa = reboiler hot-side pressure. as shown by Figure 17.406 Pressure and AP Control These assumptions permit us. We may now write a heat-balance equation similar to equation (15. = heating-medium latent heat. fi5/lbf A R = reboiler heat-transfer area. pa = heating-medium (vapor) density. in turn.4 where: C.C p [ ( W c + R ) + sJK011 T&) + c~T'wR(s) cp@RTR(s) . pcu/sec "C fi? .and an average pressure. to lump the column contents together and to assume an average temperature. Reboiler with Direct Throttled Steam For the case where the steam valve is manipulated by some variable other than flow or flow ratio.1.30) but with terms for reflux added: CPTFFwF(S) = + CP@FTF(S)+ qdJ).wB(s)] (17. we may make a partial signal flow diagram as shown on Figure 17.2. at least for annospheric or pressurized columns. lbf/fi? Q.~ p T ~ p f ~ .1) + g ~ ) ~l(s) ( A p + C p T p ) %(-r) 1 -[wF(s) + 5 The material-balance equation is: wcol(s) = wR(s) . Pep. = heating-medium flow. Note that the reboiler is lumped with the column base but the condenser and overhead receiver are not lumped with the column top.(~ p ~ T ~ p W ~(17. lbm/ft3 A .

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

we often control pressure in a column by a pressure-dividingnetwork such as shown in Figure 3.6. note that aQ.408 Pressure and AP Control This may be reduced to the form of Figure 17.3 and 17.2 Preliminary signal flow diagram for column heat storage dynamics . 17.3 PRESSURE CONTROL VIA VENT AND INERT GAS VALVES As mentioned in Chapter 3. (17.5 where: URA..7.5 may now be combined into the signal flow diagram of Figure 17. Figures 17. = 0.3) P(4 = URARX aTaIaPa 1+LPa c R s .aQv/apa For critical flow. FIGURE 17. The control-valve inputsignal spans are usually the same but one valve opens while the other closes./aP.

3 Pressure Control v a Vent and Inert Gas Values i 409 m g U E z - i? e rc B 0 a . E m m U E bz m e F m v - E 8 5 e.17. s WZ .

3 and 17.410 Pressure and AI? Control FIGURE 17.6 Combined signal f o diagram for figures 17.5 lw .

7.3 into the signal flow diagram of Figure 17.G.8 and 17. K.10 where: (17. dwIG/aO.G. lbf/fi? PR = pressure downstream of vent valve. lbm/sec Pq = column pressure. Note that the valve gains. and the condenser. lbm/sec we = vent flow. The term C. awIG/d6. Since in most cases the inert gas bleed and the vent flow are fairly small. are both assumed to be positive.and awe/aO.(s). Together with the first-order dynamics. = controller output signal These equations may be combined into the partial signal flow diagram of Figure 17. reverse action of the inert gas valve is obtained by the .1 term. K. tend to be small. the equation for the vent valve is: (17. lbf/@ 6. and the addition of the pressure measurement.and aw.17.7 to the form of Figure 17../aP... vapor line to condenser.5) where WIG = inert gas flow. this commonly leads to large controller gains (small proportional bands) and control valve saturation for fairly small disturbances. For the case where reboiler steam is flow or flow-ratio controlled we can now combine Figures 17.6) From equation (17. we may reduce Figure 17.(s).9 can be reduced to the form of Figure 17./aO. Ibf/ft2 PIG = inert gas supply pressure. and the controller. Figure 17. Note the pressure feedback on w.8.6) we can see that open-loop pressure dynamics are essentially first order. If we can assume that PIG and PR are sufficiently constant. through dT. . is the acoustic capacitance of the column.4) Similarly.3 Pressure Control via Vent and Inert Gas Valves 411 The equation for the inert gas valve is: (1 7.9. the valve gains.


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




or nearly so. Z&).13) and (17. Mathematically the entire network may be studied by the methods of transmission-line analysis. and the capacitance is the acoustic capacitance of the space between the trays.(s). Hamett. Two cases are now of primary interest. is simply P&)/Q. respectively. and 1 is the total number of trays. the resistance is that of the tray and layer of liquid to vapor flow.(s) tanh nl + ZT(s)tanh nl + Z&) (17. This means that we treat the acoustic impedance Z .16) As can be seen. The terminal impedance. then the impedance looking up from the reboiler is approximately: Z+) Z. If ZT = 0. however. between the bottom and top trays as a pure resistance approximately equal to 2 (2 z. then we must treat the impedance as that of an RC chain as shown in Figure 17. . Here each RC section represents one tray.15) -2 R C r2 1 4 2 15 2 -. If the individual RC sections are equivalent. as would be true for an atmospheric column or for a column with tight overhead pressure control. the impedance becomes IR at low fiequencies.11) where (17. and Rose2 If.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.) =+_- where the subscripts s and Y refer to the stripping and rectification sections of the column. 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.16.*.12) ( 17. for each tray and vapor space. . ] (17. we are interested in the high-frequency behavior of the column. then: ( 17.14) where R and C are the resistance and capacitance.

6 Column AI? Control Pia Heat to Reb& 423 FIGURE 17.17.16 Equivalent network for vapor f o and pressures in column lw .

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

17. 2 C 8 A 2 VI 2 E $ a n m r W E 5z sa L O !5E .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 .

as shown by Figure 17. Bremer. and A. . REFERENCES 1. Characteri&qJJ edited by A. Academic Press. Wlim. Note that for constant top pressure control. the secondary or slave loop. and A.8 and add the AP controller as shown in Figure 17....18. D.426 Prmure and AP control The other case of interest is that of a tall column (many trays). Paper W-2-58.T. Day.3). the flow loop must have a linear flow meter or an orifice meter followed by a square root extractor. Eqg. the AP controller is connected directly to the steam valve positioner. This calls for another rearrangement of Figure 15. Wilmington.17. B. 48(6): 1957. L. J. I d . Hamett. ila s Young. columnbase pressure control and AP control are equivalent.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. Af Control Via Direct Steam Valve Manipulation If.J. B. We may now simpMy and rearrange Figure 15. R. T. 3.. Del. is tuned to be much faster than the primary loop. ISA 1008-1019 (1956). we need a somewhat different analysis.8. Chem. New Yo& Rose.. more rigorous treatment of this subject will be found in Day. 2. as is usually the case. Note that P(s) is defined by equation (17. Stanton. Then at high fiequencies: (17. “Plant and ProcessDynamic 1958. As in the case of level control cascaded to flow control.17) A more detailed.

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

-dt and (18.4) aLn L ( t )21 -Mn(t) aMn (18.v.Lnxn d + vn-lj’n-1- vnj’n (18.1 Flows t o and from basic tray .j’n-1 j’n .3) The next two equations concern the total tray material balance: dM.j’n-1 and the vapor-liquid equilibrium equation: E = j’n * (18. = 1 + (a .5) FIGURE 18.1) Then we have the Murphree tray efficiency: .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 . .L.Ln+l.l)xn + vn-l. axn (18.2) y.

2) gives: yn = Ey: + (1 .(S)= V(s) (18. we can write: (18..7) A rearrangement of equation (18.12) to give : .1)ZJ2 ( 18.6). [l and (a . so: V&) = V.12) We can now combine equations (18. (18.l)xn + (1 . we get: yn = 1 (18.9) leads to: aY.._ ax...8) + (a . (18.E)yn-l Ea - Eax.11) As discussed in Chapter 17. and (18.3).10) (18.18.9) Partial differentiation 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.E)yn-l Substituting for yi from equation (18.11). 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. and yn.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.-. (18.(for any tray efficiency less than unity). is a h c t i o n of both x.7).

13) v.8).1 ( ~ ) n For most trays Vne1 = V. The transmissions V(s)and y.4) and (18. . (18. But.. = and so the last term cancels out. = active area of the tray.17) Typical values of rm calculated by the authors for sieve tray columns are in the range of 2.5) may be transformed and combined: - + + on-1 (vn-1 .E) V(J) ..Vn)y.14) may be combined (after some reduction) into the signal flow diagram of Figure 18. Finally equations (18. 1 - 1 (18. (18.15) where A.. the derivation of rm is valid only if the column does not have sigmficant inverse response.. as mentioned previously.5-8 seconds. and (18.l2 (18. which is usually the column cross-sectional area minus the area of two downcomers. fi3/sec fi By using the approximate form of the Francis weir formula: Q = kf$.9).2 for a basic tray.W L ( 5 ) + E + 1xn + 1 ( ~ )+ TnEyn .-%JMn(J) . ft2 change in flow over the outlet weir per change in height over the weir.14) Harriott’ has shown that: (18. the transmissions L.16) we obtain: ( 18. Next equations (18. In this way a signal flow diagram for any number of trays can be prepared.-l(s) (18.(s) go to the tray above.(s) and xn(5) go to the tray below.13).430 Composition Dynumiu-Binu~ D & M . although the feed tray and two terminal trays have slightly modified diagrams.

2 Bmu Tray l?ynam# 43 1 + P . 1 V 0 0 e E E b P N= .18.0 6 9 3: sg Giii . .

7): f. we obtain: (18.26) Next the total material balance for the tray is: a34 f = E: + Lf+. the feed flow. by partial differentiation.24) where zF = low boiler mole fraction in feed. (18.18. .18) (18.22) and (18.19) into (18.Lfxf + Vf-1Yf-1 .25) .Lf + Vf-1 dt - v .21) (18.27) .VflfN .20) from which.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 .3 FEED TRAY DYNAMICS The feed tray differs from the basic tray in two ways: (1) it has an additional flow.18) and rearranging gives us: Vf= F ( l .19) Substituting equation (18.q) + Vf-1 (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.r) (18.YfVf(4 We can write an equation for y(F analogous to equation (18. Let us first write an equation for the summation of steady-state flows at the feed tray: ( 18.VfYf (18. 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(. When linearized and Laplace transformed.) (18.

29) .2) if we make the simplifying assumption that the reflux. it is usually possible to control internal reflux rate.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. so the reflux enthalpy does not change significantly.28) Note that From equations (18. subcooling has only a small effect on reflux enthalpy. q is approximately the heat necessary to vaporize 1 pound mole of feed divided by the molar latent heat of vaporization of the feed. We assume further that the vapor from the top tray is totally condensed. Note that q is a measure of the thermal condition of the feed.1).3. On a McCabe-Thiele diagram (see Section 2.20) through (18. the slope of the so-called q line is q / ( q . We can then write the following transfer function relating the vapor composition yr and reflux composition xR: e .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. This we are doing more frequently today.%(S) TDs + 1 where al = vapor-flow transport delay from top tray to condensate receiver.4). Lo. enters the top tray at its boiling point. in most cases leads to small errors in the calculated values of static gains and top-tray mixing time constant. the intercept on the 45" line is always zF. but if the reflux is subcooled only a few degrees. minutes e . This is not usually true. minutes TD = = mixing time constant of well-mixed condensate receiver. minutes liquid-flow transport delay from condensate receiver back to top tray. therefore. the reflux temperature is sometimes controlled.(a1 + a2 )r -xR(S) (18. Then LR = Lo. 18. For those cases where it is not practical to control reflux temperature.28) we can prepare the signal flow diagram of Figure 18. Ignoring subcooling. Further.

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

-(Ul+U2)5 ax.33) This is the mixing time constant of the top tray. Note that: .34) (18.4 may now be prepared. The term LR($)/6$) relates reflux flow to controller output signal. If dead time is negligible.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. or the ratio in a ratiocontrol system.31) where (18. then we may write: (18. This has been shown experimentally by Aikman to be true of a plant column. P(s) 1 = zRay' 1 (TD$ .4 Top-Tray and Overhead System Compostth Dynamia 435 The signal flow diagram of Figure 18. Note that a composition control loop is also shown.5. 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. (18. The mathematical explanation is simple. Then: (18.l is the mol fiaction of low boiler in the vapor leaving the tray below the top tray.35) . By collapsing the overhead composition loop and the two transmissions for Y .32) and rT = m (18. the set point to the secondary flow controller. then the transmission xR(s)is broken and there is no feedback of xR(s)down the column.18.~ ( S )we are led to the signal flow diagram of Figure 18. Note that y . for the moment there are no implications as to whether the manipulation is simply a valve.

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 .

18.4 Top-Tray and Overhead System Composition W a m h 437 E E x 9 E 8 ! i U 5 ) I c. 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. .

it should be noted that the first procedure-feedforward control-is more desirable than the use of a large condensate receiver. Therefore.37) (18.35) may now be factored into two firstorder lags such that the larger is: (18. Although we have indicated two methods of breaking the xR(s) thereby simp-g composition control system design and improving the control of?&). It can be shown that blending is much more effective if done outside the composition control loop.K ) is large enough. transmission.39) and the smaller is: (18. .Otherwise conditional stability might occur if 1/(1 . 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).36) and (18.38) The denominator of equation (18.438 ConapwitMn Dynunaia-Binuty DktiUahn where (18. Actually the control of top-tray vapor composition is not as important as the control of composition of condensate to the next step. if the resonant frequency of the closed-loop composition control system is well above l/rD.40) Substituting back into equation (18. we get: (18.42) At frequencies below (1 . then there is in effect no transmission of xR.K ) while at frequencies above l / r Dit has a gain of unity.35). As a practical design consideration.[ p ( j v ) ]has a gain of 1/( 1 .K)/rD.

This momentary change of concentration in the wrong direction gves rise to the term “inverse response. too. Usually. This means that transport times in the vapor and reflux lines should be as small as possible. in the short run it increases it. the reboiler has considerably more holdup than a typical tray. 18. The chief difference is that the reboiler has no entering vapor flow from a lower tray. should have as small a holdup as possible.2-18. if unagitated.(s)is the transfer function relating vapor flow to composition controller output.1). is that of the reboiler itself plus that of the column base. This is due to a decrease in foam density on the trays and a consequent momentary overflow into the downcomer. Therefore.6.4 and 18. It is assumed that steam flow is controlled by a steam flow controller.6 Inverse Response 439 Finally it should be noted from equation (18.6.” The signdicance of inverse response in the design of feedforward compensators and feedback controllers for composition is discussed later. A composition control system is shown in which V(s)/O. The holdup. 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. As may be seen from Figure 18.18. the discussion in Chapter 3.6 INVERSE RESPONSE The phenomenon of inverse response was apparently first noted by Rijnsdorp? As mentioned in Chapter 13. ~ .30) that it is quite important to keep the dead time in the overhead system to a minimum. See. 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 . however. 18. MB. The condensate receiver. an increase of liquid flow onto a tray increases low boiler concentration. the reboiler/column-base composition dynamics are essentially those of a simplified basic tray (see Figure 18. 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. L1 and x1 are the liquid flow and its composition from the lowest tray. We therefore may prepare the signal flow diagram of 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 .

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.. 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. overhead system. Symbolism such as .7 Overall Composition Dynamirs 441 18. and bottom system. and (2) by preparing all of the column equations in matrix form. and even when the shorthand notation of signal flow graphs is used. For this reason more recent studies have usually been based on matrix methods. one can combine the various equations affecting column composition dynamics and thereby calculate overall column behavior. Forcing functions in the original Rippin and Lamb equations were limited to LR. Binary Distillation-Rippin and Lamb Model For a binary distillation column. V. The signal flow diagram approach to deriving individual transfer functions is exceedingly tedious even for a short column. and F . zF. the authors have added terms for p and q. or a third method-simulation on an analog or digital computer.18. From the material previously presented.43) where p is the top pressure in psia and where the individual transfer functions are determined by one of the methods indicated above. convenient. 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.' A small amount of work has been done with hybrid computers. the "stepping" technique.7 OVERALL COMPOSITION DYNAMICS In previous sections we examined in some d e t d the composition dynamics of the feed tray.

Applymg asymptote techniques to the Bode plots. Rijnsdorp. Equations . More comor~Iy today the individual tray differential equations are combined and solved in the time domain. however. A comprehensive review of the literature of distillation dynamics and control through about 1974 is given by Rademaker.43) and ( 1 8 . The advent. In the meantime we should determine terminal volumes by material-balance and protective-control calculations before calculating column composition dynamics. 12-14 It should also be noted that the open-loop transfer functions given in equations (18. accuracy falls off.43) and ( 1 8 . for only one combination of holdup volumes in the condensate receiver and the column base. they determined approximate Laplace transforms. Feedback control for top and bottom composition could then be designed with ease. predicted behavior checks with that of more rigorous models. Other Models Two other models for binary columns are those proposed by Wahl and Harriott16 and by Waller. it is shown that the former is much faster. For large columns in particular.4-42 Cmpacitim DynamiLs-Binuty DirtillutiOn means the open-loop transfer function of yT with respect to 2. .” These are similar in that they are both based on circulation rates and liquid holdup in the column. nonideal separations. Low-order transfer functions are calculated &om steady-state data. suggest that if these holdups are no larger than the total holdup on the trays. 4 ) are valid.7. 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. For columns with moderate relative volatitities-say 2 to 5-and for terminal purities not greater than 98 to 99 percent. of more p o w e m and less expensive digital computers has shifted the emphasis to digital simulation. It should be noted that some of the work of modeling column composition dynamics has been concerned with multicomponent rather than binary separations. (18. later extended by Luyben and others. But for highpurity. and Maarleveld. Rippin and Lamb developed a stepping procedure. 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.18 Tolliver and Waggoner” have published an exhaustive review of more recent additions to the literature. then the open-loop transfer functions are relatively insensitive to variations in terminal volumes. 4 ) may now be represented in signal flow diagram form as shown in Figure 18. More work needs to be done to check the range of validity of these models.sL14 computing these transfer functions in the frequency domain for with a digital computer. This topic definitely needs more study. Studies to date. strictly spealung. The two models give essentially the same results except that the Waller method leads to higher order-and intuitively more accurate-transfer functions. however.

I m U e 4 n L E t a a a 5 e e P t-= .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.

W. Rijnsdorp. W. 1980.” presented at American Institute of ChemicalEngineers Meeting. Luyben. J.. E. Fund. Anderson. Oct. 14. W. T. Tex. 80-508. 149 (1961). 2. Harriott.. Dm. L. E. Rothfus. Wiley.. W. Elsevier. L. and R.. E...” Fifth World Petroleum Congress. Harriott. E. En&. Vol. 11. 1-5.. Rademaker.. Rippin. M. London. ISA Trans. “Simple Models for Distillation Dynamics. 32. New York.... 57. T. Ph. Wahl. 12. R. L. 7. P. Chem.” Institute of Chemical Engineers Symposium Series no. 18. 5(4):386-394 (1966). “Comparison of Stepping and General Complex Matrix Inversion Techniques in Calculating the Frequency Response of Binary Distillation Columns. 1975.. Thesis. “Distillation Column Control. 4. Rijnsdorp.E. T. Aikman. 17. and A. 16. Nos. Eng.9(3):396-407 (1970). Maarleveld. Thesis. 19. “Dynamics and Control of Continuous Distillation Columns. E. M. D. Waller. Series. J. P. R. 6. T. and P. L. J. Process Cuntrol. presented at Houston. K.” ISA Paper C. Harper. 8. Tolliver.I..D. University of Delaware. 3. “Feedback and Feedforward Control of Distillation Columns with Inverse Response..” Bulletin. and R... F. Campbell. I. 36.. Lamb. I.P. Rothfus. pp. R.. 1964. 12:5-14 (1961). Shunta.. 1979. . H. T. L. Des. Sc. Tex.C. Houston. 1964. G:39. Kermode. C.E. Rijnsdorp. W. 7.” I. P r g . and J... Rademaker. 1960. D... 5. R. O. IEC Proc. Birmingham Univ. Proms Dynamics. Fund. 1958.. 59-79. P. L. 10.444 Cmpaitirm Dynamics-Binay Dktillutk REFERENCES 1.D.E. and D.. Teager. Thesis. 9. Apr. ISA Trans.. I. et al.C. New York. 1963.. Baber. Edwards. 13. D.. E. 8:838 (1969). A Review and Perspective from the C. University of Delaware. Ph. Cbem. and N. Cadman. 1969. McGrawHill. R. and R. and W. Luyben. L. T. W. 6(3):421-431 (1967). O. 1959. A. Carr. Cadman. Cadman. Kermode. Waggoner. 1955. University of Delaware. New York. E. Massachusetts Institute of Technology. ISA Paper 56-28-3. Symp. 15. Luyben. Dynamrczr and Control of Continuow DtitillatMn Columns. “A Theoretical Study of the Dynamics and Control of Binary Distillation. J. 1( 1):7481 (1967).

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

calculate mol fi-action low boiler. on a binary basis and relative volatility at each tray. Pick a light key and a heavy key-or a low boiler and a high boiler. Some workers have used the Smoker analytic method rather than the trayto-tray method. and boilup ratio. Using printouts of multicomponent design calculations. the second problem must be dealt with first. Use curve-fitting techniques to relate xu and relative volatility at each tray. As shown by Strangio and Treybal. How to derive a relationship between liquid composition on each tray and relative volatility. and so forth. Sometimes a multicomponent system . 19. Go through binary design calculations to check the total number of theoretical trays and feed-tray location. usually the design case. We feel that the advantages of the trayto-tray method outweigh the disadvantage of longer run times. we may calculate the effect of changing reflux ratio. The following procedure is suggested: 1. xu.and bottom-product drawoffs. Two fimdamental problems therefore conflont the engineer at the outset: 1.2 DESIGN PROCEDURE Since we will use the tray-to-tray method. and reads out any or all tray compositions. 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. it is appropriate to b e p by stating the required assumptions: 1. 4. the tray-totray method easily handles relative volatility as a function of composition. Such an approximation greatly reduces computation and time. 3. Binary or pseudobinary separation. 2. As is shown here. With these in hand. If the match with the multicomponent calculations is not adequate. Simple column with a single feed and top.446 C&&k o Stedy-State Gains f through the points. 2. The basic approach then consists of defining a base case. 2. However. Then the slope of either curve is the “gain” of top or bottom composition as a function of reflux ratio. reflux ratio. Most industrial columns are not binary and do not have constant relative volaulity fi-om top to bottom. go to step 3 and use a higher order function for curve fitting. surprisingly good accuracy can be obtained in many cases by approximating a multicomponent system as a binary system. to firm up feed-tray location (number of trays above and below the feed tray). How to make a reasonable binary approximation of a multicomponent column.

In the first programmable calculator programs we developed.3) This is they corresponding to the intercept of the operating and q lines. 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. 7. find: yn = 1 + (a . the base is the usual starting point. The tray-to-tray calculations come next. 6 . The design program (see also Chapter 2) starts with data provided by the column designer or plant.l)xn axn (19.1) (19. Find: (19. LR/D. It also includes top. For optimum design it is also the composition of vapor above the feed tray.4) This is the equation defining the vapor-liquid equilibrium relationship. 4.2 Dabn Procedure 447 can be modeled as one binary in the bottom section and as a different binary in the upper section. 3. it proved inadequate. One bottom product that is all liquid. composition. Feed that may be single-phase vapor or liquid. (2) a linear relationship between a and liquid composition x. it is customary to start at the top. Partial reboiler and total condenser. 1. three choices were available for relative volatility: (1) constant. But the first time we tried these on a real column for MeOH-H20. 3. 3. 5. If trays are numbered from the top down.1 Starting at the column base and working upward. and (3) a quadratic relationship between relative volatility and liquid composition on a tray.Running a design program is not absolutely necessary for our purposes.19. and enthalpy or q. Find: (19. . but it is one way of initializing subsequent programs and of checkmg assumptions. We will follow the second practice.and bottom-product compositions and reflux ratio.2) 2. Equimolar overflow. If trays are numbered from the bottom up. The following steps are involved. This includes feed flow rate. or may be mixed phase in nature. One top product that is all liquid.

4. it does not count as a stage.5) This is the material-balance equation and its initial application gives xl. we switch operating lines. and so on until y 2 y j . 19. 3. Note that the first x is xB. The first x calculated is that of the first tray above the feed tray. the value o f y used is that of the feed tray. The number of times that we perform 3.2 Find: (19.448 Calcuhbn of Steady-State Gains It is the x versus y curve for a McCabe-Thiele diagram.2 to calculate x2. 3.2 [equation (19. designated “Type A”: 1.2 Find: Xn+l = Pyn + XB P + l (19.1 and 4. 4.3 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. “exact” value of R that causes the column to produce top and bottom purities that exactly match specifications. Since the condenser is assumed to be a total condenser.5)] is one less.6) This is the material-balance equation for the rectification section. Then use yl in 3. .1 [equation (19. 2.2 untily 1xD (total condenser assumed). the number of stages there are in the stripping section. 4. This may be determined by the following procedure. and therefore represents the number of theoretical trays. The number of times 4.1 and 4. liquid composition on the bottom tray.2 have been repeated is NR.4)] is N.7) 4. Assume a new value of R slightly smaller than that used for design.1 Find: (19. using x1 to calculate yl. the number of stages in the rectification section. Starting with the tray above the feed tray.1. The number of times that we perform 3.. it produces bottom-product purity slightly better than specification).3 Repeat 3. A partial condenser would count as a stage.3 Repeat 4. including the feed tray.2). including feed tray and the reboiler (assuming there is one). What we need to do next is to find a slightly different. Calculate P from equation (19.

we multiply the “error” yT 7 xD by a constant with the proper polarity for negative feedback. Column terminal composition sensitivity to changes in feed rate. we need to know the change in top composition due to a change in distillate or reflux flow with other variables held constant.l)x.. Many different modes of column operation and control can be visualized.4) has been used equals ni 4. We have. 6. but for preparing programs. find: y n = 1+ ax. we must keep in mind at least four needs: 1. Then switch operating lines and find: (19.1. Compare last y = yr with specified xD. (19.4 COLUMN OPERATION PROCEDURE . used a simple “proportional controller. assume a slightly larger value of R and start again at step 1. 2.+ < x.6) yn = ax. for example.. Use they thus calculated for the feed tray.” that is. Here.4 Column Operation ProMdure 449 3. for example. Repeat equations (19. Originally we used 0. This proved to be too large for some purposes so we switched to lo-’. feed composition. 1 + (a . go back to step 1.7) utl they have been used Nr ni N s = NR times.0002 as a tolerance: [3”r . Starting at the column base and working upward. A number of convergence procedures are available for step 6. feed enthalpy. we must be carell to provide program logic to prevent “taking off to the left of the equilibrium curve.19. A better method is a modified interval-halving technique that provides moderately rapid convergence without instability.6) and (19. This reqmres some trial and error to achieve a balance between oscillation and excessively slow response..5) Repeat u t l the number of times equation (19.l)xn (19. (a .4) (19. As indicated by Figure 19.xD]. assume a new value of R. and repeat steps 2-6. If yr does not match within a preselected tolerance.7) If the calculated x. Column gains for feedback control. and column pressure (usually defined at column f N S 19. 5.

00000 . is too close to 1. or B. Guess for x.1 Effect on calculation of rectifling m i o n when: A. Guess for R is too small.450 Calcrrlation o Steady-State Gains f FIGURE 19.

boilup is fixed by flow control of steam or other heating medium. The first of these we have labeled ‘Type A” and the second we call ‘Type B. or (2) with fixed reflux or boilup. and . For example.e. we wish eventually to control column top composition by manipulating distillate flow. and reflux is varied by trial and error u t l the materialni balance equation and tray-to-tray calculations converge. steady-state accuracy is required. we use one of two types of programs: (1) Column terminal conditions are fixed at new. for most of these concerns. we change a terminal condition such as distillate flow and find the corresponding change in terminal compositions. Again.” Gains for Feedback Control case 1 Consider Figure 19. Let us assume that feed rate. and bottom-product flow is determined by column-base level control.2 where top-product flow is set by flow control. 4.4 Column O p e r h Pmmdure 451 overhead). the top-product rate. As shown by the dotted line. and boilup are fixed and that we wish to find the changes (i. The starting point is a set of “prep” equations that reflect the desired change in distillate.The new distillate flow is: D’=D+AD and B’ = = B - AD AD (since F is fxed at 1.19. column gains for feedforward compensation. if feed composition changes. Ability to estimate the cost penal? of producing excessively pure product by excessive boilup and reflux.00) p‘ LA R’ VJB’ = LR = LAID‘ 1. different values fiom those of design. These sensitivities ofien give good clues to the need for feedforward compensation or for terminal composition feedback control. Shinskey and Douglas and Seemann’ have been particularly interested in this. 3.. assume a new value of xD calculate a new x i : = x. With these equations in hand. what changes in boilup and reflux do we need to hold terminal compositions constant? Steady-state accuracy is required. AD. As it turns out. feed enthalpy. feed composition. “gains”) of top and bottom compositions in response to a change in D. reflux flow is set by condensate receiver level control.

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

This means that bottom product is also fixed. first to obtain converged values of xB and xD for the specified reflux and boilup. + AVf LA = L R + AVs p’ = V:/B R’ = LAID Then use the Type B routine.1 Calculate: xn+l = (R‘+ l)yn R’ ax. &. are therefore: V.Ns = NR. when the Type B program is to be used. Determine when the calculation is complete. and new top composition.l)xn and y. 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..4 Column Operation Promdure 453 4. Continue until N = NT . Since the original design calculations are unconverged. The m a p t u d e s of the interaction terms give clues to the need for decouplers. we have also calculated the change in xB in response to a change in D.The second gain. is another interaction term. it is necessary.19.2 If the first calculated x shows that xk < x k . 5.l (Figure 19. = 1+ 4.! = V. x i . assume a new and larger value of xB and resume at step 4. The “prep” equations for ultimately finding new base composition.l and repeat steps 1-5. we would start by assuming that both feed and top-product rates are fixed. we need the gain of xB with respect to V. Switch to r e d c a t i o n operating line. To control base composition via boilup. If bT . case 2 If we desire to control bottom composition by changing boilup as shown on Figure 19. We will term the procedure for steps 1-5 the Type B procedure in contrast to the Type A procedure. Note that changing B is equivalent for this case to c h a n p g D.x D ] is greater than the chosen tolerance. change in xD with respect to V. This second gain is an interaction term. that is. (a .. 4. .2. when the estimated xb I at the top tray. Note that we call the preceding Type B. y choose new x.1).

3.454 Calnclatron @Steadystate Gains ’ case3 Another commonly encountered control scheme is that of Figure 19. bottom product via base level control .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. This is similar to that of Figure 19. perhaps eventually cascaded fiom top composition control. I f f d and boilup are fixed at design conditions. while reflux is flow controlled.2 except that overhead condensate receiver level is controlled by throttling distillate.3 Distillate via reflux dNm level control. top and bottom composition &.x.

4 Column Operation Pmxdure 455 The Type B routine. we wish to find the responses x. described in the discussion of Case 1. If feed and bottom rates are fixed.. &/D’ The Type B routine. would then be required. case 5 A third basic control scheme is that of Figure 19. and if it is desired to find the responses xh.l to a change in reflux flow. . Bottomproduct flow is on flow controi. x.3. + AVs V:/B’ p’ D’ R’ =D = + AV. is then required.4. perhaps ultimately cascaded fiom overhead composition control. described in discussion of Case 1.19.l to changes in boilup. The following “prep” equations are required: LA = L + ALR R p‘ = V:/B R‘ = LAID The Type B routine. V: V. x. while reflux is flow controlled. perhaps ultimately cascaded fiom bottomproduct composition control. Note that if the original design calculations were done on the assumption that reflux is not subcooled.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. Base level is controlled by adjusting heating-medium flow control. Here overhead condensate receiver level is controlled by throttling distillate flow. described in the discussion of Case 1. the following “prep” equations are required: B’ =B = = AV. if feed and reflux are fixed. is required. the effect of subcooling may be considered as a change in LR: ALR = :(To .

we may wish to find the responses A&. feed and reflux are fixed. is required. described under Case 1.AB V. x. FIGURE 19.4 Distillate via reflux d u level control: boil up via base level control rm .456 Calculation ofsteady-State Gainr Case 6 If./B’ - p’ = D’ = D AB R’ = LJD’ The Type B routine.4. for Figure 19. The following “prep” equations are required: B’ = = B + AB V: V. .l to a change in bottom-product flow.

2. case 9 For the control system of Figure 19.B’ = R‘ LAID’ = In the first equation above.2. Since D is fixed: B’ B + AF - But our program requires that all flows be relative to F = 1.4 Column 0pera. The following “prep” equations are needed: B’ = B + qAF 1 + AF . Therefore let: B‘ = B+AF F+AF B+AF l+AF For simplicity of symbolism. we represent B” by B ’ .19.3. we may wish to find the responses xb and x i to changes in q.q)AF R = 1 . The following “prep” equations are needed: B’ = = = B + AF 1 + AF VJB’ p‘ LA D’ L + (1 . we may wish to determine overhead and base composition responses to changes in feed rate. The following “prep” equations are needed: q’ L A = = = q + Aq L .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”. we may wish to determine overhead and base composition responses to changes in feed rate. Case 8 For the control system of Figure 19.(Aq)F R LAID R’ Use the Type B program as described under Case 1.

For a change in F to F’ = F + AF: B’ But: = L R = + q(F + AF) L R - V.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. B‘ again might more properly be written B . B’ . for simplicity of symbolism. we may wish to find the responses of xD and xB to changes in q. A Type B program. we may wish to determine overhead and base composition responses to changes in feed rate. case 10 For the control system of Figure 19.3. case 11 For the control system of Figure 19.V. we use B’ instead of B .4. let us define: B” B” = qAF + B F + AF qAF + B 1 + AF = Again.B or qAF B’ = qAF +B Since OUT programs are based on 1 mole feed per unit time. should be used. discussed under Case 1. B and therefore: + qF = . The following “prep” equations are required: .

. The following “prep” equations are needed: F q‘ = q + Aq V. B.2. The variables chosen for feedforward compensation will depend on which feedback control scheme is used-Figure 19. We wish to find the changes in LR.2. in the Type B program (see Case 1).. 19. and V.4. This dormation could be used to design feedforward compensators to minimize transient changes in terminal compositions. Case 14 For any of the control schemes. D. case 12 For the control system of Figure 19. and C.19.3. bzF. 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..00 p’ = v:/B Case 13 Use the Type B program (see Case 1). rerunning the program for the quadratic coefficientsA. 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). ..! = V. or 19.+ qAF p’ D’ = = = V:/B’ 1 - B’ R’ LR/D’ Use the Type B program (see Case 1). required to hold compositions constant in the face of a feed composition change.4 Column Operation Procedure 459 v: = v. B. B.4assume that top composition and bottom composition are held constant. For any of the control schemes.4. + (Aq)F = 1. or 19. the effect of a column pressure change fiom P to P’ on xD and xB may be found by recalculating a’s. and C. we may wish to determine overhead and base composition responses to changes in feed enthalpy factor.3.. q. case 1s For any of the control systems discussed-Figure 19.19.and entering the new values of A.

case 19 If we have a situation similar to that of Case 18 except that both a new x.460 Calm- OfStedy-State Gaim The equations required to recalculate the other variables are: 2: = ZF + AZF F = 1.: . Case 16 This is the same as Case 15 except that q is changed. and C. and a new x.3. We will need to know LR and V. say xD. B. is set at a new. described in Section 19.. Case 17 This is the same as Case 15 except that column top pressure is changed to P = P AP. are chosen. we will get changes in D and B as well as LR and V...3). we may wish to estimate the cost penalty of the increase boilup.3). higher value. the other is readily calculated: X. Then if one new terminal composition is chosen.00 B=F-D' Use the Type A program. the other will also be fixed. constant.3). = XD + AxD Use the Type A program (see Section 19. reevaluated. to ensure that product purity is always at least as good as the original specification. If the column control scheme fixes either D or B. + Economic Penalty o Overrefluxing f Case 18 If product purity at one end of the column. The only equation needed is: q' = q Aq + Use the Type A program (see Section 19. Then run the Type A program (see Section 19. Relative volatilities should be recalculated and the quadratic ' coefficients A.

+ AV. 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.q ) l+R . 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).: R’=R+AR LA = = R’D’ LR LA or + D’ + D’ +D +D RID’ so: = RD B’ = = 1 .3).D’ p VJB’ Use the Type B procedure (see Case 1). the following “prep” Vi D’ But: = V.q) D’ v: (1 . the following “prep” equations apply for constant V. - + LA = VA = Vi (1 LA = q) RD’ so: D’ or + RD’= v: (1 .19. For a column with automatic control of R equations apply for constant R: = case 21 L/D.4 CoLumn Opera&n Procedure 461 Use the Type A program (see Section 19.

we ran a Type B program twice. V. dx.00./dD and d%B/dD. 19. We arbitrarily chose F = 1. D The gain is a very nonlinear function. One could use this information to design a suitable nonlinear or adaptive control scheme.98 0. 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. . decreasing as xD increases and as D decreases.7 = XB = q = a = 2. With this card.which is Case 1.80 R = 2. We then chose to find the column @IS.462 Calculation @Steady-State Gains Use the Type B procedure (see Case 1).50 N. and once for D slightly larger than design.was held constant. = 7 N T = 16 Our HP-97 program predicts: (NR = 9) p = 2.20 a = 2.53608.01 0.1.469 0.46392 and B = 0. It is also a function of F.306.001. A = 0. once for D slightly smaller than design.46 0. Although we chose small changes.80 a t x = 0. so for the design case. Mer running the design and exact R (Type A) cases. D = 0. we made a data card.9930 (xg)find = Next we found the exact R to be 2.d+Bx+C The results are tabulated in Table 19.20 atx = 0.5 EXAMPLES Example 1:Calculation o Column Gains f Let us use a test case where the following conditions apply: ZF = XD 0.

780 -0.950. that N. reflux is set by condensate receiver (reflux drum) level control. including the reboiler. as well as the tray-to-tray method.9520. For the design case.063 -1. dD -0.01000 0.87 They assumed a control system in which distillate is flow controlled.46492 c 0.575 XD XB = = = 0. and NR = 31.009273 dx 0 dD dx. based on a simplified Smoker equation. Overhead composition is controlled by boilup. Predicted overhead composition by our HP-97 program is 0. The exact R turns o'it to be 5. = 18 stages. Design conditions were: ZF = 0. D I xo x* 0. Table 19. The exact Smoker equation.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.65 trays.2 shows the results obtained with various values of xD greater than the design value of 0.010835 0. Douglas and Seemann calculated by their analytlcal method.98000 -1.98113 0.46292 0.670 0. = 21 stages and N.332 = 5.852.30 stages.00 q R a = 1.314 .950 0. predicts N. = 26.19.97875 -1. Again we made a data card at this point to minimize subsequent calculations. and bottom-product rate is controlled by column-base level.189 -0.050 1.46392 0.891 0.

4.. 5. V.33 31. Eng. “Effect of Control Point Location Upon column Control. Luyben. 1977. Cbm. McGraw-Hill.72 11.. Douglas. Seemann.442 4. Shinskey. Wood. 50(6) (June 1954).47 REFERENCES 1..98 11. E.050 XI3 0. Cox.193 4. 49-55 (June 1978).” presented at American Institute of Chemical Enpeers Meeting.261 0 5. 1968).464 C&ulatMn of Steady-State Gains TABLE 19. C.950 0. Ey.8521 6. Eng. J. Cbem.SEEMANN ANALYTICAL METHOD 0 1. P. L.6145 7. New York. M. S.0185 3.61 3. Des.043 5.955 0. Pmg. . Bauer.1834 8. Uitti.87 4.997 4. G.022 0. A.13 6.970 0. D o .93 5.S EEMANN ANALYTI CAL METHOD % ENERGY INCREASE DOUGLA S. 7. F. Buckley. Pny.029 0.65 REFLUX RATIO R=LR/o DOUGLAS. P. Treybal.960 0. 64(1) (Jan. Etg. and R.036 0. K. L.62 6. Chicago. 13( 3 ):279 ( 1974).12 19. I d .. R. 6..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. Cbem. Cbem.. 1976. Strangio. and C.Pmg. K.44 6. “Steady-State Control of Distillation Columns.1880 6. Dirtillation Control.965 0.774 5.70 6. and R. E... R. Pmc. 3. Orr. 2. and W.” ISA Paper 49-9-2.

Real-life industrial columns are often highly nonideal (nonlinear) and often must produce products with high purities-0.’ For simplicity‘s sake. 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. But particularly for composition control at both ends of the column. At least dual composition control was accomplished without decouplers. including those of modem control theory. for example) apparently are not badly affficted with interaction. Most of those reported in the literature (references 8 and 9. has not yet been published-and the reader will not find one here. we strongly recommend a simulation study using a nonlinear model to compare different control schemes.1 INTRODUCTION truly d e h t i v e treatment of composition control. 9 465 . The last consideration results in very nonlinear behavior. even for simple binary distillation. Tyreus” discusses multivariable control of an industrial column whose design was performed via a technique called the inverse Nyquist array. some columns are so equipped. we will assume linearity in this chapter to give some insight into column-composition control principles. 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.” These papers are largely aimed at multivariable control.20 a Composition Control-Binary Distillation 20. usually involving “modern control theory. 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.1 mol percent impurities or less. however. a laudable objective. Although most existing columns do not have composition control at both ends.

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. Overhead composition control cascaded to reflux flow control 3. Base level control cascaded to bottom-product flow control 4.1. 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). Condensate receiver level control cascaded to &stillate flow control 2.466 Composition C ~ t r o l . then for the system of Figure 20.1 we may prepare the closed-loop signal flow diagram of Figure 20. This permits us to ignore flow control loop dynamics. based on the Rippin and Lamb model.2 FEEDBACK CONTROL OF COMPOSITION In Chapter 18 overall composition dynamics.2 for: 1.1 Distillate via reflux drum level control b t o product via base level control otm . were represented graphically in Figure 18. FIGURE 20. reflux. If we assume that boilup. bottom-product flow. and top-product flow are the manipulated flows.

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.

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

3 Interadon Compens&km 469 si 8 U E E 3 a 3 2 - I t. EP.. 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. .20.

B i t ~ t Dtitdhtk y E p.470 COmparitiOn C ~ n t d . 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 .

5) into the large block of the top loop.6) Since the two loops are now decoupled.6 that shows that the two loops are now independent. xT. has the transfer function: LR/fpR KmGFT(s) = - -[I xB(s) OL Ke LR(s) (20. xB. we may prepare the partial signal flow diagram of Figure 20. By inspection we can see that the eftkt of reflux flow changes on bottom composition.4 may now be reduced by means of signal flow diagram transformation theorems to the form of Figure 20. may be canceled out if we make: In solving equation (20.4): (20.7) .5.5) We can also find the other compensator transfer function from equation (20.3) we find that the compensator. may be canceled out if we make: while the effect of vapor flow changes on top composition. KFTGFT(s). By substituting equation (20.3 Intmactiun COmpenratMn 471 The signal flow diagram of Figure 20. we get: (20.20.

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

20.3 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 .

0.0000036 2 : (20.6 becomes [equation (20.11) (20.1) while the base composition is controlled by adjusting steam flow.10) (20.8)] becomes: .0000097 = 21.0.000215 EloL= X 28s-k 1 The time constants are in minutes. From these we find that the large top block of Figure 20.5s + 1 . The following transfer functions were derived via the stepping technique: [%IoL p$IoL 0.13) Correspondingly.9) (20. Top composition is controlled by manipulating reflux (see Figure 20.000207 31.6 [Equation (20.6s 1 .7)] : [%IOL [IOL = (?e)( ) 0.0000062 21.474 Comparitiun Control-Binay Dictdmkn As an example let us consider a binary distillation column designed to separate water from nitric acid.12) 21. the large block in the bottom loop of Figure 20.6s 1) + (20.6s +1 [EloL + 0.

16) Equations (20.000 18 28. to make yT(s) = 0.1s + 1 (20.15) 1." but that offers limited benefits and is more complex to implement. and there will be no contribution fiom the feedback controllers. Since the column as built had a composition analyzer at only one end. Then. as shown by Figure 20.13) and (20. Later Luyben and Vinante13 tested decoupling experimentally on a semiworks column. benefits of decoupling were minimal.4 FEEDFORWARD COMPENSATION Feedforward compensators may be added for almost any disturbance.4 Feedjmvard G m t p e m h 475 = (e)( ) 0. that a PI controller with fairly high gain could be used. we can see by inspection of Figure 20. Luyben12 has discussed two approaches to decoupling. including one he terms "ideal. Physically each may be implemented. Niederlin~ki'~ Wallerl5 have also studied decoupling. Let us.20. If the compensators do a perfect job.14) indicate simpler process dynamics than probably exists in reality. as dscussed in Chapter 12. Individual feedforward compensator functions may be determined in a very simple fashion.7 that: . Since the column had little interaction. then yT(s) = 0 and xB(s) = 0.7.02 (20. and KFBGFB(S) 20. F. in any event. we had no opportunity to test operation with decouplers.14) The decouplers themselves become: (20. for example. But we would expect. 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. look at feedforward compensation for feedrate changes.

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 .

we can derive the following feedforward compensator functions: (20.18) and (20.23) (20. we may write: + Kf4Gf4(5) x a ["'I..20) By simultaneous solution of equations (20.21) and (20. to make Y ~ ( s ) = 0 and ~ ~ ( 5 ) 0 in = response to disturbances zF(s)and q(s).19) Similarly. to make xB(s) = 0.20.17) and (20.22) By a similar analysis we find that.24) .4 Feea'jcmard Compensdm 477 where (20.18) we find that (20.) (4 LR = 0 (20.

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

By the principle of superposition we may write two equations: (20.aYT at. Top composition controlled by reflux.29): A According to the theory: = A3 A4 (20.27) we get: -. 2. Finally. is constant. Condensate receiver level controlled by distillate. Base level controlled by bottom product withdrawal. there may be little preference between various control system Structures. Bottom composition controlled by boilup (usually steam flow). Relative-Gain Matrix for Binary Column-I Let us consider our normally preferred scheme: 1.29a) constant v. 4. if one is going to use decouplers.30) . AV. ReMve-Gain M a r k 05 479 implied by the literature. 3. See reference 6. in matrix form: The relative-gain array is related to the first term in brackets on the righthand side of equation (20.2 . as is entirely practical today.27) and (20. (20. = 0 and from equation (20. Some work has been done on defining a dynamic RGA. constant xs If V.28) whence.

33) Hence : (20. is constant. av. (20. then AX.28) we get: (20.32) L o r av.31) av.If x. = 0 and from equation (20.34a) (20. On substituting this back into equation (20.34b) L av. (20. J .34) Next: ay.27). we obtain: (20.

2. AX --AD+-AV. J = = 1- A1 1- A1 A4 = A1 = A3 = A4. ax. Base level controlled by bottom product withdrawal. We can see that each column and each row must add up to 1.36) (20. If AI zz A2 Relative-Gain Matrix for Binary Column-I1 This scheme is the one that Shinskey" terms "material-balance'' control: 1. Condensate receiver level controlled by reflux.38) .aD a5 v so: (20. interaction is severe. the theory says that: A2 A3 aL.5 Rehive-Gain M& 481 L Further.35) and ax. 3. Top composition controlled by distillate. The two starting equations are: (20.37) Here: (20. Bottom composition controlled by boilup (usually steam flow). 4. .20.

39a) (20.ax.- Next: (20. ax. Relative-Gain M t i for Binary Column-Ill arx The last scheme we consider here has the following features: 1. 2.39b) L av. . av. Top composition controlled by manipulation of LJD. Condensate receiver level controlled by distillate flow.

40) Then: ?YT AyT .aR ay. avs (20. + -AV.44) By a procedure similar to that used i the two previous sections.20. Bottom composition controlled by boilup (usually steam flow).5 RCld~e-Gain MatrrjC 483 3. + -AV.41) (20. 4.42) AX . Column-base level controlled by bottom-product withdrawal.AR .43) (20. once more: A = Here: & B ax. we find: n (20..45) Next: (20. First let: (20.45a) A2 = .aR whence. av.-AR . [: :1] (20.

The design specification for yT is 0. is 0. V versus y T (Figure 20.45~) Again: L aR J &mpfe. but this led to serious errors in calculating A’s.8).47 A3 = -1.47 2. _ .45) we find: & and A2 = 2. are as follows: a * avs aV S Hence: = -0. The h l results. LJD versus XB (Figure 20. At the specified operating conditions.1.0000422 - mEA m/hr mEA ax. Oripally lps these slopes were drawn by eye.10).000165 mfA.-0. As an example let us consider a binary system with low boilerA and high boiler B.000483 m/hr From equation (20.47 1 .484 (20.11). and V versus xB (Figure 20. determined by digital Merentiation.47 = so: .985710 mEA and for x. s o e were drawn on each of these curves to obtain the four gains. An HP-41C program was used to calculate curves of LR/D versus yT (Figure 20.9).

8 YT VS.20. R .5 Rehive-Gain Matrix 485 FIGURE 20.

R . vs.486 C~mPmiriOn Control-Binaty Dktdlath FIGURE 20.9 X .

10 YT vs-vs .20.5 Reldve-Gain M & 487 FIGURE 20.

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

In view of the ease of calculating steady-state column gains rigorously with programmable calculators or small computers. For column bottom composition control. For some columns overhead vapor or condensate composition measurements may be feasible.6 COmparitiOn Measurement Lo& 489 Shinskef has derived simplified equations for column gains based on what he terms a “separation factor. In making comparisons of the two methods with other columns.33 These results are clearly quite different from those obtained by the more rigorous method. 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. considerable attention has been devoted to selecting measurement locations for composition control.Y r ) h S + (ZF . One might t i k hn offhand that to control terminal composition.20. Then: R A = V 6.47) so For the illustrative problem. and (2) it has been shown that the closer the measurement to the point of manipulation.486933 d.XB h1 = (YT . we recommend the latter approach.46) and (20. .X B ) YT(1 .33. This may be partly due to the high terminal purities.6 COMPOSITION MEASUREMENT LOCATION In the distillation literature.47) gives hl = 6.Z F ) X B (1 . the better is the control.33 -5. The range of validity of Shinskey’s separation factor unfortunately has never been published. = 0. For this system his equation for X 1 is: 1 1+ Y T . This turns out indeed to be the case. 20.solution of equations (20.Y T ) where (20. column-located measurements should be as close to the top tray as possible. z. For overhead composition control. one should measure compositions at the column top and bottom. we found that for terminal purities of 95 mol percent or less. the two methods gave comparable results.X B ) Y T (1 .” S . the measurement should be located in the liquid line to the reboiler or in the vapor space below the bottom tray. but not on the top tray where there is usually poor mixing.

J. L. G. Shinskey. N. “control Both Ends for Profit.. and D. presented at Houston. Pmg.. J.” Proceedings. W. Luyben. Luyben. 3. This was true even though he provided no feedforward compensation for these changes. Interactiota A d y s k . F. Ray.” Institute of Chemical Engineer Symposium Series No. Eng. Pm. and W. “New Strategy Improves Dual Composition Control. and (2) that tray where the change in temperature per change in overhead (or bottoms) composition is a maximum. W. H. e.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.? 122:361 (1983). Fuentes. ISA. 5 . 1977.. Ryskamp. has swested a feedforward scheme for changing temperature controller set points.” ISA Paper C.” IEC Pmc. “A Comparison of . L. C. ISA Conference #27. W.” Wyahu. 9. 1980.” Chem. Research Triangle Park. Ellis. mas ell^. New York. Tolliver. and K. and R. O t c. New York.I. 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. “Feedback and Feedforward Control of Distillation Columns with Inverse Response. Hammerstrom.80-508. 61:74 (1965). m i McGraw-Hill. S. the concentration of the same component in the overhead decreases. 1983.I. “control of Hgh Purity Disdlation Columns... Since holding interior tray temperatures constant does not hold terminal compositions constant in the face of feed-composition changes. Two common ones are (1) that tray where the temperature change per tray is a maximum. T. “Steady-State Gains in Distillation Column Fdorward Control. 4. Fagervik..P. Uim16 has shown that for an ideal binary system. Waggoner. 7. London. 1981. an increase in feed concentration of the more volatile component led to an increase in overhead concentration of the same component.’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. O t c. T. Tex. REFERENCES 1. New York.. Uim showed that by maintaining a constant reflux-to-feed ratio.. it has occurred to some authors that one should vary interior tray temperatures as feed composition changes. For the same system. G. “Distillation Column Control: A Review and Perspective from the C. Advanad Pmcess Gmtml. Di-tdlaziota G m. Waller. 2. and this is seldom the case with today’s measurement technology. 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.. 51-59 (June 1980). 8. C. D s e Dev.d. 32..C. L. 1968. 1972. Luyben. McAvoy. 10.18 for example. C. L. C. Furthermore. he got f&ly good control of overhead composition for feed composition changes.. L. McGraw-Hill. K. 6. Luyben.

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

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

and how their parameters are selected.e. Adaptive sampling rates and gain tuning are typical examples. T . but also may afTect controlled variables in other loops. for example. to cancel the effect of the disturbance on the controlled variable (Chapter 20). A PID controller that is approximated by simple rectangular integration is: ek 70 + -(e. The majority of articles written about interaction compensation have been for continuous control (Chapter 20). A change in one manipulative variable. Control engineers meanwhile have devised simpler methods to handle nonlinearities and the multivariable nature of a distillation column.’3714 chapter discusses the application This of some of these concepts to sampled-data control <. not only affects the variable it is controlling. one at a time. Feedforward control is yet another type of multivariable control. 21.f a distillation column. modem control theories to the place where they are convenient to implement. their forms. Override control allows a valve to be manipulated by several different controllers. More thorough discussions on this can be found in several texts on &@tal ~ 0 n t r 0 1 . called “minimal prototypeyy control.Another kind of optimal control. A measured load disturbance is used to adjust the manipulative variable before feedback control occurs.2 CONTROL ALGORITHMS Control algorithms have been the subject of numerous papers over the years and are still evolving today. Interaction compensation is a case in point. The types of algorithms people have proposed range from simple proportional and discretized analog types (PI and PID) to highly complex optimal control algorithms. Therefore..7-” Much still needs to be done to bring some of the new. interaction compensation attempts to eliminate the effect one manipulative variable has on other controlled variables. ~ ~ 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. It is worthwhile here to review some of the more common types. Some simple adaptive control schemes have also been used. according to a built-in logic that selects the controller to meet the most urgent need of the process (Chapter 9). 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. while Fitzpatrick” and Shunta” designed sampled-data interaction compensators.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..-l (21.1) .

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

496 Sampled-Data Control . 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. Fundamentally we are interested in achieving good control in the face of set-point changes and load disturbances-traditionally called servo and regulator control. and its output is the set point to the analog steam-flow controller.or bottom-product composition-control loop. Let us consider a top.fDhiUatMn Columns 21. we now look specifically at how sampled-data control is applied to a distillation column. 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. The temperature control loop is in the computer. and that the temperature undergoes a tolerable deviation when a load disturbance FIGURE 21.1 Sampled-data control .3 SERVO AND REGULATOR CONTROL Having discussed some general points about sampled-data control techniques and algorithms. A typical loop is shown in Figure 21.1 where steam flow is manipulated to control the temperature on a stripping tray.

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.

DLS. Chapter 20 discusses methods to eliminate interaction for analog controls.5 INTERACTION COMPENSATION Interaction occurs in multivariable systems when a change in a manipulative variable causes a deviation in more than one controlled variable.DLRare designed according to Section 21.(z) when a change is made in LR(z).18) We desire zero change in X.21) . 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. A classic example is a distillation column in which a rectlfjrlngtray temperature is controlled by manipulating reflux.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.18): (21. The purpose of this section is to illustrate how interaction compensators can be achieved by sampled-data algorithms for the control system discussed above. Consider first the stripping section control loop.20) A similar calculation in the re+g loop leads to Dm(z): (21. Basically the interaction compensators will function like feedforward algorithms with reflux and vapor boilup treated as load disturbances on the opposite controlled variables. If the loops are tightly tuned. and stripping tray temperature by manipulating steam.'2 The compensators are used in conjunction with the dual feedback algorithms. DsR. this interaction can result in unstable control. and it causes a further deviation in the rectifying temperature. The interaction algorithms DIS and Dm are designed as follows. The controlled variable X .19) Equation (21. 21.19) defines DIS(z): (21. The load and set-point algorithms DSS.6. The steam flow is adjusted to compensate. is given by equation (21.3. A change in reflux causes a deviation in the stripping tray temperature. the necessary change to make in V ( z )is: (21. Therefore. The control system is illustrated in Figure 21.

21.5 Inter& Gmm ? p & 509 FIGURE 21.6 Interaction compensation .

The secondary variable measurement is the proper feedback for a primary controller in a cascade loop. for this discussion. The override control concept can be implemented readily in the computer as well. 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 algorithm is structured to track an external feedback signal. Analog systems require changing hardware and rewiring. D&) and Dm(z)make the appropriate change in the manipulative variable in the opposite loop to compensate for it. therefore. however.510 Sa?nphd-Da.7a and 21.8 illustrates a conventional loop. applies to a DDC loop as well. The improvement in control with interaction compensators is sipficant. 7 ~ and 21. With no means of tracking the output of the secondary variable (assumed equal to the output of the signal selector).7d for a 10 percent change in feed composition. We are again controlling the temperature in the stripping section of the distillation column by manipulating steam flow and using the dual algorithm. The flowcontrol dynamics are very fast compared with the composition dynamics and. is devoted to the design of sampled-data algorithms that prevent windup. for safety considerations.preferred for critical overrides. Figure 21. the dual algorithm in Figure 21. The advantage of computer implementation is the ease with which control schemes can be modified by simply reprogramming. can be neglected for this analysis. for sampled-data algorithms. However.30 Tracking Dual Algorithm We assume. The secondary . 21.7b show the closed-loop responses of composition on trays 2 and 18 for set-point changes and Figures 2 1 .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. that the overrides are implemented with analog hardware. when a change in reflux or vapor boilup occurs. Therefore. Figures 21. This section. therefore. hard-wired circuits are still. A tracking mechanism to prevent controller saturation or "windup" is needed for sampled-data systems as well as analog. 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. The principles of override control discussed in Chapter 9 suffice for sampleddata systems and will not be restated here. Overrides can enter the loop in the set-point path of the flow controller through the signal selector.taControl ofDimU&im Columns Thus. Similarly. but requires a different approach. This discussion. the algorithm must be restructured as in Figure 21.8 saturates when an override occurs.9.

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

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

21.Loops with Ovewides m l 513 FIGURE 21.6 Samphd-Data G m fb.9 Tracking “dual” control in loop with overrides .

[C"'*(s)D?*(s) .*(s) + [C"*(S)D~(S) C*(s)]K (21.28) Ds(z)is the same for both loops. C(s) = [ C " * ( ~ ) D W . C ~ ( Z ) D S ( Z ) G P W ) (K 1 .22) The asterisk (*) denotes a variable that has been sampled. (21.24) is z-transformed and rearranged to solve for the controlled variable C(z). and the controlled variable. Ds(z) is solved from equation (21.C*(s)]K 1 .DL*(s) Therefore: C(S)= [C"*(s)Df(s) .C*(s)l M(s)Gp(s) +M*(s)Dt*(s)H(s)Gp(s) L(s)G.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.26) K is specified to make DL(z) physically realizable. .DL(z)) + G J ( 4 C(z) = + K GPH(z)(l .DL(41 G. so that zo terms in the numerator ofDL(z) are zero. M*(s) = M*(s)D. Dt(z)is related toDL(z)by equation (21. .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). The algorithm D is solved by first writing the loop equation.22).C*(S)] KH(s)GP(s) Equation (21. K is a static gain and its input is the difference L between the set-point algorithm 0.23) . The manipulative variable M*(s) is solved and substituted into equation (21.27).25) Dt(z) is solved by setting C"(z) 0.DL(z)) = (21.(s) + (21. The overrides L are assumed to be zero. that is. (21.H(z)KC"(z) (21.

(21.1 into equation (21. T~ is the reset time. Dt(z) alternatively could have been determined by equation (21. Tracking PI Algorithm The tracking structure is not unique to the dual algorithms.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 21.1. 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.13 show the closed-loop responses to a 20 percent step increase in feed composition.21.29) Figures 21.27).01 for the conventional and tracking dual algorithms.10).27).730 to make DL (2) physically realizable for a step-load disturbance.057 to 0. Again control is improved with the tracking design.8 and 21.6) for a feed-composition disturbance. The tracking algorithm does not saturate and resumes control quickly. Figure 2 1.7) and (21.9 for the 20-tray distillation column whose transfer functions are given in Table 21.Dt(z) is solved by substituting the appropriate terms in Table 21.9).10 and 21.30) K. C ( z ) was specified as in equation (21. 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”). is the controller gain and D ( z ) and equation (21. The conventional algorithm saturates and takes almost an hour to unwind (Figure 21. 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. (21.K has a value of .12 and 21.26).14 compares the conventional loop and the restructured tracking form.11 show the closed-loop responses to a set-point change in x2 fi-om 0. D‘(z) is calculated from . The sampling period is 1 minute. Figures 21.

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

21.6 Smnpled-Data G m fiLoops with Ovemdes m l 517 FIGURE 21. with setpoint disturbance .1 1 Tracking sampled-data control of X.

12 Conventional control o X. with feed composition disturbance f .518 Sampled-Data Control o Didhtiim Columns f FIGURE 21.

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

14 Comparison o conventional and tracking PI control f .520 Sampled-Data Control of Dtjtllation Columns FIGURE 21.

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

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

C. C. 522 Cutler... G.. 521 Fitzpamck. 1 . C. 258. 242 Hamilton. W. C. 292 Arant. 292 Holland. R. M. 522 Day. 135 Fuentes. L. 310. 426 Harper. L.. G. D. A. H. R. R. 521 Berger. 292 Edgar.. 310 htrom. K.. E. 444 Carr. E.. 66 Hollander. W.. J. 310. R. 310.. M.. 490 Gaines. F. B.. T. 135. B. 522 6. 192 Chin. C. R. D. 373 Hengstebeck.. 445. 292. 490 Handey. M. L...... H. N. 180 Driskell. B. 108. 464 Beck. 521 Aikman.. 108. 192 Fehervari. 311 Baber. D. 23.404. R.. 444 Harriott. E. D. C. 373.. T. 297.. P. 242 Grote. 521 Hammerstrom. 227 Gallier. 107. B... 241.. L. L. F. I?.. K. R. T. J... K. C. 23. 311. 521 Buckley. P. R. L.. 311 Dobratz. R. 444 Ellis. 404 Chiang. W. R.. 311.. T. C.. 522 Cheung. 444 Hempel. S. 490 Fagervik.J. J. 464 Cadman. 227. R.. S. H. M. A. D.. 426 530 .445. 79. F... 340. L. T. 522 Frank.. 521 Harbert.. 23. 292 Archamboult. 310 Beesley. D. 305. 66 Binder. 310. L. 66 Gould. 180. D. 107 Church. E. 522 Cox. 311 Geyer. M.. E. 313. R.. Cundall. 292 Bolles.. G. 23 Hepp. 192 Bergen. T. 442. M. P. L. 464 D o h . 311 Douglas. D. J.. 107 Bonilla. B. 297.. J. M.. 23. 426 Bristol. L. A.. D. A. K.. 227 Gilliland. C. J. C. 135 Compio. J... K.. J. A. L.. 441. 135 Griffin. 242. T. 445. 242 Boyd. E. W. 310.. 404 Fisher.. 240. P... 242 Bremer... T. 230. 231. 444 Bauer. 292 Billet. E... W. T. 444 Anderson.. 495.. F. J. 444 Campbell. 23. 347. 135 DOSS... 192 Giles. 135.374 Arnold. 430.. D. 4 4 500. F. J. 444 ANSUISA. A.. N... D. J. 310 Edwards.. R. W. 2 3 Harnett. 167. A. O. E. C. 435. 107 Dahlin. 490 Brosilow... 445. 527 Garcia. 347. V.. 306.. A.404. J.Author Index Ahlgren. 444 Chao Yung-Cheug.. 490 Fahmi. G. R. 373 Grabbe.. E.

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

.. D..532 Smith. 231. 73. 490 Wahl.. W.490 . W. V. 521 Snyder. 444.465. 326 Williams.490. 491 Wood. 66.. 444. 3 11 Treybal. 310 Waggoner.465. 521 Strangio.465.. 444 Thal-Larsen... 108. 135 Stevens. J. C. E...464. 464. K.. 490 Webber. 361. 242 Vinante. H. 441. D. N. N. L. 151 Stanton. E. J.. 326 Tivv. 348.. K.464 Sutherland. 135 Speicher. T.. 522 Teager. H.. 347. D.. 23. A. H. A. M. 444. R. 135 Smith. R. 475. L. 11. D.. 446. H. 325... V. T. E. W. E. 491 Wade. 404 Author I& Uim.. 522 Touchstone. 66. 107 Williams. T. 192 Vermilion.. R. V. 242 Tolher. 374. 242 Wild.490.491 Van Winkle.465. J. B. 231... M. M. 192. F. B. 446. 426 Stempling. V. 374 Thistelthwaite. K. 310.. A. A. 475. 490 Tou. E.. A...T. K. 310 Wright. C.. 444 Waller.. 23. B. 0. L:. A. H. J. J. 426 Wood. 242 Weber. F.464 Tyreus... 310... 306. 326 Vanwormer.. R. C. 313..445. C. O. J. 231.

303-305 Conventions. 4.. 11-12. 126-130. 25 Bottoms. 200-202 Augmented PI level controller. 72-79 pressure. 181-182 Constraints. 2 proportional-integral (reset). 12-13 Divider.167 Controller tuning. 385 Average temperature measurement. 193-227. 279 impedance. 299-303 Feedforward plus overrides. 4. 213-214. 60-61 columns atmospheric. 199 proportional-only. 100-107. 199 single-loop (SISO). 19 integral (floating). 70. 303 Column design. 25 Bubble point. 11-12. 1 product quality. 126-130. energy. 211-213 Autoovemde level control time constant.465-491 design approach to. 141-143 Feedforward compensation. 166. 5 15 External reset feedback. 137-151 Feed rank size. 327-3% multivariable (MIMO). 25 Disturbances. 405426 Balancing energy and material handling capacities. 70. 2 1 C. 153. 375-404 Averaging pressure control. 15-16 objectives. 233 Air-cooled condensers. 241-242 Computations. 299 Feed enthalpy control.Subject Index Activity coefficients 39. 80-84 vacuum. 336-337 via feed. heat flow. 308-309 Base level control via bottom product. 202. 14-15 unfavorable schemes. 39-40 Calandria. 375-404 composition. 255-256 Condensate receiver level control. 83. 4. 241 Averaging level control. 74. 80-84 Composition control. composition. 241-242 Euler integration. 273 Dew point. 231-234 Distillate. 166 unreasonable schemes. 200 material balance. 140-141 527 . 144-145 Feed temperature control. 110 Cascade control.40-41 Differential vapor pressure. 70. 70-72 Conservation. 100-107 holdup. 295-297 design procedure. 58-60 AP measurement. 250 Double-differential temperature. 34. 307-308 Control averaging level. 11-12. 389-390 via steam to reboiler. 229 Anti reset-windup. 100-107.465-491 estimators. 202-205 Feed systems. pneumatic. 71. 382-383 Automatic start-up/shutdown. 6-10. 1. 99 Analyzers. 240 Estimators. 104-107 Condensers. 390-399 Bottom product. 420-426 rating. 34.

468 Multiplier. 256-273 characterized displacers. 260 external damping. 261-262 flush diaphragm transmitter and 1: 1 repeater. 117-119 thermosyphon. 489 Sidedraw columns. 100. 266-268 specific gravity compensation. 45-49 Nonlinear PI controllers. 28 column overhead. 4. 243-249 Reflux cycle. 262-263 purge system errors. 258-259 Nonideality. 195 Low selectors. 39 Reboiler dynamics. 193-227 Pump bypass. 272 displacer-type. 249-2 55 Overhead level control. 169-180 composition control. 87-90. 468-475 Interlocks. 366-370 Flow and flow ratio conventions. 15-16 Murphree tray efficiency. 110-116 Rectification section. 43-45 Reset cycle. 57 q-line. 194 Internal r d u x calculation. 154-166. 170-180 Single-loop control (SISO). 243-249 external. 70. 109. 478-489 Relative volatility. 215-217 McCabe-Thiele diagram. waves. 193-227 for sidedraw columns. 288-289 Gravity return reflux. 54-55 Reflux. 143 Flash. 349-359 Flooded reboiler. 333. 4. 92-99 Relative gain array (matrix). 313. 357-365 Reboiler types forced-circulation. 383 On-line identification. 116-1 17 internal. 62-63 Subject I d t x Modern control theory. 246-249 Multivariable control. 119-122 kettle-type. 394. 214-215 q. 195 Minimum number of trays. 256 damping chamber.107 Overrides. 189 Separation factor (Shinskey). 220-227 PI controller. 14 . 273 high viscosity fill AP transmitter. 182-192 High base pressure override. 255-256 Heat recovery schemes. 57-58 Raoult's law. 243-249 Inverse response. 259-260 Material balance control. 51-54 Orifice. 6-10. 309-310 On-line models 305-310 Operating lines. 296-297. 195 Hot vapor bypass.528 Feed tray location. 49-64 Median selector. 234-239 Pressure control. 27 internal. 405-426 Product quality control. 385-386 Pressure-compensated temperature. 301 PI level controller tuning. 195 High selectors. 46-91 Protective controls. pneumatic. 264-266 two flush diaphragm transmitters. 439 Level control. 16-17. 90-99 Heat flow computations. 428 Noise. temperature and pressure compensation. 375 Level measurement. 69 Maximum capacity overrides. 195 Manometer. 86-87 Interactions. 109. 3-4. 266 column base. 210 High limiters. enthalpy factor. 210 High column AP override. 260-261 delta P transmitter with double remote seals. 114-1 16. 381 Sensible heat recovery. 327-346 Shinskey scheme. 63-64 Minimum reflux ratio. 109. 41-43 Flooded condenser. averaging. 272-273 Low limiters.

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

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.