36 views

Uploaded by William Ruiz

- Likeness Between Erd p
- Jiri Klemes, Ferenc Friedler, Igor Bulatov, Petar Varbanov-Sustainability in the Process Industry_ Integration and Optimization (Green Manufacturing & Systems Engineering)-McGraw-Hill Professional (20
- Liew, 2012 - numerical technique for Total Site sensitivity analysis.pdf
- WCECS2008_pp123-127
- Advanced Computational Method
- Cross Layer Sensor
- Mamdouh Article 2.pdf
- pinchanalysisatoolforefficientuseofenergy-140920130116-phpapp01.pdf
- Energy Recovery+Pinch
- A Minlp Model for the Rigorous Design of Shell and Tube
- MATLABSIMULINK-Based-Model-of-Single--Machine.pdf
- Handbhandbook of modeling for discert optimizationook of Modeling for Discert Optimization
- Analysis and Evaluation of Contemporary Nursing Knowledge
- 2005_Tuning PID Controllers using the ITAE.pdf
- Batch Size JIT
- 955209902D03
- Multi-Objective Numerical Optimization 2008.pdf
- Ijee1673
- CSI4124_SYS5110_CaseStudies_Fall_2014.pdf
- I, S and IS

You are on page 1of 385

Process Industry

About the Authors

Ji Kleme, H DSC, is one of the key personalities

of the world-leading Centre of Excellence in Process

Integration at the University of Manchester Institute

of Science and Technology in the United Kingdom.

the University of Pannonia in Hungary.

Integration at the University of Manchester in the

United Kingdom.

Process Integration and Intensification (CPI2) at the

University of Pannonia in Hungary.

Sustainability in the

Process Industry

Integration and Optimization

Ji Kleme

Ferenc Friedler

Igor Bulatov

Petar Varbanov

Lisbon London Madrid Mexico City

Milan New Delhi San Juan

Seoul Singapore Sydney Toronto

Copyright 2011 by The McGraw-Hill Companies, Inc. All rights reserved. Except as permitted under

the United States Copyright Act of 1976, no part of this publication may be reproduced or distributed

in any form or by any means, or stored in a database or retrieval system, without the prior written

permission of the publisher.

ISBN: 978-0-07-160555-7

MHID: 0-07-160555-X

The material in this eBook also appears in the print version of this title: ISBN: 978-0-07-160554-0,

MHID: 0-07-160554-1.

All trademarks are trademarks of their respective owners. Rather than put a trademark symbol after

every occurrence of a trademarked name, we use names in an editorial fashion only, and to the benefit

of the trademark owner, with no intention of infringement of the trademark. Where such designations

appear in this book, they have been printed with initial caps.

McGraw-Hill eBooks are available at special quantity discounts to use as premiums and sales

promotions, or for use in corporate training programs. To contact a representative please e-mail us at

bulksales@mcgraw-hill.com.

Information contained in this work has been obtained by The McGraw-Hill Companies, Inc.

(McGraw-Hill) from sources believed to be reliable. However, neither McGraw-Hill nor its authors

guarantee the accuracy or completeness of any information published herein, and neither McGraw-Hill

nor its authors shall be responsible for any errors, omissions, or damages arising out of use of this

information. This work is published with the understanding that McGraw-Hill and its authors are

supplying information but are not attempting to render engineering or other professional services. If

such services are required, the assistance of an appropriate professional should be sought.

TERMS OF USE

This is a copyrighted work and The McGraw-Hill Companies, Inc. (McGrawHill) and its licensors

reserve all rights in and to the work. Use of this work is subject to these terms. Except as permitted

under the Copyright Act of 1976 and the right to store and retrieve one copy of the work, you may

not decompile, disassemble, reverse engineer, reproduce, modify, create derivative works based upon,

transmit, distribute, disseminate, sell, publish or sublicense the work or any part of it without McGraw-

Hills prior consent. You may use the work for your own noncommercial and personal use; any other

use of the work is strictly prohibited. Your right to use the work may be terminated if you fail to comply

with these terms.

GUARANTEES OR WARRANTIES AS TO THE ACCURACY, ADEQUACY OR COMPLETE-

NESS OF OR RESULTS TO BE OBTAINED FROM USING THE WORK, INCLUDING ANY

INFORMATION THAT CAN BE ACCESSED THROUGH THE WORK VIA HYPERLINK OR

OTHERWISE, AND EXPRESSLY DISCLAIM ANY WARRANTY, EXPRESS OR IMPLIED, IN-

CLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY OR FIT-

NESS FOR A PARTICULAR PURPOSE. McGraw-Hill and its licensors do not warrant or guarantee

that the functions contained in the work will meet your requirements or that its operation will be

uninterrupted or error free. Neither McGraw-Hill nor its licensors shall be liable to you or anyone else

for any inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting

therefrom. McGraw-Hill has no responsibility for the content of any information accessed through

the work. Under no circumstances shall McGraw-Hill and/or its licensors be liable for any indirect,

incidental, special, punitive, consequential or similar damages that result from the use of or inability to

use the work, even if any of them has been advised of the possibility of such damages. This limitation

of liability shall apply to any claim or cause whatsoever whether such claim or cause arises in contract,

tort or otherwise.

Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Energy Efficiency . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Screening and Scoping: Auditing,

Benchmarking, and Good Housekeeping . . . 5

1.4 Balancing and Flowsheeting Simulation

as a Basis for Optimization . . . . . . . . . . . . . . . . 7

1.5 Integrated Approach: Process Integration . . . 7

1.6 Optimal Process Synthesis and Combinatorial

Graphs ................................ 8

1.7 How to Apply the Process Integration and

Optimization Technology . . . . . . . . . . . . . . . . . 9

2 Process Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1 Introduction: The Need for Process

Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2 What Is Process Integration? . . . . . . . . . . . . . . . 12

2.3 History and Development of Process Integration 12

2.4 Pinch Technology and Targeting Heat

Recovery: The Thermodynamic Roots . . . . . . 14

2.5 Supertargeting: Full-Fledged HEN Targeting 15

2.6 Modifying the Pinch Idea for HEN Retrofit . . 16

2.7 Mass Exchange and Water Networks . . . . . . . 17

2.8 Benefits of Process Integration . . . . . . . . . . . . . 18

2.9 The Role of PI in Making Industry Sustainable 20

2.10 Examples of Applied Process Integration . . . . 20

2.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3 Process Optimization . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.2 Model Building and Optimization: General

Framework and Workflow . . . . . . . . . . . . . . . . 24

v

vi Contents

Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.3.1 What Is Optimization? . . . . . . . . . . . . . 25

3.3.2 Mathematical Formulation of

Optimization Problems . . . . . . . . . . . . 25

3.4 Main Classes of Optimization Problems . . . . 26

3.5 Conditions for Optimality . . . . . . . . . . . . . . . . 28

3.5.1 Conditions for Local Optimality . . . . 28

3.5.2 Conditions for Global Optimality . . . 28

3.6 Deterministic Algorithms for Solving Continuous

Linear Optimization Problems . . . . . . . . . . . . . 29

3.7 Deterministic Algorithms for Solving Continuous

Nonlinear Optimization Problems . . . . . . . . . . 29

3.7.1 Search Algorithms for Nonlinear

Unconstrained Problems . . . . . . . . . . . 30

3.7.2 Algorithms for Solving Constrained

Nonlinear Problems . . . . . . . . . . . . . . . 31

3.8 Deterministic Methods for Solving

Discrete Problems . . . . . . . . . . . . . . . . . . . . . . . . 31

3.9 Stochastic Search Methods for Solving

Optimization Problems . . . . . . . . . . . . . . . . . . . 32

3.10 Creating Models . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.10.1 Conceptual Modeling . . . . . . . . . . . . . 34

3.10.2 Mathematical Modeling of Processes:

Constructing the Equations . . . . . . . . 35

3.10.3 Choosing an Objective Function . . . . 37

3.10.4 Handling Process Complexity . . . . . . 38

3.10.5 Applying Process Insight . . . . . . . . . . 40

3.10.6 Handling Model Nonlinearity . . . . . . 41

3.10.7 Evaluating Model Adequacy and

Precision . . . . . . . . . . . . . . . . . . . . . . . . . 43

Energy Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.1 Introduction to Heat Exchange and

Heat Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.1.1 Heat Exchange Matches . . . . . . . . . . . . 46

4.1.2 Implementing Heat Exchange

Matches . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.2 Basics of Process Integration . . . . . . . . . . . . . . . 47

4.2.1 Process Integration and Heat

Integration . . . . . . . . . . . . . . . . . . . . . . . 47

Contents vii

4.2.3 Performance Targets . . . . . . . . . . . . . . . 48

4.2.4 Heat Recovery Problem Identification 48

4.3 Basic Pinch Technology . . . . . . . . . . . . . . . . . . . 50

4.3.1 Setting Energy Targets . . . . . . . . . . . . . 51

4.3.2 The Heat Recovery Pinch . . . . . . . . . . 54

4.3.3 Numerical Targeting: The Problem

Table Algorithm . . . . . . . . . . . . . . . . . . 56

4.3.4 Threshold Problems . . . . . . . . . . . . . . . 60

4.3.5 Multiple Utilities Targeting . . . . . . . . . 61

4.4 Extended Pinch Technology . . . . . . . . . . . . . . . 69

4.4.1 Heat Transfer Area, Capital Cost, and

Total Cost Targeting . . . . . . . . . . . . . . . 69

4.4.2 Heat Integration of Energy-Intensive

Processes . . . . . . . . . . . . . . . . . . . . . . . . 71

4.4.3 Process Modification . . . . . . . . . . . . . . 80

4.5 HEN Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . 81

4.5.1 The Pinch Design Method . . . . . . . . . . 81

4.5.2 Superstructure Approach . . . . . . . . . . 93

4.5.3 A Hybrid Approach . . . . . . . . . . . . . . . 95

4.5.4 Key Features of the Resulting

Networks . . . . . . . . . . . . . . . . . . . . . . . . 96

4.6 Total Site Energy Integration . . . . . . . . . . . . . . 96

4.6.1 Total Site Data Extraction . . . . . . . . . . 97

4.6.2 Total Site Profiles . . . . . . . . . . . . . . . . . . 97

4.6.3 Heat Recovery via the Steam System . . 99

4.6.4 Power Cogeneration . . . . . . . . . . . . . . . 101

4.6.5 Advanced Total Site Optimization and

Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 102

5 Mass Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.1 Water Integration . . . . . . . . . . . . . . . . . . . . . . . . 105

5.2 Minimizing Water Use and Maximizing

Water Reuse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5.2.1 Legislation . . . . . . . . . . . . . . . . . . . . . . . 106

5.2.2 Best Available Techniques . . . . . . . . . . 107

5.2.3 Water Footprint . . . . . . . . . . . . . . . . . . . 108

5.2.4 Minimizing Water Usage and

Wastewater . . . . . . . . . . . . . . . . . . . . . . 111

5.3 Introduction to Water Pinch Analysis . . . . . . . 113

5.4 Flow-Rate Targeting with the Material

Recovery Pinch Diagram . . . . . . . . . . . . . . . . . . 116

viii Contents

5.6 Water Minimization via Mathematical

Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.6.1 Introduction to Mathematical

Optimization . . . . . . . . . . . . . . . . . . . . . 118

5.6.2 Illustrative Example: A Brewery Plant 120

5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

6.1 Design and Management of Hydrogen

Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

6.2 Oxygen Pinch Analysis . . . . . . . . . . . . . . . . . . . 125

6.3 Combined Analyses, I: Energy-Water,

Oxygen-Water, and Pinch-Emergy . . . . . . . . . 126

6.3.1 Simultaneous Minimization of

Energy and Water Use . . . . . . . . . . . . . 126

6.3.2 Oxygen-Water Pinch Analysis . . . . . . 128

6.3.3 Emergy-Pinch Analysis . . . . . . . . . . . . 130

6.4 Combined Analysis, II: Budget-Income-Time,

Materials Reuse-Recycling, Supply Chains,

and CO2 Emissions Targeting . . . . . . . . . . . . . . 131

6.4.1 Budget-Income-Time Pinch Analysis . 131

6.4.2 Materials Reuse-Recycle and Property

Pinch Analysis . . . . . . . . . . . . . . . . . . . 133

6.4.3 Pinch Analysis of Supply Chains . . . . 136

6.4.4 Using the Pinch to Target CO2

Emissions . . . . . . . . . . . . . . . . . . . . . . . . 138

6.4.5 Regional Resource Management . . . . 139

6.5 Heat-Integrated Power Systems:

Decarbonization and Low-Temperature Energy 142

6.5.1 Decarbonization . . . . . . . . . . . . . . . . . . 142

6.5.2 Low-Temperature Energy . . . . . . . . . . 143

6.6 Integrating Reliability, Availability, and

Maintainability into Process Design . . . . . . . . 144

6.6.1 Integration . . . . . . . . . . . . . . . . . . . . . . . 144

6.6.2 Optimization . . . . . . . . . . . . . . . . . . . . . 146

6.7 Pressure Drop and Heat Transfer Enhancement

in Process Integration . . . . . . . . . . . . . . . . . . . . 146

6.8 Locally Integrated Energy Sectors and

Extended Total Sites . . . . . . . . . . . . . . . . . . . . . . 148

6.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

Contents ix

7.1 Classic Approach: Mathematical Programming 151

7.2 Structural Process Optimization: P-Graphs . . 153

7.2.1 Process Representation via P-Graphs 154

7.2.2 The P-Graphs Significance for

Structural Optimization . . . . . . . . . . . 155

7.2.3 The P-Graphs Mathematical Engine:

MSG, SSG, and ABB . . . . . . . . . . . . . . . 157

7.3 Scheduling of Batch Processes: S-Graphs . . . . 159

7.3.1 Scheduling Frameworks: Suitability

and Limitations . . . . . . . . . . . . . . . . . . . 159

7.3.2 S-Graph Framework for Scheduling . . 161

8.1 The Role of Optimization in Process Synthesis 165

8.2 Optimization Tools for Efficient

Implementation of PI . . . . . . . . . . . . . . . . . . . . . 166

8.3 Optimal Process Synthesis . . . . . . . . . . . . . . . . 167

8.3.1 Reaction Network Synthesis . . . . . . . . 167

8.3.2 Optimal Synthesis of Heterogeneous

Flowsheets . . . . . . . . . . . . . . . . . . . . . . . 169

8.3.3 Synthesis of Green Biorefineries . . . . . 171

8.3.4 Azeotropic Distillation Systems . . . . . 173

8.4 Optimal Synthesis of Energy Systems . . . . . . 176

8.4.1 Simple Heat Integration . . . . . . . . . . . . 176

8.4.2 Optimal Retrofit Design . . . . . . . . . . . 177

8.5 Optimal Scheduling for Increased Throughput,

Profit, and Security . . . . . . . . . . . . . . . . . . . . . . . 179

8.5.1 Maximizing Throughput and Revenue 179

8.5.2 Heat-Integrated Production Schedules 180

8.6 Minimizing Emissions and Effluents . . . . . . . 183

8.7 Availability and Reliability . . . . . . . . . . . . . . . . 186

8.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

9.1 Overview of Available Tools . . . . . . . . . . . . . . . 191

9.2 Graph-Based Process Optimization Tools . . . 191

9.2.1 PNS Solutions . . . . . . . . . . . . . . . . . . . . 191

9.2.2 S-Graph Studio . . . . . . . . . . . . . . . . . . . 193

9.3 Heat Integration Tools . . . . . . . . . . . . . . . . . . . . 195

9.3.1 SPRINT . . . . . . . . . . . . . . . . . . . . . . . . . . 195

x Contents

9.3.3 STAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

9.3.4 SITE-int . . . . . . . . . . . . . . . . . . . . . . . . . . 198

9.3.5 WORK . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

9.3.6 HEXTRAN . . . . . . . . . . . . . . . . . . . . . . . 199

9.3.7 SuperTarget . . . . . . . . . . . . . . . . . . . . . . 200

9.3.8 Spreadsheet-Based Tools . . . . . . . . . . . 200

9.4 Mass Integration Software: WATER . . . . . . . . 201

9.5 Flowsheeting Simulation Packages . . . . . . . . . 202

9.5.1 ASPEN . . . . . . . . . . . . . . . . . . . . . . . . . . 202

9.5.2 HYSYS and UniSim Design . . . . . . . . . 203

9.5.3 gPROMS . . . . . . . . . . . . . . . . . . . . . . . . . 204

9.5.4 CHEMCAD . . . . . . . . . . . . . . . . . . . . . . 205

9.5.5 PRO/II . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

9.6 General-Purpose Optimization Packages . . . . 206

9.6.1 GAMS . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

9.6.2 MIPSYN . . . . . . . . . . . . . . . . . . . . . . . . . 207

9.6.3 LINDO . . . . . . . . . . . . . . . . . . . . . . . . . . 208

9.6.4 Frontline Systems . . . . . . . . . . . . . . . . . 209

9.6.5 ILOG ODM . . . . . . . . . . . . . . . . . . . . . . 209

9.7 Mathematical Modeling Suites . . . . . . . . . . . . . 210

9.7.1 MATLAB . . . . . . . . . . . . . . . . . . . . . . . . 210

9.7.2 Alternatives to MATLAB . . . . . . . . . . . 211

9.8 Other Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

9.8.1 Modelica . . . . . . . . . . . . . . . . . . . . . . . . . 211

9.8.2 Emerging Trends . . . . . . . . . . . . . . . . . . 212

9.8.3 Balancing and Flowsheeting Simulation

for Energy-Saving Analysis . . . . . . . . . 215

9.8.4 Integrating Renewable Energy into

Other Energy Systems . . . . . . . . . . . . . 216

10.1 Heat Pinch Technology . . . . . . . . . . . . . . . . . . . 219

10.1.1 Heat Pinch Technology: First Problem 219

10.1.2 Heat Pinch Technology:

Second Problem . . . . . . . . . . . . . . . . . . . 224

10.2 Total Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

10.2.1 Total Sites: First Problem . . . . . . . . . . . 226

10.2.2 Total Sites: Second Problem . . . . . . . . . 231

10.3 Integrated Placement of Processing Units

and Data Extraction . . . . . . . . . . . . . . . . . . . . . . 234

Contents xi

10.4.1 Utility Placement: First Problem . . . . 238

10.4.2 Utility Placement: Second Problem . . 243

10.5 Water Pinch Technology . . . . . . . . . . . . . . . . . . 247

10.5.1 Water Pinch Technology: First Problem 247

10.5.2 Water Pinch Technology:

Second Problem . . . . . . . . . . . . . . . . . . 249

11.1 Energy Recovery from an FCC Unit . . . . . . . . 253

11.2 De-bottlenecking a Heat-Integrated Crude-Oil

Distillation System . . . . . . . . . . . . . . . . . . . . . . . 256

11.3 Minimizing Water and Wastewater in a

Citrus Juice Plant . . . . . . . . . . . . . . . . . . . . . . . . 262

11.4 Efficient Energy Use in Other Food and

Drink Industries . . . . . . . . . . . . . . . . . . . . . . . . . 268

11.5 Synthesis of Industrial Utility Systems . . . . . . 271

11.6 Heat and Power Integration in Buildings and

Building Complexes . . . . . . . . . . . . . . . . . . . . . . 275

11.7 Optimal Design of a Supply Chain . . . . . . . . . 277

11.8 Scheduling a Large-Scale Paint Production

System ................................ 279

12.1 Data Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . 283

12.1.1 When Is a Stream a Stream? . . . . . . . . 284

12.1.2 How Precise Must the Data Be at

Each Step? . . . . . . . . . . . . . . . . . . . . . . . 285

12.1.3 How Can Considerable Changes in

Specific Heat Capacities Be Handled? 286

12.1.4 What Rules and Guidelines Must Be

Followed to Extract Data Properly? . . 287

12.1.5 How Can the Heat Loads, Heat

Capacities, and Temperatures of an

Extracted Stream Be Calculated? . . . . 289

12.1.6 How Soft Are the Data in a Plant or

Process Flowsheet? . . . . . . . . . . . . . . . . 290

12.1.7 How Can Capital Costs and Operating

Costs Be Estimated? . . . . . . . . . . . . . . . 290

12.2 Integration of Renewables: Fluctuating

Demand and Supply . . . . . . . . . . . . . . . . . . . . . 292

12.3 Steady-State and Dynamic Performance . . . . . 292

xii Contents

12.5 Making It Happen . . . . . . . . . . . . . . . . . . . . . . . 293

13.1 General Sources of Information . . . . . . . . . . . . 295

13.1.1 Conferences . . . . . . . . . . . . . . . . . . . . . . 295

13.1.2 Journals . . . . . . . . . . . . . . . . . . . . . . . . . 297

13.1.3 Service Providers . . . . . . . . . . . . . . . . . 297

13.1.4 Projects . . . . . . . . . . . . . . . . . . . . . . . . . . 301

13.2 Heat Integration . . . . . . . . . . . . . . . . . . . . . . . . . 301

13.2.1 Conferences . . . . . . . . . . . . . . . . . . . . . . 301

13.2.2 Journals . . . . . . . . . . . . . . . . . . . . . . . . . 301

13.2.3 Service Providers . . . . . . . . . . . . . . . . . 301

13.2.4 Projects . . . . . . . . . . . . . . . . . . . . . . . . . . 303

13.3 Mass Integration . . . . . . . . . . . . . . . . . . . . . . . . . 304

13.3.1 Conference . . . . . . . . . . . . . . . . . . . . . . . 304

13.3.2 Journals . . . . . . . . . . . . . . . . . . . . . . . . . 304

13.3.3 Service Providers . . . . . . . . . . . . . . . . . 305

13.3.4 Projects . . . . . . . . . . . . . . . . . . . . . . . . . . 306

13.4 Combined Analysis . . . . . . . . . . . . . . . . . . . . . . 306

13.4.1 Conferences . . . . . . . . . . . . . . . . . . . . . . 306

13.4.2 Journals . . . . . . . . . . . . . . . . . . . . . . . . . 307

13.4.3 Service Providers . . . . . . . . . . . . . . . . . 307

13.4.4 Projects . . . . . . . . . . . . . . . . . . . . . . . . . . 308

13.5 Optimization for Sustainable Industry . . . . . . 309

13.5.1 Conferences . . . . . . . . . . . . . . . . . . . . . . 309

13.5.2 Journals . . . . . . . . . . . . . . . . . . . . . . . . . 310

13.5.3 Service Providers . . . . . . . . . . . . . . . . . 310

13.5.4 Projects . . . . . . . . . . . . . . . . . . . . . . . . . . 311

14 Conclusions and Further Information . . . . . . . . . . 313

14.1 Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . 313

14.1.1 Books and Key Articles . . . . . . . . . . . . 313

14.1.2 Lecture Notes and Online Teaching

Resources . . . . . . . . . . . . . . . . . . . . . . . . 315

14.2 Development Trends . . . . . . . . . . . . . . . . . . . . . 316

14.2.1 Top-Level Analysis . . . . . . . . . . . . . . . . 316

14.2.2 Maintenance Scheduling,

Maintainability, and Reliability . . . . . 316

14.2.3 Hybrid Energy Conversion Systems . . . 317

14.2.4 Integration of Renewables and Waste 317

Contents xiii

14.2.6 Energy Planning That Accounts for

Carbon Footprint . . . . . . . . . . . . . . . . . 319

14.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320

Index ....................................... 345

This page intentionally left blank

Preface

T

his book describes and analyzes an amalgamation of two

effective ways to considerably improve the efficiency and

sustainability of processing industries: Process Integration

and optimization. It is the result of collaborative efforts by two

groups: researchers at the Centre for Process Integration, the

University of Manchester, UK (a renowned center of excellence in

this field), and the Faculty of Information Technology at the

University of Pannonia, Hungary, as represented by the Centre for

Process Integration and Intensification (CPI2) and the Centre for

Advanced Process Optimization. The University of Pannonia centers

are highly regarded for their achievements in optimization, and the

P-graph and S-graph approaches to process optimization originated

at this university.

The book should provide support for graduate and postgraduate

students worldwide as well as for Continuing Professional

Development (CPD) courses and for practitioners from various fields

of the processing industry. Its chapters analyze a number of problems

of practical significance and also suggest various options for solving

them to the benefit of modern society. The book provides a wealth of

material for postgraduate teaching and further professional training.

It is supported by the expertise stemming from the authors work as

well as by a pool of case studies of varying complexity that have been

collected over the years. This wide-ranging material presented here

has been selected and refined over years of postgraduate teaching,

further CPD courses, and training for the industry. It includes eight

industry-based demonstration case studies and nine testing examples

with the solutions developed. An unbiased evaluation and overview

of the software tools available for learning, teaching, and industrial

applications are also included. Text discussions are complemented

by many figures to clarify details and enhance understanding. The

book contains 14 chapters and a comprehensive bibliography.

Chapter 1 is devoted to introducing and defining the field, and it

also includes a basic assessment of energy efficiency. It starts with

screening and scoping, which include auditing, benchmarking, and

recommendations for good housekeeping. Next it describes an

xv

xvi Preface

optimization. It is on this basis that an integrated approach to

optimization, Process Integration, is introduced. This approach is then

connected to optimal process synthesis and combinatorial graphs.

The important question of how to apply the Process Integration and

Optimization Technology arises and is dealt with. This is further

tackled and analyzed in the following chapters.

Chapter 2 deals with the basic outline and definitions of Process

Integration (PI). It begins with a historical and methodological

introduction, briefly reviewing refinements and extensions of PI over

its years of development. The thermodynamic roots of the Pinch

Technology and of targeting heat recovery are introduced, followed

by one of the key graphical constructions in the PI methodology:

Composite Curves (CCs) for targeting process heat recovery.

Supertargeting, or targeting for a full-fledged Heat Exchanger

Network (HEN), is the next logical step. These tools are used to assess

modifications of the Pinch idea for HEN retrofitting. Although PI

was initially based on Heat Integration, an important spin-off was

the development of integration for mass exchange and water

networks. The chapter concludes with remarks on the role of PI in

making industry sustainable.

Chapter 3 introduces the other key methodology used for

sustainable process design and synthesis: optimization. It presents a

general framework and workflow of model building and follow-up

optimization, including models that incorporate black boxes or

gray boxes. It is critical to understand both the meaning and the

mathematical formulation of optimization. Toward this end, the

following questions are answered: What is optimization? What are

the main classes of optimization problems? How are optimization

problems formulated mathematically? What are the conditions for

local or global optimality? This is followed by introducing

deterministic algorithms for solving continuous linear and nonlinear,

constrained and unconstrained, optimization problems. The most

popular optimization methods and algorithms that employ stochastic

search are also reviewed. The middle part of Chapter 3 is devoted to

model creation. It includes a detailed description of conceptual

modeling: extracting data about the operating units, identifying

network and topology data, constructing equations to represent the

processes, and finally choosing a right objective function. The last

part of this chapter discusses how to handle complexity and

nonlinearity as well as how to apply process insight when evaluating

model adequacy and precision.

Chapter 4 covers the core topic from which the development of PI

began: improving the energy efficiency of individual processes and

the PI extension into Total Sites. The chapter starts by introducing

heat exchange, heat recovery, and heat exchange matches so that

readers will be prepared for the remaining chapter content. The

Preface xvii

strategic view of Process Integration is outlined; this includes an

overview of the hierarchy of process design, the meaning of

performance targets, and the practical issue of identifying heat

recovery problems from process flowsheets. These three topics are

closely interrelated. Without properly applying the process design

hierarchy, the performance targets cannot find a practical application

and use. However, the process design hierarchy would be difficult to

apply without employing performance targets and estimating upper

bounds on process performance or lower bounds on total cost.

Meaningful and practically useful heat-integrated designs cannot be

obtained without appropriate identification of the heat recovery

problem, a process referred to as data extraction. The chapter

proceeds to describe the use of Composite Curves to set heat recovery

targets, the Problem Table Algorithm for numerical targeting, and

the Heat Recovery Pinch. These tools and concepts form the basis of

Pinch Technology, defining the thermodynamic capabilities of the

heat recovery problems. The more advanced aspects are discussed

next; these include threshold problems, targeting multiple utilities

via the Grand Composite Curve, and establishing targets for heat

transfer area, capital cost, and total cost. There is a short overview of

options for modifying the core process (which defines the heat

recovery problem) that highlights the usefulness of Pinch Technology

and targeting for improving its energy efficiency. The chapter then

focuses on the synthesis of Heat Exchanger Networks; the approach

mainly follows the Pinch Design Method, but there is discussion of

the superstructure-based and hybrid methods used for HEN

synthesis. The next step is Total Site Integration, which provides the

necessary knowledge for energy recovery over complete industrial

complexes and sites.

Chapter 5 deals with an extension of PI known as Mass Integration,

the most widely used instance of which is Water Integration (WI).

The chapter begins with a description of the methodology and bases

for minimizing water use and maximizing water reuse, and the

importance of legislatively imposed constraints is discussed. Best

available techniques are analyzed and recommended for usage, and

the concept of a water footprint is described. At this point, the stage

is set for the main task: minimizing freshwater usage and wastewater

effluents. For this, the methodology of Water Pinch Analysis is

introduced. Also described is a related Mass Integration and

targeting technique, the material recovery Pinch diagram. The

chapter concludes with water minimization using the mathematical

optimization approach. Both the WI and the mathematical

approaches to optimization are illustrated with case studies.

Chapter 6 addresses further PI opportunities that have arisen as

the methodology was developed. These include: Hydrogen Networks

Design and Management; Oxygen Pinch Analysis; combined analyses

(energy-water, oxygen-water, Pinch-emergy, budget-income-time,

xviii Preface

regional resource management; heat-integrated power systems with

decarbonization and low-temperature energy systems; and the

integration of reliability, availability, and maintainability with

process design. An example is given of a real-life PI problem involving

pressure-drop considerations during heat transfer enhancement.

Several recent applications are mentioned, including a Locally

Integrated Energy Sector and extended Total Sites with multiple

energy carriers.

Chapter 7 presents an overview of process optimization from the

perspectives of Mathematical Programming (MPR) and the P-graph.

The main features of these frameworks are analyzed, and it is shown

that the P-graph is better suited than MPR for solving combinatorial

optimization problems and, in particular, problems involving the

synthesis of process networks. Optimization of process scheduling is

the next topic. The most popular models and representations of

process schedules are analyzed, and an efficient tool for obtaining

them is introduced: the S-graph.

Chapter 8 presents an integrated view of PI and optimization. It

discusses how to efficiently apply them jointly in process synthesis

and how to combine them. The chapter presents a number of examples

of the P-graph and S-graph frameworks applied to combinations of

PI and optimization. These applications are grouped thematically:

(1) optimal process synthesis, including examples on reaction networks,

green biorefineries, and azeotropic distillation; (2) synthesis of

general energy systems involving Heat Integration and optimal

retrofit; (3) optimal scheduling for maximizing throughput and

revenue; (4) minimizing emissions via optimal synthesis of advanced

energy conversion systems using Fuel-Cell Combined Cycles; and

(5) availability and reliability features.

Chapter 9 reviews the software tools for process modeling,

integration, and optimization. The engineering field of sustainable

design is complex in terms of scales and relationships, which makes

information technology and computer software essential for solving

problemspreferably with a user-friendly interface. The chapter

reviews a wide spectrum of tools, as follows: (1) graph-based process

optimization (process network synthesis solutions implementing

P-graphs and the S-Graph Studio software); (2) Energy and Mass

Integration tools designed to optimize the implementation of Heat

Integration (SPRINT, HEAT-int, HEXTRAN, SuperTarget, spreadsheet-

based tools), Total Site Integration (STAR, SITE-int), power generation

and combined heat and power (STAR, WORK), and water systems

integration (WATER); (3) process flowsheeting and simulation

packages developed or supported by the major players (Aspen Plus,

HYSYS and UniSim, gPROMS, CHEMCAD, PRO/II); (4) general-

purpose optimization systems (GAMS, MIPSYN); (5) computer

algebra systems; and (6) other tools.

Preface xix

Chapter 10 is a collection of examples and case studies that

support the material presented in previous chapters. This selection

of problems was collected by the authors over decades of teaching

and consulting. Most of the problems have step-by-step solutions

that feature comments and guidance for mastering the methodology.

The examples are organized in several thematic groups: basic Heat

Integration, Total Sites, integrated placement of energy-intensive

processes, placing utilities, and Water Pinch Technology.

Chapter 11 contains advanced examples that are based on

industrial case studies, most of which were performed and published

by the authors. The case studies presented include: (1) retrofit of a

fluid catalytic cracking unit process that featured a large-scale heat

recovery network; (2) de-bottlenecking of a heat-integrated crude-oil

distillation process; (3) water system optimization of a citrus juice

plant; (4) efficient water use in other food industries; (5) synthesis of

industrial utility systems; (6) heat and power integration in buildings

and building complexes; (6) optimal supply chain design; and

(7) scheduling a large-scale paint production system.

Chapter 12 distills some important and rare expertise: typical

pitfalls and how to avoid them. It addresses possible (and probable)

difficulties encountered during problem formulation and data

extraction. The chapter raises seemingly trivial yet fundamental

questions: When is a stream a stream? How precise must the data be

at each step? How can extreme changes in specific heat capacity be

handled? What rules and guidelines must be followed to properly

extract data? How can the heat loads, heat capacities, and temperatures

of an extracted stream be calculated? How soft are the data in a

plant or process flowsheet? How can capital costs and operating costs

be estimated? The provided answers are based on long-term

experience and could help readers to solve the right problemthat

is, a problem that accurately reflects the reality of the process

under consideration. The chapter also addresses the integration of

renewables that exhibit fluctuating demand and supply rates as well

as steady-state and dynamic performance considerations. The chapter

concludes with recommendations for interpreting results and

suggestions on the successful advocacy and implementation of

sustainable and optimal design.

Chapter 13 consists solely of information on sources for further

reading and information gathering. The various sources of

information about optimization and integration in the process

industry are arranged as follows: (1) general sources of information,

(2) Heat Integration, (3) Mass Integration, (4) combined analysis, and

(5) optimization for sustainable industry. Each topic is subdivided

into sections on conferences, journals, service providers, and

projects.

Chapter 14 presents conclusions and sources of further

information. Suggestions for further reading point to books and

xx Preface

adequately described in a book of this size. The book ends with a

comprehensive bibliography that provides details of works cited in

the text and serves also as a source of information and directions for

further study.

This book is intended to increase awareness of the principal

methodologies that could contribute to improving energy and water

efficiency in processing plants while reducing their environmental

impact. Many additional illustrative examples could well have been

discussed, but space limitations required that the authors select the

most important features from a variety of fields and processing

industries; even so, hundreds of references are included for seekers

of more information. If the authors have managed to raise the interest

and awareness of readers, then this book has fulfilled its intentions.

Ji Kleme

Ferenc Friedler

Igor Bulatov

Petar Varbanov

Acknowledgments

W

e acknowledge the support from the European

Communityfunded project entitled Integrated Waste to

Energy Management to Prevent Global Warming, or

INEMAGLOW. We would like to thank Prof. Robin Smith and

Mr. Simon Perry from the University of Manchester for numerous

discussions about the finer points of PI and optimization

methodology. We also acknowledge with much gratitude the input

of all collaborators who contributed to this book. Their dedication,

timely responses, and willingness to accept editorial comments and

suggestions are greatly appreciated. We received invaluable help

from the staff of the Faculty of Information Technology at the

University of Pannonia in Hungary: Dr. Rozlia Pigler-Lakner,

Dr. Istvn Heckl, Dr. Boton Bertk, Dr. Zoltn Sle, Mr. Mt

Hegyhti, and Ms. Adrienn Sas. Substantial contributions were also

made by Hon Loong Lam and Zsfia Fodor, Ph.D. students

at CPI2.

Special thanks go to colleagues and close collaborators

worldwide who shared their latest methodologies and case studies.

It would be impossible to list all those involved, but we should like

to mention a few: Prof. L. T. Fan, Department of Chemical

Engineering, Kansas State University, Manhattan, Kansas, USA; Dr.

Dominic C. Y. Foo, Department of Chemical and Environmental

Engineering, University of Nottingham, Malaysia Campus; Prof.

Zdravko Kravanja, Faculty of Chemistry and Chemical Engineering,

University of Maribor, Slovenia; Prof. Valentin Plesu, Centre for

Technology Transfer for the Process Industries, Department of

Chemical Engineering, University POLITEHNICA, Bucharest; and

Prof. Petr Stehlk, Brno University of Technology, Institute of

Process and Environmental Engineering, UPEI VUT Brno, the

Czech Republic.

The authors appreciate the editing and production efforts of the

following people: Taisuke Soda, Michael Penn, Stephen Smith,

Richard Ruzycka, and Michael Mulcahy of McGraw-Hill, and

Aloysius Raj and the staff at Newgen.

It has been a great pleasure for us to work with all of you.

xxi

This page intentionally left blank

CHAPTER 1

Introduction and

Definition of

the Field

1.1 Introduction

In recent years there has been increased interest in the development

of renewable, non-carbon-based energy sources in order to combat

the increasing threat of carbon dioxide (CO2) emissions and subsequent

climatic change. More recently, the fluctuations and often large

increases in the prices of oil and gas have further increased interest

in employing alternative, lower-carbon or non-carbon-based energy

sources. These cost and environmental concerns have led to increases

in the industrial sectors efficiency of energy use, although the use of

renewable energy sources in major industry has been sporadic at

best. In contrast, domestic energy supply has moved more positively

toward the integration of renewable energy sources; this movement

includes solar heating, heat pumps, and wind turbines. However,

there have been only limited and ad hoc attempts to design a

combined energy system that includes both industrial and residential

buildings, and few systematic design techniques have been marshaled

toward the end of producing a symbiotic system.

This book provides an overview of the Process Integration and

optimization methodologies and its application to improving the

energy efficiency of not only industrial but also nonindustrial energy

users. An additional aim is to evaluate how these methodologies can

be adapted to include the integration of waste and renewable energy

sources.

Industrial production requires a considerable and continuous

supply of energy delivered from natural resourcesprincipally in

the form of fossil fuels such as coal, oil, and natural gas. The increase

in our planets human population and its growing nutritional

1

2 Chapter One

Furthermore, many nations have accelerated their development in

the last 10 years, and countries (such as China and India) with large

populations have seen significant increases in energy demands.

This growing energy consumption has also resulted in unsteady

climatic and environmental conditions in many areas because of

increased emissions of CO2, NOx, SOx, dust, black carbon, and

combustion process waste (Kleme et al., 2005a; Kleme, Bulatov,

and Cockeril, 2007). It has become increasingly important to ensure

that the production and processing industries take advantage of

recent developments in energy efficiency and in the use of

nontraditional energy sources (Houdkov et al., 2008; Lam,

Varbanov, and Kleme, 2010). The additional cost is related to the

amount of emitted CO2 and often takes the form of a centrally

imposed tax. A workable solution to this problem would be to

reduce emissions and effluents by optimizing energy consumption,

increasing the efficiency of materials processing, and also increasing

the efficiency of energy conversion and consumption (Kleme et al.,

2005b).

Although major industry requires large supplies of energy to

meet production, it is not the only sector of the world economy that is

increasing its demands for energy. The particular characteristics of

the other sectors (e.g., transport, residential) make optimizing for

energy efficiency and cost reduction more difficult than in traditional

processing industries, such as oil refining, where continuous mass

production concentrated in a few locations offers an obvious potential

for large energy savings (Al-Riyami, Kleme, and Perry, 2001). In

contrast, for example, agricultural production and food processing

are distributed over large areas, and these activities are not continuous

but rather structured in seasonal campaigns. Hence, energy demands

in this sector are related to specific and limited time periods, so the

design of efficient energy systems to meet this demand is more

problematic than in traditional, steady-state industries.

This chapter proceeds by first outlining the field of energy

efficiency, including its scope, actors, and main features. The next

step is to describe energy-saving techniques generally and then to

specify an integrated approach: Heat Integration. An increasingly

prominent issue is assessing and minimizing emissions and the

carbon footprint. The carbon footprint (CFP) is defined by the U.K.

Parliamentary Office for Science and Technology as the total amount

of CO2 and the other greenhouse gases emitted over the full life cycle

of a process or product (POST, 2006). There have been numerous

studies (see, e.g., Albrecht, 2007; Fiaschi and Carta, 2007) that

emphasize the carbon neutrality of renewable sources of energy.

However, even renewable energy sources make some contribution to

the overall carbon footprint, and assessment studies frequently do

not account for this. The carbon footprint should also be incorporated

Introduction and Definition of the Field 3

into any products life-cycle assessment (LCA); see, for example,

Masruroh, Li, and Kleme (2006).

The task of saving energyespecially at a time of rising energy

costs, demand, and carbon emissionsmust be taken seriously by

all communities and industries. Society is driven by the economics

of individual situations, and no section of society worldwide can

be expected to save energy at any cost. Thus, energy-saving

measures must be considered within the context of such issues

as environmental factors, legislatively imposed constraints, and

pressure from conscientious consumers. The simplest and most

obvious technique involves energy auditing and applying good

housekeeping measures. In many cases even these simple measures

are not fully understood or completed in sufficient detail. To

undertake a worthwhile energy audit, correct measurements are

necessary. Also, because in many cases energy demand is not

constant and instead fluctuates considerably, the monitoring of

energy consumption has to be performed over specific (or extended)

periods of time. Recommended monitoring techniques are described

by various sources: utility companies, such as SEMPRA ENERGY

(2009); governmental agencies, such as the U.S. Department of

Housing and Urban Development (2009); and international groups,

such as the International Energy Agency (Mandil, 2005).

Improvements in energy efficiency must often be achieved by

more complex means, such as those associated with improved

design and operation. It is of paramount importance that all

energy-related processes operate with maximum efficiency and

minimum energy input. These systems should also ensure that

they are fueled as much as possible by low-value inputs or recycled

wastes, such as process outputsfor example, off-gases and hot-

water waste (AEA Technology, 2000). To ensure that systems are

designed to be as efficient as best practice allows, optimization

methods are frequently employed for grassroots design, retrofit,

control, and intelligent support systems for processes, plants, and

buildings. One technology that has a strong reputation for

improving energy efficiency through better design is Pinch

Technology (Linnhoff and Vredeveld, 1984), which has been in use

for more than 20 years. This technology, through feedback from

practical applications and industry professionals, has been

continuously developed and expanded (Kleme et al., 1997; Smith,

2005; Kleme, Smith, and Kim, 2008). Details on the successful

applications of Pinch Technology in various industrial sectors are

described in Chapter 11.

The sustainability of energy systems can also be considerably

improved by making use of renewable energy sources (e.g., biofuels,

4 Chapter One

greenhouse gases. Implementing Combined Heat and Power (CHP)

systems (AEA Technology, 2000), rather than a separate power system

and heat system, can also substantially improve the efficiency of

energy supply. In addition, the overall situation can be improved by

certain fast-advancing technologies: heat pumps, compact heat

exchangers, fuel cells (FCs), and intensified technologies. Some of

these approaches are not yet fully commercialized but are gradually

becoming available. Some examples discussed in AEA Technology

(2000) are as follows:

Advanced gas turbines for both utility and industrial

applications, including cogeneration (CHP).

Fuel cells are electrochemical devices that may be fueled by

hydrogen, methane, or other organic fuels. High-temperature

FCs (MCFC and SOFC) can also use cleaned and conditioned

synthesis gas directly. These systems produce high-grade

heat (above 500C) in addition to electrical power, and they

are well suited to cogeneration. It is estimated that FCs

typically emit 25 percent less CO2 than a gas turbine. Yet

further advances are required before their full application

becomes economically practical. One option is to integrate

CHP and FCs (Varbanov et al., 2006).

Dividing-wall distillation technology (Triantafyllou and Smith,

1992; Hernndez and Jimnez, 1999). This technology

involves the separation of three components (or groups of

components) in a mixture. In the past this would have

required two distillation columns, with heating and cooling

provided for each column. The dividing-wall technology

combines the separation process into a single vessel to yield

energy savings of about 30 percent and capital savings of

about 25 percent (MW Kellogg, 1998).

Compact heat exchangers are generally made of thin metallic

plates rather than tubes. The plates form complex and small

flow passages that result in a large surface area for heat

transfer per unit volume. Multistream versions of these

exchangers can incorporate 12 or more streams. Compact

heat exchangers can yield energy savings and also reduce the

costs of capital and installation. In a case study at a U.K.

refinery, potential capital savings ranged from 69 to 84 percent

(EEO, 1993).

Cogeneration is being increasingly applied in most sectors. For

example, many oil refineries satisfy a large portion of their power

demands by on-site generation, with the balance being supplied by

externally purchased electricity. Usually all or almost all heating

Introduction and Definition of the Field 5

needs are met by on-site generation of heat carriers (hot oil, steam,

flue gases). CHP generation and even tri-generation (simultaneous

production of heat, power, and cooling) offer an opportunity to

reduce greenhouse gas emissions from the combined power grid

refinery system by utilizing fuel heat content more completely than

do most existing power generation technologies. The improved

utilization rate is achieved by recovering the heat left in the exhaust

streams of the various power generating facilities (gas turbines,

FCs).

and Good Housekeeping

Over the years, screening and scoping tools have had a considerable

effect on reducing the costs of energy and of treating effluents,

thereby improving plant profit margins. For example, energy audits

performed on various food and drink processes have resulted in cost

savings of 1530 percent and in attractive returns on investment

(NRCan, 2007; U.S. DOE, 2007). Because profit margins are generally

small in this sector, efficient management of energy is crucial for

increasing profits while simultaneously reducing the production

plants environmental impacts.

The Carbon Trust (2009) has suggested the following steps for

reducing energy consumption and thus improving energy efficiency.

An analogous approach can be used for optimizing the use of water

and wastewater:

(2) locating heat leakages; (3) preventive maintenance;

(4) insulation; (5) justifying use of heating, cooling, and

lighting; (6) prevention or reduction of fouling; and

(7) monitoring and control.

Energy audits: (1) examining records of energy cost and

consumption; (2) producing an energy balance sheet;

(3) providing high-quality data on energy consumption and

costs; (4) collecting and processing data regularly

(recommended for the analysis and review of energy

information); and (5) establishing a benchmark of energy

consumption based on other organizations or accepted

standards.

Energy Efficiency Environmental Management System, ISO

14001: (1) the management system is a network of interrelated

elements; (2) these elements include responsibilities,

authorities, relationships, functions, processes, procedures,

practices, and resources; (3) the management system

6 Chapter One

applying the policies to achieve the objectives.

Responsible use of energy: in (1) energy procurement,

(2) metering and billing, (3) performance measurement,

(4) policy development, (5) assigning energy management

responsibility, (6) energy surveying and auditing, (7) training

and education, and (8) capital project management.

The steps listed here have been found by the U.K. Energy Efficiency

Best Practice Programme (EEBPp, 2002) to be central to any resource

and waste management program. Energy screening and scoping

audits fall into three main categories whose use (either individually

or in combination) is based on the required depth of the study, the

process to be analyzed, and the plant size: (1) the walk-through audit,

which provides a quick snapshot of certain opportunities; (2) the

detailed audit, which conducts an in-depth analysis of specific

components; and (3) the Process Integration audit, which analyzes the

plant as a whole and takes a systematic look at all processing steps and

their interconnections.

Audits may be performed by plant personnel, by external experts,

or by teams with members from both groups. Although there are

many possible options and levels of detail, typical activities include

the following (NRCan, 2007):

reference period.

2. Collection of energy total consumption and cost (information

usually available from fuel records, electricity invoices, etc.).

3. Development of a process flow chart that shows materials,

energy inputs, and energy outputs for the main processing

steps.

4. For the largest consumers, collection of energy data from

the plant metering devices, control systems, and process

flow diagrams (if current operating conditions are close to

design data).

In terms of the time required for (and other costs of) identifying

opportunities to save energy, one efficient approach is the top-level

analysis (Varbanov et al., 2004). This procedure accounts for investment

limits or allowances and identifies economically justifiable energy-

saving opportunities, listing them in the order of their expected

economic return. Measurements can be performed using portable

instruments (flow rate, temperature, humidity, etc.) to determine the

overall plant production of steam, refrigeration, compressed air, and

hot water. Interviews with key personnel and operators can also

provide valuable information about plant operations.

Introduction and Definition of the Field 7

as a Basis for Optimization

Balancing reconciliation and flowsheeting simulation tools are

frequently used for sustainability design and savings analysis; in

fact, they have become the main tools in a process engineers toolbox.

These tools help engineers to develop complete material and energy

models based on measurements and/or design values and

mathematical models. Consequently, these simulation tools play an

important role in the technical and economic decision-making

activities related to the planning and/or design stage of processes

under development and to the operation of existing equipment.

A number of computer-based systems have been developed to

help process engineers calculate energy and mass balances. However,

ongoing development costs have left only a few of these systems on

the market: those whose positions have been secured by a substantial

number of sales. An early overview of flowsheeting simulation was

presented by Kleme (1977). The balancing, data validation, and

reconciliation technology involves a set of procedures incorporated

into a software tool. Process data reconciliation has become the main

method for monitoring and optimizing industrial processes as well

as for component diagnosis, condition-based maintenance, and online

calibration of instrumentation. According to Heyen and Kalitventzeff

(2007), the main goals of this technology are: (1) to detect and correct

deviations and errors of measured data so that they satisfy all balance

constraints; (2) to use knowledge about the process system and

structure along with measured data to compute unmeasured data

wherever possible, especially key performance indicators (KPIs); and

(3) to determine the postprocessing accuracy of measured and

unmeasured data, including KPIs. More information about available

software tools is provided in Chapter 9.

Heat Integration is the first part of Process Integration, which

provides the design foundation for CHP systems, refrigeration, air

conditioning, and heat-pump systems. Process Integration is equally

applicable to small, medium, and large industrial sites (e.g.,

power stations and oil refineries engaged in the production of

petrochemicals). The technology answers one of the major challenges

in the design of heating and cooling systemsnamely, the complexity

of energy and power integrationvia a mapping strategy based on

thermodynamically derived upper bounds on the systems thermal

and power performance. The efficient use of available heating and

cooling resources for serving complex systems of various sizes

and designations can significantly reduce energy consumption and

emissions. This methodology can also be used to integrate renewable

8 Chapter One

thermal into the combined heating and cooling cycles. Since 1995, the

energy consumption of European Community (EC) member countries

has risen by 11 percent to the equivalent of 1637 Mt (megatons) of oil

equivalent (Eurostat, 2007). This increase in energy consumption

contrasts with the trend of the EC population, which is growing at

only about 0.4 percent annually (Eurostat, 2007). The overall share of

total energy consumption by industry is declining in most countries.

However, domestic energy consumption is rising. In the United

Kingdom, for example, residential consumption rose from 35.6 Mt

(oil equivalent) in 1971 to 48.5 Mt in 2001an increase of 36 percent

despite increases in energy efficiency (DTI, 2006).

Process Integration Technology (Pinch Technology) has been

extensively used in the processing and power generating industry

for more than 30 years. It was pioneered by the Department of Process

Integration, UMIST (now the Centre for Process Integration, CEAS,

the University of Manchester), in the late 1980s and 1990s. Heat

Integration is introduced in Chapter 2 and is described in more detail

in Chapter 4. Water and mass integration is covered in Chapter 5, and

recent developments in the field are reviewed in Chapter 6.

Combinatorial Graphs

Process synthesis is a complex engineering activity that involves

process modeling (e.g., chemical engineering) as well as combinatorial

challenges. Although the basic process modeling has reached a

considerable level of maturity, the combinatorial aspects of the

engineering problem still leave significant room for improvement.

One innovative approach to process synthesis is to exploit the

combinatorial nature of network optimization. This approach is used

by the process-graph (P-graph) framework, which explicitly defines

sets of process materials and operations and then uses efficient

combinatorial algorithms to build a rigorous network superstructure

that can be reduced to the optimal network topology. This is different

from the Mathematical Programming (MPR) approach, where the

combinatorial aspects are modeled by algebraic equations and the

structural features are blended with the underlying process models.

These approaches are covered in Chapters 3, 7, and 8.

The P-graph framework has been successfully applied to and

demonstrated on several cases of energy system design. For example,

Varbanov and Friedler (2008) explored FC-based systems in a case

study that evaluated energy conversion systems to reduce CO2

emissions via Fuel-Cell Combined Cycle (FCCC) subsystems that

utilize biomass and/or fossil fuels. The combinatorial complexity of

the problem is efficiently handled by using P-graph framework and

Introduction and Definition of the Field 9

algorithms. The authors developed a methodology for synthesizing

cost-optimal FCCC configurations that accounts for the system

carbon footprint. Their results show that the high energy efficiency

of such systems when using renewable fuels makes them economically

viable for a wide range of conditions. This study and other case

studies (and guided applications) are provided in Chapters 10

and 11.

Optimization Technology

The crucial topic of applications is addressed in Chapter 12, and

Chapters 13 and 14 discuss sources of further information for those

who would like to learn more or who seek qualified help from

leading researchers and providers worldwide. Although every effort

has been made to ensure that the provided information is as

comprehensive as possible, it cannot be fully exhaustive. The field of

Process Integration and optimization is developing rapidly, and

every month brings important advancements. For this reason, the

reader is strongly encouraged to remain up-to-date by reading about

and exploring new developments.

This page intentionally left blank

CHAPTER 2

Process Integration

Energy and water saving, global warming, and greenhouse gas

emissions have become major technological, societal, and political

issues. These issues are of strategic importance because they are all

closely related to energy supply. Numerous studies have been

performed on the subject of improving energy efficiency while

reducing emissions of greenhouse gases, volatile organic compounds,

and other pollutants.

In response to these industrial and societal requirements, several

novel methodologies emerged since 1970. They included process system

engineering (Sargent, 1979; Sargent, 1983) and Process Integration

(Linnhoff et al., 1982; Linnhoff et al., 1994) followed by a number of

works from the UMIST Group. Both disciplines were involved in dedicated

conferences such as ESCAPE (European Symposium on Computer

Aided Process Engineering), which was facilitated by the European

Federation of Chemical Engineering Working Party on Computer

Aided Process Engineering (CAPE, 2009), and PRES (Conference on

Process Integration, Modelling and Optimisation for Energy Saving

and Pollution Reduction; PRES, 2009), which is supported on an annual

basis by chemical and chemical engineering societies (e.g., Hungarian

Chemical Society, Czech Society of Chemical Engineering, Italian

Association of Chemical Engineering, Canadian Society for Chemical

Engineering). It has gradually become evident that resource inputs and

effluents of industrial processes are often connected to each other.

Examples of this connection include the following:

an equivalent reduction in the cooling utility demand

(Linnhoff and Flower, 1978; Linnhoff et al., 1982; Linnhoff

et al., 1994); obviously, this also tends to reduce the CO2

emissions from the corresponding sites.

2. Reducing wastewater effluents usually leads to reduced

freshwater intake (Wang and Smith, 1994; Bagajewicz, 2000;

Thevendiraraj et al., 2003).

11

12 Chapter Two

increasing internal recycling and reuse of energy and material

streams instead of fresh resources and utilities. Projects for improving

process resource efficiencies can offer economic benefits and also

improve public perceptions of the company undertaking them.

However, motivating, launching, and carrying out such projects

requires proper optimization studies that are based on adequate

models of the process plants.

Process Integration (PI) is a family of methodologies for combining

several processes to reduce consumption of resources and/or to

reduce harmful emissions. It started as mainly Heat Integration (HI),

stimulated by the energy crisis of the 1970s (Hohmann, 1971; Linnhoff

and Flower, 1978; Linnhoff, Mason, and Wardle, 1979; Linnhoff et al.,

1982; Linnhoff and Hindmarsh, 1983; Linnhoff and Vredeveld, 1984).

This energy-saving methodology has been used extensively in the

processing and power generating industry over the last 30 years.

Heat Inetegration examines the potential for improving and

optimizing the heat exchange between heat sources and heat sinks in

order to reduce the amount of external heating and cooling required,

thereby reducing costs and emissions. A systematic, rule-based

design procedure has been developed that yields the maximum

energy-saving design for a given system.

There are several definitions of HI (using Pinch Technology);

most refer to the thermal combination of steady-state process streams

or batch operations for achieving heat recovery via heat exchange.

More broadly, the definition of PI, as adopted by the International

Energy Agency (Gundersen, 2000) is as follows: systematic and

general methods for designing integrated production systems

ranging from individual processes to total sites and with special

emphasis on the efficient use of energy and reducing environmental

effects.

It is remarkable that PI continues to interest researchers even 35 years

after its emergence. The HI, which developed as the first part of PI,

deals with the integration of heat in Heat Exchanger Networks (HENs).

This methodology has been shown to have considerable application

potential for complete chemical processing sites, reducing overall

energy demand and emissions across the site and thus leading to a

more effective and efficient site utility system. The PI and its subset HI

method have also been successfully applied to the cogeneration of

heat and shaft power. Further details are available elsewhere

Process Integration 13

(Linnhoff et al., 1982; Linnhoff et al., 1994; Shenoy, 1995; Smith, 2005;

El-Halwagi, 2006; Kemp 2007; Kleme, Smith, and Kim, 2008).

One of the first works in this field was Hohmanns (1971) PhD

thesis, which introduced a systematic thermodynamics-based

reasoning for evaluating the minimum energy requirements for a

given HEN synthesis problem. In the late 1970s this work was

continued by Linnhoff and Flower, who used Hohmanns foundation

to develop the basis of Pinch Technologynow considered the

cornerstone of HI. As is often the case with a pioneering innovation,

this work was difficult to publish. Yet the authors strong commitment

eventually led to the publication of their ideas in Linnhoff and

Flower (1978), which has since become the most cited paper in the

history of chemical engineering. Similar work (Umeda et al., 1978;

Umeda, Harada, and Shiroko, 1979) was independently published in

Japan, but it was Linnhoff (supported by teams from UMIST and

later Linnhoff March Ltd.) who pushed the new concept through

academia and industry. The publication of the first red book by

Linnhoff et al. (1982) played a key role in the dissemination of HI

methodology. This users guide to Pinch Analysis detailed the most

common process network design problems, including HEN synthesis,

heat recovery targeting, and selecting multiple utilities.

These methodologies were developed and pioneered by the

Department of Process Integration, UMIST (now the Centre for

Process Integration, CEAS, the University of Manchester) in the late

1980s and 1990s (Linnhoff et al., 1982; Linnhoff and Vredeveld, 1984;

Linnhoff et al., 1994; Kleme et al., 1997; Smith et al., 2000; Smith,

2005). A second edition of Linnhoffs users guide was published by

Kemp (2007). Applications of HI in the food industry were presented

in Kleme and Perry (2007a) and in Kleme, Smith, and Kim (2008).

Tan and Foo (2007) successfully applied the Pinch Analysis approach

to carbon-constrained planning for the energy sector, and Foo, Tan,

and Ng (2008) applied the cascade analysis technique to carbon-

footprint-constrained energy planning.

Another important part in process design and optimization is

the synthesis phase of process flowsheets. From the earliest stages of

PI there have been attempts to combine it with optimization (see, e.g.,

Giammatei, 1994). Such combining is usually performed after the

targeting phase mentioned previously. Ideally, the structure of the

entire processand the configurations of the operating units within

itshould be simultaneously designed and optimized, because the

performance of each unit influences the others. The main source of

complexity in this synthesis is the problems dual nature of being

both continuous and discrete. There are several known methods for

performing the task, including heuristic, evolutionary, and

superstructure-based approaches. Two major classes of methods for

process synthesis are heuristic and algorithmic (or Mathematical

14 Chapter Two

these approaches incorporate heuristic rules as well as Mathematical

Programming (MPR). Much as with any decision making, process

synthesis (or design) is an activity for which no past experience can be

ignored, especially when it comes to localized details of the design.

The most popular approach is to create a superstructure for the

network being designed and then choose the best possible solution

network from the superstructure options.

The Thermodynamic Roots

Furman and Sahinidis (2002) have compiled a comprehensive review

of works tracing the development of the research on HENs in time.

Their study shows that there was only mild interest in heat recovery

and energy efficiency until the early 1970s, by which time just a few

works in the field had appeared. But between the oil crises of 1973

1974 and 1979 there were significant advances made in HI. Although

capital cost remained important, the major focus was on saving

energy and reducing related costs. It is exactly this focus that

resulted in attention being paid to energy flows and to the energy

quality represented by temperature. The result was the development

of Pinch Technology, which is firmly based on the first and second

laws of thermodynamics (Linnhoff and Flower, 1978).

In this way HEN synthesisone of the most important and

common tasks of process designhas become the starting point for

the PI revolution in industrial systems design. HENs in industry are

used mainly to save on energy cost. For many years the HEN design

methods relied mostly on heuristics, as necessitated by the large

number of permutations in which the necessary heat exchangers

could be arranged. Masso and Rudd (1969) is a pioneering work that

defines the problem of HEN synthesis; the paper proposes an

evolutionary synthesis procedure that is based on heuristics. An

alternative HEN synthesis method is described in Zhelev et al.

(1985), one that exploits synergies between heuristics and

combinatorics. In this approach, several suboptimal networks are

synthesized, and in most cases they are less integrated (consist of

more than one subnetwork). A complete timeline and thorough

bibliography of HEN design and optimization works is provided in

Furman and Sahinidis (2002). The paper covers many more details,

including the earliest known HEN-related scientific article: Ten

Broeck (1944).

The discovery of the Heat Recovery Pinch concept (Linnhoff

and Flower, 1978) was a critical step in the development of HEN

synthesis. The main idea behind the formulated HEN design

procedure was to obtainprior to the core design stepsguidelines

Process Integration 15

and targets for HEN performance. This procedure is possible

thanks to thermodynamics. The hot and cold streams for the

process under consideration are combined to yield (1) a Hot

Composite Curve representing, collectively, the process heat

sources (the hot streams); and (2) a Cold Composite Curve

representing the process heat sinks (the cold streams). For a

specified minimum allowed temperature difference Tmin, the two

curves are combined in one plot (see Figure 4.7), providing a clear

thermodynamic view of the heat recovery problem.

The overlap between the two Composite Curves represents the

heat recovery target. The overlap projection on the heat exchange

axis represents the maximum amount of process heat being internally

recovered. The vertical projection of the overlap indicates the

temperature range where the maximum heat recovery should take

place. The targets for external (utility) heating and cooling are

represented by the nonoverlapping segments of the Cold and Hot

Composite Curves, respectively. The methodology is described in

more detail in Chapter 4.

After obtaining targets for utility demands of a HEN, the next logical

step is to estimate targets for capital and total costs. Capital costs in

HENs are determined by many factors, of which the most significant

is the total heat transfer area and its distribution among the heat

exchangers. Townsend and Linnhoff (1984) proposed a procedure for

estimating HEN capital cost targets by using the Balanced Composite

Curves, which are obtained by adding utilities to the Composite

Curves obtained previously (see Figure 4.7). The HEN heat transfer

area target is computed from the enthalpy intervals in the Balanced

Composite Curves by using the heat transfer coefficients given in the

HEN problem specification (assuming vertical heat transfer and

spaghetti-type topology). Improvements to this procedure that

have been proposed involve one or more of the following factors:

that exhibit nonuniform heat transfer coefficients (Colberg

and Morari, 1990; Jegede and Polley, 1992; Zhu et al., 1995;

Serna-Gonzlez, Jimnez-Gutirrez, and Ponce-Ortega,

2007).

2. Accounting for practical implementation factors, such as

construction materials, pressure ratings, and different heat

exchanger types (Hall, Ahmad, and Smith, 1990).

3. Accounting for additional constraints such as safety and

prohibitive distance (Santos and Zemp, 2000).

16 Chapter Two

predicted profitability. Taal and colleagues (2003) summarized the

common methods used for cost estimation of heat exchange

equipment and the sources of energy price projections. This paper

showed the importance of choosing the right cost estimation method

and a reliable source of energy price forecast when retrofit projects

are evaluated for viability. Both retrofit and grassroots design projects

require accurate cost estimates in the projects early stages so that the

correct choices are made. There are several methods available, most

of which lead to FOB (Free On Board) cost estimates. As for operational

costs, the range of possible future energy prices may prove crucial

when the margins of a proposed design or retrofit project are

analyzed. This is especially true for a large project that has a long

payback period and consumes a lot of energy. Such projects are

typical of plants in the chemical, petrochemical, refinery, and

paper industries. Taal et al. (2003) include a brief review of oil and

natural gas price projections reported by centers of excellence in the

energy economics field. There are numerous forecasts, sometimes

contradictory, but general trends can be observed. Some general

insights in oil and gas price and production behavior are mentioned

and refer to more detailed sources.

Bochenek, Jezowski, and Jezowska (1998) compared the approaches

of optimization versus simulation for retrofitting flexible HENs.

This is an important work that should have generated additional

research. Zhu, Zanfir, and Kleme (2000) proposed a heat transfer

enhancement methodology, for HEN retrofit design, from which HI

could benefit substantially. This approach is worthy of wider

implementation, especially in the context of retrofit studies.

Heat Exchanger Network retrofit is a special case of optimization.

In retrofit problems, one must accommodate an existing network

with existing heat exchangers that are already paid for. This

circumstance substantially alters the economics of the problem as

compared with a new design. One example of this approach to

retrofit was the paper of Tjoe and Linnhoff (1986), which suggested

identifying heat exchangers with cross-Pinch heat transfer and,

where appropriate, attempting to replace such heat exchangers with

others that do not transfer heat across the Pinch, thereby reducing

energy consumption.

However, an ideal new design would not account for the existing

HEN equipment and topological constraints. One method for

overcoming these drawbacks is the Network Pinch (Asante,

1996; Asante and Zhu, 1997), which uses continuous nonlinear

optimization to identify the bottlenecking heat exchangers within

Process Integration 17

the existing network. The network bottleneck occurs at a heat

exchanger, which constrains the load shifting to gain further

improvement in heat recovery. At least one of the heat exchanger

sides exhibits the minimum allowed temperature difference between

the involved streams, and the corresponding point in the networks

temperature-load plot is referred to as the Network Pinch. To

overcome this Network Pinch, the network structure must be

modified. Possible modifications include the relocation of an existing

heat exchanger, the addition of a new exchanger, or a change in the

stream splitting arrangement. To identify the most promising

modifications, the Network Pinch method uses MPR guided by

thermodynamic insights. Nonlinear optimization is employed to

evaluate the capital energy trade-offs and to produce a new optimal

structure. This approach allows retrofit to be carried out one step at

a time, leaving the designer in control to accept or reject suggested

modifications at each step.

There have been some successful practical applications of these

targeting and retrofit methodologies. Pleu, Kleme, and Georgescu

(1998) demonstrated the wide applicability of PI in the Romanian

oil refining and petrochemical industry. In one of the first

comprehensive retrofit case studies, Hassan, Kleme, and Pleu

(1999) presented a PI analysis and retrofit suggestions for a fluid

catalytic cracking (FCC) plant. Pinch Technology and its recent

extensions offer an effective and practical method for designing the

HEN for new and retrofit projects. Al-Riyami, Kleme, and Perry

(2001) demonstrated a HI retrofit analysis for the HEN of an FCC

plant. Their study found significant room for improvement in the

heat recovery process, and the new network was designed using

the Network Pinch method.

Water is widely used in the processing industries as an important

raw material. It is also frequently used as a utility (e.g., steam or

cooling water) and as a mass transfer agent (e.g., for washing or

extraction). Large amounts of high-quality water are consumed in

many industries that face strict requirements for product quality and

the associated manufacturing safety issues. The processing industry

is characterized by complex design and operation of storage and

distribution systems for water management. Todays industrial

processes and systems that use water are subjected to increasingly

stringent environmental regulations on the discharge of effluents.

Increases in population and its quality of life have led to increased

demand for freshwater. The rapid pace of these changes has made

more urgent the need for improved water management. Adopting

techniques to minimize both water consumption and wastewater

18 Chapter Two

the amount of effluent generated by processing.

For these reasons, the success of HI has inspired researchers to

apply the Pinch and PI concepts to other areasin particular, to

mass exchange networks (El-Halwagi and Manousiouthakis, 1989).

Wang and Smith (1994) developed a method for industrial water

networks as a special case of mass exchange networks (see Figure 5.1).

Their main objective was to minimize the consumption of freshwater

and the disposal of wastewater simultaneously by maximizing the

reuse of internal water, again exploiting the idea of recycling and

reusing valuable streams and materials in order to save resources

and reduce emissions. Wastewater can be further reduced by

applying additional techniques for water regeneration that enable

further reuse or recycling. For the case of a single contaminant,

translating the method of Pinch Analysis to water minimization is

straightforward: the waters Composite Curve is used to construct a

plot of contaminant concentration versus contaminant load.

Extending the Water Pinch Analysis to multiple-contaminant

problems is a complicated and difficult procedure. The principal

issue concerns determining which contaminant to use on the Y axis

when plotting the Composite Curves. Several approaches have been

proposed. One option is to employ MPR, in which case Water Pinch

serves as a preliminary visualization tool (Doyle and Smith, 1997).

Foo, Manan, and Tan (2005) applied Water Pinch Analysis to

synthesize optimal water recovery networks for batch processes.

These authors introduced a numerical technique (time-dependent

water cascade analysis) that has the advantage of clearly depicting

the time-dependent nature of batch water networks. Majozi (2005)

also employed mathematical modelingthe mixed integer nonlinear

programming (MINLP) approachto devise an effective technique

for wastewater minimization in batch processes. Here, too, wastewater

was minimized via application of reuse and recycle.

Heat recovery targeting for HEN synthesis problems is based on

Composite Curves described in Section 2.4 (Linnhoff, Mason, and

Wardle, 1979). The Composite Curves plot is a visual tool that

summarizes the important energy-related properties of a process in

a single view (see Figure 4.7). It was the resulting recognition of the

thermodynamic relationships and limitations in the underlying

heat recovery problem that led to development of the Pinch Design

Method (Linnhoff and Hindmarsh, 1983), which is capable of

producing maximally efficient heat recovery networks. As already

discussed in this chapter, PI has been considerably expanded in

scope since these initial applications. It is now also used for HEN

Process Integration 19

retrofits, both water and combined water-and-energy minimization,

as well as to minimize total site energy consumptiona process

that includes Combined Heat and Power, locally integrated energy

sectors, integration of renewables and waste-to-energy techniques,

and combinatorial tools (P-graph and S-graph; see Chapter 7). In

addition, recent applications have extended the PI approach to

regional energy and emissions planning, financial planning, batch

processes, and the targeting of other constrained resources, such as

land, renewable energy, and emissions. In the wake of the initial

breakthrough of Pinch Analysis for HEN synthesis, all these new PI

applications follow the same simple logic: target setting should

precede designing. In the most straightforward casessuch as HEN

synthesis for Maximum Energy Recovery (MER) and water network

synthesis for maximum water reusethe targets can be interpreted

as indicators of what a rigorous application could actually achieve.

However, the applicability and benefits of PI are not limited to

these straightforward cases. In fact, the target setting can be applied

in various contexts and still yield enormous benefits in terms of

reduced computational and project development time. Kleme,

Kimenov, and Nenov (1998) described several applications of Pinch

Technology within the food industry, work that was further

developed in Kleme et al. (1999). This research showed that Pinch

Technology can provide benefits far beyond oil refining and

petrochemicals.

The most important property of thermodynamically derived

heat recovery targets is that they cannot be improved upon by any

real system. Composite Curves play an important role in process

design; for HEN synthesis algorithms, they provide strict MER

targets. For process synthesis based on MPR, the Composite Curves

establish relevant lower bounds on utility requirements and capital

cost, thereby narrowing the search space for the following

superstructure construction and optimization.

The preceding observation highlights an important character-

istic of process optimization problems, and specifically those that

involve process synthesis and design. By strategically obtaining

key data about the system, it is possible to evaluate processes based

on limited informationbefore too much time (or other resources)

are spent on the study. This approach follows the logic of oil

drilling projects: potential sites are first evaluated in terms of key

preliminary indicators, and further studies or drilling commence

only if the preliminary evaluations indicate that the revenues

could justify further investment. The logic of this approach

was systematically formulated by Smith in his books on PI for

process synthesis (Smith, 1995; Smith, 2005) and by Daichendt and

Grossmann (1997), whose paper integrated hierarchical decom-

position and MPR to solve process synthesis problems.

20 Chapter Two

Sustainability in the design and operation of industrial processes is

becoming a key issue of process systems engineering. There are MPR

tools for designing chemical processes that minimize costs (and thus

maximize profits). These MPR formulations typically are highly

complex; therefore, in practical applications, the original problem is

usually decomposed into smaller problems that are solved

sequentially.

There are several definitions of sustainability. It can be defined as

the capacity to sustain the viability of a given system. In this context,

sustainable development implies that current actions should not harm

future generations and that measurable indicators are needed to

ensure compliance. For industrial processes, the relevant indicators

are the rates of resource intake and effluent emissions. These factors

are usually expressed in terms of footprints (De Benedetto and

Kleme, 2009)for example, the carbon footprint (CFP) and/or water

footprint (WFP) of a process. The sustainability of an industrial process

depends on minimizing these footprints. For synthesis and design

problems, the measured indicators should apply to the proposed

systems complete life cycle as part of a life-cycle assessment (LCA).

In short, PI has a direct impact on improving the sustainability

of a given industrial process. All PI techniques are geared toward

reducing the intake of resources and minimizing the release of

harmful effluents, goals that are directly related to the corresponding

footprints. Hence, employing PI and approaching the targeted

values will help minimize those footprints.

Footprint considerations, however, are only part of an industrys

ultimate indicatorthat is, its economic performance. This metric

takes the form of either cost or profit, depending on the selected system

boundaries. Thus, no matter how environmentally attractive a given

project may be, it will probably not be implemented unless doing so is

economically feasible. There are several ways to merge the objectives

of sustainability and profitability. The two most popular are

(1) expressing footprints in strictly economic terms and then folding

them into the cost or profit objective function; and (2) employing

multiobjective optimization, whereby footprint and economic

indicators are evaluated in parallel as objectives to be achieved.

In this section, four examples illustrate the potential of PI to reduce

not only resource demand but also operational and capital costs.

The problem areas vary widely, yet in each case the application of

the PI techniques yields significant benefits. Only the main problem

points and outcomes are described here; see Chapter 11 for more

details.

Process Integration 21

The FCC unit is a major process element in oil refineries, and

improvements in yield and efficiency have been attempted over the

years in response to various external drivers. Retrofitting the HEN

associated with an FCC usually leads to improved energy recovery

and thus to reduced energy use and/or increased throughput. In this

example, the HEN of the FCC process includes a main column and a

gas concentration section (Al Riyami, Kleme, and Perry, 2001). The

particular FCC plant considered had 23 hot streams and 11 cold

streams, and the associated cost and economic data required for the

analysis were specified by the refinery owners. The task was to

analyze the existing process and then propose a HEN retrofit plan for

improving energy recovery. Incremental area efficiency was used for

the targeting stage of the retrofit design, which was carried out using

the Network Pinch method (Asante and Zhu, 1997) consisting of a

diagnosis stage and an optimization stage. In the retrofit, four heat

exchangers were added and one existing exchanger was removed.

The resultant design produced energy savings of 8.955 MW or 74

percent. This translated into a 27 percent decrease in the plants utility

bill for an annual savings of $2,388,600. The modified HEN required

an investment of $3,758,420, so the payback period was less than 19

months.

Herrera, Islas, and Arriola (2003) studied a hospital complex that

included an institute, a general hospital, a regional laundry center, a

sports center, and some other public buildings. The diesel fuel used

to generate the steam required for the complex amounted to

75 percent of its total energy consumption and 68 percent of its total

energy cost ($396,131 in 1999). The Pinch Analysis that was performed

estimated the minimum need for external heating at 388.64 kW.

Because the actual heating used was consuming 625.28 kW, there

existed a potential energy savings of 38 percent.

As reported by Kleme, Kimenov, and Nenov (1998), an oil production

process operated with a minimal temperature difference of 65C at

the Process Pinch has been analyzed. The external heating required

by the system was provided by two types of hot utilities, and the

required external cooling was provided by two cold utilities. The

study resulted in increasing heat recovery and reducing the minimum

temperature difference to 814C. This reduced the hot and cold utility

requirements and eliminated the need for water steam and cooling

water, which considerably simplified the overall design.

A Whisky Distillery

In a study by Smith and Linnhoff (1988), the authors found that steam

was being used below the Process Pinch; this resulted in unnecessarily

22 Chapter Two

from the use of a heat pump, and so the steam below the Process

Pinch was eliminated by reducing the heat pumps size. Although

this approach required that the steam now be used for process

heating above the Process Pinch, overall energy costs were still

reduced owing to the reduced compressor duty.

2.11 Summary

This chapter offered an introduction to and an overview of the field

of PI, its HI roots, the expansion of its application areas, and the

evolution of process optimization methods toward combining

PI techniques and insights with the tools of MPR. The chapter

discussed the philosophy of PI as well as its contribution to

sustainable development and the economic efficiency of projects.

There is a wealth of useful information on these topics, and a

number of excellent sources of further information are described in

Chapters 13 and 14.

CHAPTER 3

Process Optimization

T

his chapter deals with process optimization: its definition,

goals, and application areas within sustainable industrial

process design and integration. The aim is to provide

information on how to formulate sustainability tasks as optimization

problems and on what tools to employ for solving them. The chapter

begins with a brief description of the general framework for model

building and optimization; this is followed by basics of optimization

problems and their classes as well as descriptions of the most common

algorithms for solving optimization problems. Finally, the chapter

discusses how to build models efficiently, how to handle complexity,

and how to ensure model adequacy and sufficient precision. The

details of computational implementations of optimization solvers

and other software tools are given in Chapter 9.

3.1 Introduction

Building and operating industrial processes entail costs and

environmental impacts. Emissions and effluents include: gaseous

waste streams, which may harbor CO2, SOx and NOx; wastewater and

various aqueous streams; and flue gases. When attempting to

improve the environmental and economic performance of process

systems, it is important to keep in mind that the processing paths,

which connect the various system inputs and outputs, usually

interact with each other. Therefore, minimizing resource demands

and environmental impacts is greatly facilitated by properly modeling

the process systems and then deciding which designs and operating

policies to pursuein what priority and to what extent.

Maintaining a balance between model accuracy and simplicity is

necessary in order to derive meaningful results with minimal

computational expense. A system model can be created for different

purposes. A lumped steady-state model (i.e., one that neglects

variations in time and space) will contain only algebraic equations.

To simplify the problem, steady-state models assume that the

operating units are black boxes or gray boxes (a black box is a model

that represents an empirical process in terms of its input, output, and

transfer parameters but does not describe any internal physics; a gray

23

24 Chapter Three

some of the physics is approximatedsee, e.g., Hangos and Cameron,

2001). This approach is appropriate for optimizing complex systems.

If time variability has to be accounted for then steady-state modeling

can be applied to a set of operating periods, each of which is charac-

terized by its own fixed parameters.

Framework and Workflow

A good process model should contain a thorough conceptual

description of the involved phenomena, unit operations, actions,

events, and so forth. Usually this description involves text, flowsheets,

and structural diagrams. Additionally, IT-domain diagramsfor

example, UML diagramscan be used (UML is a specification of

the Object Management Group; UML, 2010). The UML diagrams

include class, object, package, use case, sequence, collaboration,

statechart, component, and deployment diagrams.

A good process model should also contain a sufficiently precise

mathematical description. The mathematical relationships are used to

reflect not only physical laws but also technological constraints and

company rules. Mathematical models include algebraic equations of

some form (i.e., equalities or inequalities) and may be supplemented

with dynamic modeling, which uses differential equations to capture

variations in time, as well as states and actions to express operational

procedures and other dynamic relationships of algorithmic nature.

Structural information is also an essential feature of process network

models. When translated to Mathematical Programming (MPR)

models, such information is expressed by integer (mostly binary)

variables. An efficient alternative to representing superstructures

with binary variables is the P-graph and its related framework

(Friedler et al., 1992b), discussed in Chapter 7.

An efficient computational implementation of the mathematical

description may take the form of a stand-alone compiled application

(e.g., PNS Editor, 2010) or may be modeled within a popular

environment for process and mathematical calculations. Examples

include MATLAB (MathWorks, 2009), Scilab (2009), simulation and

optimization tools tailored for the process industry (AspenTech,

2009c), Modelica (2009a; OpenModelica, 2010), Honeywell UniSim

(Honeywell, 2010), and the open-source DWSIM (2010). All model

components have to be well synchronized to provide appropriate

user interfaces and sufficient visual aids to help understand the

process and the optimization results.

Models often include only the computational implementation

with some mathematical descriptions, but a much better practice is to

start with the concepts before deriving the mathematical relationships

Process Optimization 25

and then finally implementing the model computationally. In the

course of formulating the mathematical and implementation

components, it may be necessary to make some corrections to the

conceptual and/or mathematical components.

Formulation

3.3.1 What Is Optimization?

Optimization can be applied to different tasks: new system design,

the synthesis of a new processing network, and the retrofit design

and operational improvements in heat exchanger, reactor, and

separation networks. Optimization is employed to find the best

available option. An objective function consists of a performance

criterion to be maximized or minimized. The system properties that

determine this function are of two types:

respect to the choice to be made

2. Variablesa set of characteristics that are allowed to vary

manipulated by the optimization tools; these are referred to as

specifications or decision variables. The remaining variables are termed

dependent variables, and their values are determined by the

specifications and the systems internal relationships. The objective

function can be formulated in terms of a single variable or a

combination of dependent and decision variables. The value of the

objective function can be changed by manipulating the decision

variables.

Optimization tasks in industry include increasing heat recovery,

maximizing the efficiency of site utility systems, minimizing water

use and wastewater discharge, and other tasks. The formulations

that are used to solve such optimization problems are known as

mixed integer nonlinear programs (MINLPs). However, they are

frequently linearized to yield the more tractable mixed integer linear

programs (MILPs), and some can be further simplified and solved

via linear programming (LPR). In general, optimization problems

can be formulated as summarized in Table 3.1.

The continuous and discrete domains, together with the

constraints, define the feasible region for the optimization. This

region contains the set of options from which to choose. The value of

the function F depends on the values of the decision variables.

26 Chapter Three

performance criterion

where x Rn (continuous variables) Continuous domain

y Z n (integer variables) Discrete domain

subject to h(x, y) = 0 Equality constraints

g(x, y) 0 Inequality constraints

and chemical concentrations) that vary gradually within the feasible

region. Integer variables are used to model the status (ON versus

OFF) of operating devices as well as the selection/exclusion of options

for operating units in synthesis problems. Using only the problem

formulations, it is possible to generate many combinatorially infeasible

sets of the integer variable values, which are later analyzed by the

optimization solver. Especially for larger problems, its a good idea to

eliminate these infeasible combinations from the search space or to

build into the optimization solver a mechanism for avoiding them

(Friedler et al., 1996).

The type of the objective function F dictates which extremum

the minimum or the maximumto seek. Common performance

criteria are to minimize the process cost or to maximize the profit.

Because some process subsystems (e.g., water networks) do not usually

generate useful product streams, no revenue is directly realized and

so minimizing the total annualized cost is used instead as an

objective function. For complete production systems and supply

chains the objective is usually to maximize the profit. Thus, additional

variables (reflecting sales and customer behavior) and their

relationships may be added to the formulation.

Equality constraints stem not only from material and energy

balances but also from constitutive relations that normalize the

stream compositions to unity. The balances include those for total

flow rates, balances of the chemical components, and energy balances

of heat exchangers, boilers, and turbines. Inequality constraints stem

from limitations on concentrations, flow rates, temperatures,

pressures, throughput, and so forth. One example of a constitutive

relation is calculation of the fluid heat capacity flow rate from its

mass flow rate and specific heat capacity.

This section discusses methods that can be applied to detect

optimality and solve optimization tasks. Choosing a particular

Process Optimization 27

method should be based on a clear understanding, and toward this

end it is useful to bear in mind three aspects of optimization

problems:

maximizes) a convex objective function and if all the

constraints are convex (Williams, 1999).

2. An optimization problem is linear if its objective function

and all the constraints are linear; note that all linear

optimization problems are also convex (Williams, 1999).

3. A variable that can assume only integer (whole number)

values is an integer variable; integer variables that are

constrained only to values of 0 or 1 are binary variables.

specific features:

Presence of constraints: Unconstrained versus constrained.

Most practical tasks involve the formulation of constrained

optimization problems.

Problem convexity: Convex versus nonconvex.

Linear and nonlinear problems: This aspect depends on the

nature of the objective function and/or the constraints. Most

process optimization problems are bilinear (i.e., containing

products of two optimization variables), for example, they

feature componentwise mass balances involving products of

mass flow rates and concentration and enthalpy balances

involving products of mass flow rates and enthalpies.

Absence of integer variables: In such cases the entire problem is

continuous and so linear programming (LPR) or nonlinear

programming (NLP) models are employed.

Presence of integer variables: In such cases the problem is

referred to as integer. Integer problems are further

subdivided into pure integer programming (IP) models,

which involve only integer variables; mixed integer linear

programming (MILP) models, which involve linear

relationships with both integer and continuous variables;

and mixed integer nonlinear programming (MINLP)

models, which involve nonlinear relationships with both

integer and continuous variables.

be found in Floudas (1995) and Williams (1999). The most important

28 Chapter Three

factors are linearity and the existence of integer variables. There are

two main reasons for the significance of these factors:

that incorporate gradient or other search mechanisms. The

presence of integer variables introduces combinatorial

complexity caused by the need to build search trees that

branch through the integer variables.

2. Nonlinearity introduces a different type of complexity.

Whereas linear problems (LPR and MILP) have been shown

to be convex (Floudas, 1995; Williams, 1999), the convexity of

nonlinear problems (NLP and MINLP) must be evaluated on

a case-by-case basis. Such problems are generally assumed to

be nonconvex until proven otherwise.

One of the most popular methods for handling optimization

formulations is the use of MPR solvers. They usually take

standardized input in the form of matrices of variables, parameters,

and equations, after which they explore the search space, using local

search and gradients, to reach an optimal solution.

When employing local search (e.g., gradient-based) algorithms, the

desired extremum (minimum or maximum) needs to be located and

proven. The optimality condition usually employed by solver

algorithms is based on the mathematical definition of an extremum.

For finding a minimum, the definition requires the existence of a

point in the search space such that any small deviation from that

point, in any direction within the search space, will result in an

increase of the objective function value or in keeping it the same:

F x * , y * d F x , y , x , y vicinity of x * , y * (3.1)

(see, e.g., Edgar and Himmelblau, 1988; Floudas, 1995; Luenberger

and Ye, 2008). Rigorous definitions can also be found by searching

the Web for KKT optimality conditions (aka KarushKuhnTucker

conditions, an extension of the method of Lagrange multipliers).

Additional conditions are applied to attain global optimality when

using local search algorithms, which in this case require the problem

Process Optimization 29

to be convex. For linear problems (e.g., MILP), the locally optimal

solution is guaranteed to be globalin other words, the best possible.

Another advantage of the linear model is that linear solvers, unlike

with nonlinear ones, do not require initialization with a feasible

solution. Nonlinear problems are much more difficult to solve, and

attaining feasibility of the solutions is a significant issue. Any result

obtained guarantees global optimality only if the optimization

problem is shown to be convex (Williams, 1999).

Linear Optimization Problems

The best-known algorithm for LPR problems is the simplex algorithm

(Dantzig, 1951; Dantzig, Orden, and Wolfe, 1954). It solves LPR

maximization problems by constructing a feasible solution at a vertex

of the search polyhedron and then walking along the edges of the

polyhedron to other vertices with successively higher objective

function values. Although this algorithm is efficient in general, in

some cases it may require exponential time to find the optimum.

Khachiyan (1979) proposed a method that, in the worst case, finds a

solution in polynomial time for the worst case. Today, the LPR

problems are usually solved using one of two methods:

developed almost continuously over the years since its initial

formulation, starting with the revised Simplex Method of

Dantzig and Orchard-Hays (1953). A thorough consideration

of the simplex algorithm and its computational techniques is

given in Maros (2003b). Modern LPR solvers use algorithms

that flexibly solve both continuous and integer linear

problems.

2. Interior point methods: In contrast to the simplex algorithm,

which finds the optimal solution by traversing points on the

boundary of the feasible region, interior point methods move

through the interior of the feasible region. One such method

(see Mehrotra, 1992) is used by MATLAB.

Nonlinear Optimization Problems

Nonlinear programming is used to solve continuous nonlinear

problems. A common feature of all deterministic search methods is

that, starting from a current solution point, they perform an iterative

search for a better feasible solution within the vicinity of the current

30 Chapter Three

possible. This strategy means that the methods find only local

optima, so there is no guarantee of global optimality. Although

modern optimization software packages are capable of solving

constrained problems, they are mostly based on the search techniques

developed for unconstrained optimization.

Problems

Many methods have been developed for performing local search for

the minimum of an objective function (Luenberger and Ye, 2008).

Deterministic search algorithms mainly assume that the objective

function is minimized and use the concept of descent: a series of

iterations are performed that aim reducing the objective function

value at each iteration. A thorough review of the deterministic search

algorithms for optimization and their performance, as applied to

chemical process models has been provided elsewhere, for example,

Kleme and Vaek (1973). Such methods include pattern search, the

Rosenbrock method, and conjugate gradients.

All descent-based algorithms rely on a common general strategy.

First, an initial solution point is chosen. Then, a rule-based direction

of movement and a step size are determined and executed. At the

new point, a new pair of direction and step size are determined, and

the process is repeated until the desired convergence is achieved.

The main difference between the various search algorithms is in the

rule applied for determining the iteration steps.

When the search domain is a given line, the process of determining

the function minimum is known as a line search. Higher-dimensional

problems are solved by executing a sequence of successive line

searches. The two simplest line search algorithms are the Fibonacci

method (e.g., Avriel and Wilde, 1966) and the golden section method

(e.g., Kiefer, 1953). Both algorithms search over a fixed interval and

assume that the function is unimodal within it. The golden section

algorithm has a linear convergence rate of dk+1/dk = 0.618. Another line

search algorithm is the Newton method, which is based on the

additional assumption of function smoothness; this technique

achieves even faster convergence than the golden section one. There

are also other line search algorithms that are grouped together as

curve-fitting methods.

Another important search algorithmone that is applicable to

multidimensional function domainsis the method of steepest

descent or gradient method. Its update rule derives the new step

based on the gradient information in the current point, identifying

the direction of the fastest decrease of the objective function. This is

one of the oldest and best-known function minimization methods.

Hence it is often used as a benchmark against which the performance

Process Optimization 31

of other methods is measured. See Luenberger and Ye (2008) for

additional information.

Many algorithms for constrained problems proceed in two main

steps: (1) The constrained problems are transformed into equivalent

unconstrained ones, and then (2) the modified problems are subjected

to unconstrained search algorithms. Groups of such algorithms

include primal methods, penalty and barrier methods, and primal-

dual methods.

The primal algorithms search for the optimal solution directly

through the feasible region. For a problem with n variables and

m equality constraints, primal methods work within the (n m)-

dimensional feasible space. Each point in the process is feasible, and

the value of the objective function constantly decreases. Thus, a

feasible solution is guaranteed even if the search is interrupted before

it finishes. Most primal methods do not rely on special problem

structure (such as convexity), so they are applicable to general NLP

problems. The weaknesses of primal algorithms are their requirement

that an initial feasible solution be found or specified and their

slowness (or even failure) to converge in the case of nonlinear

constraints.

Penalty and barrier methods utilize unconstrained approximations

of the constrained optimization problem. Penalty methods add a

penalty term to the objective function, which results in a high cost if

constraints are violated. Barrier methods instead add a term that

favors points interior to the feasible region over those near the

boundary. There are two major issues with applying these methods.

First, it is important to ensure the accuracy with which the

unconstrained NLP problem being solved approximates the true,

constrained one. Second, it is possible that the penalty or barrier term

may dominate the objective function, skewing the problem to such a

degree that the feasible solutions found are not actually optimal.

All integer programming problems (pure IP, MILP, and MINLP)

possess combinatorial features. The first solution methods used

cutting planes (i.e., added constraints to force integrality), and various

algorithms have been devised on this basis. The effective approach

has been to divide the problem into a number of smaller problems, a

method known as branch and bound. This is a strategy of problem

decomposition: it aims to partition the feasible region into more

manageable subdivisions and then, if required, into further subdivided

partitions. The method was proposed by Land and Doig (1960) for

solving LPR problems and has since evolved to a point where it can

32 Chapter Three

be used to solve IP and MIP problems. The original large MIP problem

is divided into a number of subproblems, called nodes, which form an

enumeration tree. The algorithm starts from a main node and

progresses toward the so-called leaves, adding nodes to the current

solution or discarding them as necessary. In this process, an important

role is played by the bounding function, which (in the case of objective

minimization) provides a lower bound on the remaining part of the

problem under the current branch. Thus, if the lower bound on the

current subproblem node is higher than the current best solution,

then the algorithm can safely discard (prune) the node and all its

subnodes. More information on IP solving algorithms can be found

in Nemhauser and Wolsey (1999).

Optimization Problems

An alternative to deterministic algorithms are those that employ

stochastic sampling of the search space and hill climbing: pushing the

algorithm to lower objective function values (assuming minimization)

by allowing temporary increases in the objective function value. The

main advantage of stochastic search methods is that their search is

not confined in the neighborhood of a given current solution, which

means there is a higher probability of finding the global optimum.

Such techniques have proven to be successful in applicationsfor

example, process design and synthesis studieswhere the

computational requirement for single samplings is modest and the

time frame for producing a solution is more relaxed. Their major

limitation is that many iterations are required in order to assure some

degree of optimality. This is because the algorithms blindly evaluate

even infeasible combinations of variable values, especially for the

simulated annealing variants.

Simulated annealing (SA) is the most prominent stochastic search

method (Kirkpatrick, Gelatt, and Vecchi, 1983). An interesting

application is the GAPinch toolbox for MATLAB (Prakotpol and

Srinophakun, 2004), which implements a genetic algorithm (GA)

search to solve MINLP network synthesis problems involving water

reuse. Another promising development is the use of the GA

framework to optimize schedules and supply chains (Shopova and

Vaklieva-Bancheva, 2006). Ant colony optimization (Zecchin et al.,

2006) and Tabu search (Cunha and Ribeiro, 2004) have been applied to

process design and operation. Other examples of stochastic search

methods include the following:

(GP) to identify steady-state models of integrated chemical

processes.

Process Optimization 33

2. Manolas et al. (1997) applied GA techniques to the optimal

operation of industrial utility systems.

3. Castell et al. (1998) described a novel GA application to

optimize process layout.

4. Genetic algorithms have been proposed for the synthesis of

Heat Exchanger Networks (HENs) and other processing

systems (Ravagnani et al., 2005; Xiangkun et al., 2007; Fieg,

Xing, and Jeowski, 2009); the paper by Xiangkun and

colleagues presents a hybrid GA-SA method.

5. Ahmad et al., (2008) proposed that SA be used for synthesizing

Heat Exchanger Networks. This approach employs a

completely evolutionary strategy, starting from a trivial

network topology that connects all hot streams to coolers

and all cold streams to heaters.

The procedure for building a process model is illustrated in Figure 3.1.

The modeling begins with accumulating sufficient information

about the processin order to develop an understanding of the

elements and the relationships between themand proceeds to

formulating a mathematical description of the process that is imple-

mented on a computational platform. A distinctive characteristic of

the procedure is its iterative nature. The mathematical modeling

BEGIN

YES

Conceptual YES

modeling

Mathematical Need

modeling corrections?

NO

Computational Need

implementation corrections?

NO

END

34 Chapter Three

iterative feedback loop, as shown in the figure. A similar correction

loop is also present at the output of the implementation block. The

discussion that follows addresses only those activities in Figure 3.1

that involve conceptual and mathematical modeling.

Conceptual modeling involves collecting and organizing essential

information about the phenomena in the process under consideration.

This step is often referred to as data extraction (Williams, 1999). The

process operating units are described along with the relevant features

of their behavior. The important constraints for the units capacities

and other limitations are identified and added to the description, as

is topological information about the process network. The main

purpose of this description is to serve as an interface between the

process operators and the modelers. Therefore, it is important that

the description be concise enough that modelers can efficiently grasp

the workings of the complete system. That being said, the description

must contain sufficient detail to complete the study.

When a production process is modeled for the purpose of Process

Integration, the data extraction involves four main steps:

interconnections; creation of a flowsheet that reflects this

description.

2. Identification of the heating and cooling needs of the process

through use of the flowsheet and related data about the

operating units.

3. Definition of the Heat Integration process streams: identifying

for each stream the values for its heat load, as well as the

supply and target temperatures. Some process streams may

need to be segmented. This is done if the specific heat capacity

of a given process stream varies significantly within the

interval between its supply and its target temperature.

4. Analysis of the collected data.

then the various water-using operations are analyzed and their

relevant properties are recorded in some standard form. The most

popular way to express the water requirements (Wang and Smith,

1994; Kuo and Smith, 1997) of an operation is in terms of the limitations

on (1) inlet and outlet concentrations of different contaminants and

(2) flow rates of the water to be consumed. This topic is discussed in

more detail in Chapter 12.

Process Optimization 35

Obtaining network-related information is needed to account for

constraints that are related to the system topology and to the limits

imposed by operating units on the suitability of various process

streams to serve as inputs or outputs. In water networks, such

considerations include the acceptability (or unacceptability) of using

the water output from some operations as inputs for other operations.

For instance, the final washing of sugar crystals in sugar production

would require pure water, and for this the outputs from other water-

using operations would be unacceptable. On the otherhand, used

water from blanching might be perfectly acceptable for the initial

washing or rinsing of fruits. This type of information is used to

formulate additional constraints on the compatibility of different

process streams. When supplied to automated process optimization

algorithms, these constraints serve to eliminate a number of infeasible

combinations of process units. When building pure MPR models

(Williams, 1999), network-related information is transformed into

explicit mathematical constraints involving expressions with binary

selection variables. When using the graph-theoretic approach and/or

the P-graph framework (Friedler et al., 1993) to construct a process

model, such information is explicitly encoded in the P-graph building

blocks (materials and operations) and is then used by algorithms that

generate only those topologies that are combinatorially feasible.

Constructing the Equations

After the conceptual basis has been established, it is time to begin

constructing the explicit mathematical formulations of the problem.

The standard procedure in this regard is first to build a super-

structureone that incorporates all possible options and combin-

ations of operating unitsand then to reduce the superstructure via

optimization techniques. In this context, a superstructure is the union

of several feasible flowsheets (see Figure 6.2 for an example of a water

reuse network superstructure). When this union includes all possible

flowsheets, the superstructure is called the maximal structure

(Friedler et al., 1993) or the hyperstructure (Papalexandri and

Pistikopoulos, 1996).

There are two basic approaches to formulating the superstructure

and subjecting it to reduction optimization.

followed by translation of that structure into an integer programming

model: The generated problem is then solved by the

corresponding MPR algorithm. Popular codes for solving

MILP problems are OSL (GAMS, 2009) and CPLEX (ILOG,

2009); both are included in such commercial optimization

software packages as GAMS (2009). If the model does not

36 Chapter Three

decisions on flow rates and capacities only, then continuous

optimizers (LPR or NLP) can be used.

2. Automated generation of the maximal superstructure, followed by

enumeration of all feasible network topologies using the P-graph

framework. The unit operations can be described in terms of

(a) the input to the automated procedure, (b) the compatible

connections between them, and (c) the corresponding process

and cost information. More details are provided in

Chapter 7.

formulation of the mass balances. The following key points should

be kept in mind:

wise. Overall balances are performed over the total contents

of the input and output streams of a process unit. For steady-

state models of continuous processes, this content is usually

expressed as a mass-based flow rate (e.g., in units of kg/s,

kg/h, or t/h). Componentwise mass balances reflect the mass

conservation principle between the inlet and outlet streams

for individual chemical components (or pseudocomponents).

For a given operating unit, the set of all componentwise mass

balances is exactly sufficient for completely characterizing

the material flows into and out of the unit (adding the overall

balance could lead to an overspecified system of equations).

However, if using the overall material balance is critical to

the system model, then the overall balance can be used in

place of one of the componentwise balances.

In some cases, tracking all chemical species in various process

streams is not necessary. This is true for water networks in

which mass balances are written for the water flow rates and

for the analyzed contaminants. However, an incomplete list

of the material species contained in the water streams will

naturally result in incomplete mass balances (Smith, 2005).

When complete results are necessary, rigorous simulations of

the optimized system must be performed.

The componentwise mass balances of stream mixers and the

more complex operating units involve bilinear terms that

reflect products of the stream flow rates and the component

concentrations. In this case, the result is an NLP or MINLP

problem. If either the concentrations or the flow rates are fixed

then the model could be linear (an LPR or MILP problem),

which would make for an easier computation that might

guarantee global optimality. Although such an approach is

Process Optimization 37

often viable, additional analyses are required to ensure that

the linearization does not result in an inadequate (inaccurate)

model. Similar problems also exist for Heat Exchanger

Networks, where nonisothermal mixing leads to equations

containing bilinear terms: products of mass flow rates and

enthalpies.

problem. The mass balances are supplemented in the model by lower

and upper bounds on the stream flow rates and component

concentrations. Another source of constraints is the temperature

feasibility of interequipment connections. For example, water coming

from a blanching operation may be too hot to be used for washing

fresh fruits and so may require cooling (perhaps by mixing with a

colder stream) before this use; alternatively, this may be rejected as an

unacceptable connection, leading to a forbidden match. In MPR, such

constraints usually contain integer variables.

A frequent problem when synthesizing process networks is

obtaining extremely low flow rates for some interconnections. When

the model to optimize accounts for the complete capital costs, this

problem is less likely to appear. However, if capital costs are

underestimated (or disregarded) for some reason, then operating costs

may dominate, resulting in degenerate solutions with impractically

small flow rates. Practical solutions require reasonably accurate

estimates of the capital cost, especially if there are fixed costs. A more

straightforward option is to stipulate a lower bound on all network

flow rates.

Another important part of creating a process model is identifying

the energy needs of various operating units. It is crucial to account

for the heating and cooling needs of the process operations, and these

are established by formulating enthalpy balances (Linnhoff et al.,

1982). Most process streams need to be transported between the

operating units and also moved through them, and such transport

involves overcoming certain pressure drops (for fluids) and

performing mechanical work (for solids). These operations require

mechanical shaft power, which can be supplied by direct-drive

machines or electrical motorselements that define the power

requirements of a process. In many cases, additional equations are

needed (e.g., constitutive relations as well as calculations of reaction

rates and equilibriums).

The objective function to choose depends on the goal of the

optimization. It is possible to choose from a number of criteria,

including (1) maximizing profit; (2) minimizing operating cost;

(3) minimizing total annualized cost (TAC); (4) minimizing

consumption of certain resources or consumption per unit of product;

38 Chapter Three

footprint per unit of product.

It is frequently necessary to optimize more than one criteria.

There are three main approaches to this task (Ehrgott, 2005):

then add the other criteria as constraints to the problem.

2. Combine all the criteria into one objective function by

summing them up, where each criterion is weighted with a

given coefficient.

3. Perform a multicriteria optimization, accounting explicitly

for the conflicts between the chosen objectives (criteria).

Process synthesis and process design taskswhen performed on

real-life, industrial-scale problemstend to involve substantial

number of operating units. Examples can be found in many areas:

number of possible combinations of potential heat exchangers.

Thermodynamic and process-related constraints usually

reduce this number, but even then the complexity remains

significant.

Water subsystem design is no exception, and problems with

20 or more water-using operations are common (Bagajewicz,

2000; Thevendiraraj et al., 2003). This number leads to high

levels of combinatorial complexity. In a superstructure, each

water-using operation (and each intermediate water main)

defines at least one mixer. If the number of water-using

operations is denoted by Nop, then there can be no fewer than

Nop corresponding binary variables in the network super-

structure, and the number of combinations of binary variable

values to be examined by the corresponding MIP solver

would equal 2Nop. Thus, for 20 operations there would be more

than a million (106 = 1,000,000) possible combinations.

binary variables is dictated by the number of the candidate operating

units. In the worst case, the solution algorithm will have to examine

the entire search space, which depends exponentially on the number

of the binary variables. One modeling strategy that reduces the

search space by several orders of magnitude is to use the Maximal

Structure Generation (MSG) and Solution Structures Generation

(SSG) algorithms of the P-graph framework (Friedler et al., 1993).

These algorithms effectively discard all infeasible combinations of

the binary selection variables and retain only the feasible ones.

Process Optimization 39

Another popular technique in process design and software

development is modularization or encapsulation. The complexity

management efforts in information technology and process modeling

led to development of the concepts of object-oriented modeling (see,

e.g., Modelica, 2009a; UML, 2010) and object-oriented programming

(e.g., C++, C#, Java, Delphi). With these modeling concepts, a number

of related objects or operations can be grouped together and

represented as a single object or operation. Similarly, flowsheeting

and simulation software may offer the option of representing a

distillation column as a single operating unit at the level of the entire

flowsheet while still allowing simulation of the column at the local

level; such a facility is available, for example, in gPROMS (PSE, 2009)

and HYSYS (HYSYS, 2010).

The key to object-oriented thinking is the principle of information

hiding, according to which every object should conceal as many details

as possible about its own functionality and provide to other objects

only the information relevant to interactions with them. Hence, the

interface describes the bundle of information items that are defined

as being available to an external object. An example of applying

information hiding in water network design is the abstraction of the

detailed information about water-using operations into only three

relevant pieces of information for each operation and contaminant:

the limiting inlet and outlet concentrations and the limiting flow rate

(or contaminant load). This bundle of information thus constitutes the

operations interface to the water minimization problem.

Efficiently managing the model and problem complexity is greatly

facilitated by the practice of documenting the complete modeling

process and optimization results, including their interpretation. It is

best to document all stages, starting from the conceptual modeling

and ending with the computational implementation and obtained

results. All this documentation should be systematic so that the

reasoning and results can be clearly understood and traced back to

their roots. This style of documentation is extremely useful and makes

the work of teams and individual engineers smoother and more

efficient in the long run.

Another important tool for managing complexity is targeting,

which reveals limitations in the underlying design or operation task

to which the optimization is being applied. With proper targeting it

is possible to obtain an upper bound on system performance and/or

a lower bound on system cost. In fact, it is also possible to calculate

practically achievable targets:

the first type of estimate: the maximum possible amount of

water reduction.

For HEN synthesis, the Maximum Energy Recovery (MER)

targets can be established, and HENs achieving them also

40 Chapter Three

minimum for the total annualized cost (i.e., the sum of

annual operating costs and annualized investment costs).

on its performance is this: if the best possible system performance is

still insufficient to satisfy the specified requirements, then no further

time and effort should be spent on designing that system. The next

section gives more details on the use of targeting to managing design

problem complexity.

Mathematical tools are absolutely necessary for optimizing the

design and operation of industrial processes. However, meaningful

and applicable results are obtained only when process insight is used

to guide the model building, the optimizing, and the interpretation

of results. Each optimization problem has its own particular features.

For example:

requirements and practices from those for the processing of

poultry.

When designing HENs, the various underlying process

operations that need cooling or heating should be properly

examined for data extraction. In some cases, process knowledge

may aid in the lumping of different heating/cooling needs,

thereby simplifying the flowsheet. Process knowledge is also

employed when partitioning properly into segments a process

stream whose heat capacity flow rate varies widely.

model improvement should be thoroughly documented and

implemented in the modelfor instance, in the form of constraints

or simplifying assumptions.

A good practice is to perform targeting for the desired application:

a heat recovery problem, water management, a separator network, or

reactor network design. Targeting provides information on potential

performance. Targeting procedures have been developed and well

tested for a number of applications and domains (see Chapter 2 for a

review).

The benefits of using variants of Pinch Analysis or other targeting

are twofold. First, the designer can estimate the best possible

performance of the system by using simple models and calculations,

even before using rigorous design procedures, saving valuable time.

The obtained targets can be used in preliminary sensitivity studies

to determine which operations and units should be included in the

design and which should be left out. This approach can greatly

Process Optimization 41

simplify network design if the number of candidate operations is

large and the targets can be used as guides in the network synthesis.

Usually the engineer aims either to achieve the targets exactly or to

approach them closely with the final design. If the targeting model is

too idealized, then the estimates produced will serve as loose

performance or cost bounds, not as tight bounds. Yet in many cases

this strategy results in a simple design procedure and a nearly

optimal outcome.

Second, if the targeting model is exact, as in Pinch Analysis for

Heat Integration (Linnhoff and Flower, 1978), or if it at least captures

all key factors at the corresponding design stage, as with Regional

Energy Clustering (Lam, Varbanov, and Kleme, 2010), then the

targeting procedure also provides a convenient partitioning of the

original design space. This makes it easier to decompose the problem

and to simplify the remaining actions. A good example of partitioning

the design space is the division above/below the Pinch in Heat

Integration (see Chapter 4).

As discussed in Section 3.5, convex problems are guaranteed to

produce globally optimal results when solved with deterministic

algorithms that employ local search. In contrast, a nonconvex

optimization problem is difficult to solve, and its solution is not

guaranteed to be a global optimum. All linear MP models are convex

(Williams, 1999). With nonlinear models, however, the problem

convexity must be established on a case-by-case basis. Nonlinear

models hinder the computation process of the solvers (e.g., by

requiring that feasible initial solutions be provided), and they often

result in poor numerical convergence. This is why engineers usually

seek ways to obtain linear models in some form. Crucial factors in

this task are preserving the models precision and validity.

Sometimes it is possible to linearize relationships that are inherently

nonlinear. This can be done, for instance, by replacing a single

nonlinear relationship with two or more linear ones that, together,

approximate the original function over the required range. This

technique is known as piecewise linearization. For example, it can be

applied when the available cost functions (for piping, distillation

columns, or heat exchangers) are too complex. The result of this

approach is a small reduction in the overall model precision and an

increase in the number of integer variables in the model; thus,

combinatorial complexity is increased but computational complexity

(due to nonlinearity) is reduced. The principal advantage is the

resulting linearity of the model, which almost always makes it easier

to solve than the original one. Caution must be exercised in the

process of linearization: the loss of precision must be kept as small as

42 Chapter Three

representation of the underlying process. Another pitfall is a

potentially unacceptable increase in combinatorial complexity, which

can result if too many linear segments are used to approximate the

original relationship.

Another approach to avoiding nonlinearity is to define a number of

fixed levels for some variables and then to substitute the original

nonlinear variables in the model with linear combinations of integer

variables and parameters (where the parameters are derived from

the original variables). In this way, the bilinear terms in the mass

balances of contaminants can be reduced to purely linear expressions.

In its consequences and pitfalls, this technique is similar to piecewise

linearization.

Other Techniques

There are other approaches and techniques for coping with nonlinear

models. Two of them are of particular interest to practical process

optimization: successive MILP (SMILP), which is used for model

decomposition and solving; and model reformulation.

Successive MILP can be applied to many process engineering

optimization modelsas long as the nonlinearities are not too strong.

One example is the optimization of utility systems, which consist

mainly of a set of steam headers combined with steam turbines, gas

turbines, boilers, and letdowns. Most of the nonlinearities in such

systems are bound to the enthalpy balances of the steam headers, the

steam turbines, and the letdowns. The computational difficulties

imposed by the nonlinearities can be overcome by first fixing the

values of some system properties during optimization (e.g., enthalpies

of steam mains), thereby producing a linear optimization model, and

following this with a rigorous simulation after each optimization

step. The linear optimization steps are repeated, followed again by

simulation, and so on until convergence is achieved; see Figure 3.2.

This procedure converges rapidly when applied to the optimization

of existing utility systems: usually five iterations at most are required

to reach reasonably small error levels.

Model reformulation refers to the symbolic transformation of the

original nonlinear equations into another set of equations that are

equivalent but linear. The resulting set usually contains more

equations than the initial one. One such reformulation technique,

known as the Glover transformation (Floudas 1995), can transform

equations containing the product of a continuous and a binary

variable. The essence of the technique is to replace each term that is a

product of a continuous variable and a binary variable with additional

continuous variables and an additional set of linear inequality

constraints.

Process Optimization 43

BEGIN

Initialization

(+Simulation)

MILP Simulation

Convergence? NO

YES

END

Once the model is built, the next step is validation. This process boils

down to evaluating how precisely the model predicts real-life

phenomena as well as how adequately it represents the modeled

system (Steppan, Werner, and Yeater, 1998; Montgomery, 2005). If the

model turns out to be imprecise or inadequate, then the reasons for

these shortcomings must be discovered and addressed. This iterative

process is similar to debugging during software development.

It is generally accepted that residuals (and their plots) are

sufficient for assessing whether a given model accurately predicts

the underlying process. The residual plots can be used to minimize or

even eliminate stochastic errors. In addition, parity plots are helpful

in exposing any systematic errors in the model.

The final check is to analyze the models variance (Steppan,

Werner, Yeater, 1998; Montgomery, 2005). In essence, this means

determining whether the empirically derived coefficients and the

models predictions have any statistical significance. This is

performed by means of a standard procedure for the Analysis of

Variance (ANOVA).

This page intentionally left blank

CHAPTER 4

Process Integration

for Improving

Energy Efficiency

H

eat recovery is widely applied in industrial processes and

has an extensive historical record. However, systematic

methods for performing heat recovery are relatively new

when compared with the age of modern industry.

In industry, large amounts of thermal energy are used to perform

heating. Examples of this can be found in crude oil preheating before

distillation, preheating of feed flows to chemical reactors, and heat

addition to carry out endothermic chemical reactions. Similarly,

some processessuch as condensation, exothermal chemical

reactions, and product finalizationrequire that heat be extracted,

which results in process cooling. There are several options for utility

heating; these include steam, hot mineral oils, and direct fired

heating. Steam is the most prevalent option because of its high

specific heating value in the form of latent heat. Utility cooling

options include water (used for moderate-temperature cooling when

water is available), air (used when water is scarce or not economical

to use), and refrigeration (when subambient cooling is needed). Heat

recovery can be used to provide either heating or cooling to processes.

Heat recovery may take various forms: transferring heat between

process streams, generating steam from higher-temperature process

waste heat, and preheating (e.g., air for a furnace, air or feed water for

a boiler using waste heat).

Heat transfer takes place in heat exchangers, which can employ

either direct mixing or indirect heat transfer via a wall. Direct heat

exchange is also referred to as nonisothermal mixing because the

temperatures of the mixed streams are different. Mixing heat

exchangers are efficient at transferring heat and usually have low

45

46 Chapter Four

capital cost. In most industries, the bulk of the heat exchange must

occur without mixing the heat-exchanging streams. In order to

exchange only heat while keeping the streams separate, surface heat

exchangers are employed. In these devices, heat is exchanged through

a dividing wall. Because of its high thermal efficiency, the counter-

current stream arrangement is the most common with surface

heat exchangers. To simplify the discussion, counter-current heat

exchangers are assumed unless stated otherwise. In terms of

construction types, the traditional shell-and-tube heat exchanger is

still the most common. However, plate-type and other compact heat

exchangers are gaining increased attention. Their compactness,

together with significant improvements in their resistance to leaking,

have made them preferable in many cases.

A hot process stream can supply heat to a cold one when paired in

one or several physical heat exchangers arranged in parallel or

sequence. Each such pairing is referred to as a heat exchange match.

The form of the steady-state balance equations for heat exchange

matches that is most convenient for Heat Integration calculations is

based on modeling a match as consisting of hot and cold sides, as

shown in Figure 4.1. The hot and cold part each have a simple, steady-

state enthalpy balance that involves just one material stream and one

heat transfer flow.

The main components of the model are (1) calculations of the

heat transfer flows accounted for by the enthalpy balances and

(2) estimation of the necessary heat transfer area. For the latter, both

the log-mean temperature difference and the overall heat transfer

coefficient are employed. The enthalpy balance of the hot and cold

parts, and the kinetic equation of the heat transfer, may be written

as follows:

Hot part

QHE

Cold part

Process Integration for Improving Energy Efficiency 47

where QHE [kW] is the heat flow across the whole heat exchanger, U

[kWm2C1] is the overall heat transfer coefficient, A [m2] is the heat

transfer area, and TLM [C] is the logarithmic-mean temperature

difference. More information can be found in Shah and Sekuli

(2003), Tovazshnyansky et al. (2004), and Shilling et al. (2008).

The heat exchange matches are often viewed as being identical to

heat exchangers, but this is not always the case. A given heat exchange

match may be implemented by devices of different construction or

by a combination of devicesfor example, two heat exchangers in

sequence may implement a single heat exchange match. The

distinction between the concept of a heat exchange match and its

implementation via heat exchangers is important because of capital

cost considerations.

4.2.1 Process Integration and Heat Integration

A historical review of the field was given in Chapter 2. Initially,

attention was focused on reusing any waste heat generated on

different sites. Each surface heat exchanger was described with only

a few steady-state equations, and the thermal energy saved by reusing

waste heat led to reductions in the expense of utility resources. This

approach became popular under the names Heat Integration (HI) and

the more general term Process Integration (PI). In this context, HI

means integrating different processes to achieve energy savings.

Engineers realized that integration could also reduce the consumption

of other resources as well as the emission of pollutants. Heat and

Process Integration came to be defined more widely in response to

similar developments in water reuse and wastewater minimization.

Process design has an inherent hierarchy that can be exploited for

making design decisions. This hierarchy may be represented by the

so-called onion diagram (Linnhoff et al., 1982), as shown in Figure 4.2.

The design of an industrial process starts with the reactors or other

key operating units (the onions core). This is supplemented and

served by other parts of the process, such as the separation subsystem

(the next layer) and the Heat Exchanger Network (HEN) subsystem.

48 Chapter Four

Ambient

Utilities

ng

xcha er Ne

tE t

Separator

arato

wo

ep

He

S r

rk

Feed

Reactor

Reactor

Feed + Product

Product

Steam Turbine

Boiler

demands, are handled by the utility system.

The thermodynamic bounds on heat exchange can be used to estimate

the utility usage and heat exchange area for a given heat recovery

problem. The resulting estimates of the process performance are a

lower bound on the utility demands and a lower bound on the

required heat transfer area. These bounds are known as targets for

the reason that heat recovery estimates are achievable in practice and

usually minimize the total cost of the HEN being designed.

For efficient heat recovery in industry, the relevant data must be

identified and presented systematically. In the field of Heat

Integration, this process is referred to as data extraction. The heat

recovery problem data are extracted in several steps:

heat recovery exchangers.

2. Remove the recovery heat exchangers and replace them with

equivalent virtual heaters and coolers.

3. Lump all consecutive heaters and coolers.

4. The resulting virtual heaters and coolers represent the net

heating and cooling demands of the flowsheet streams.

5. The heating and cooling demands of the flowsheet streams

are then listed in a tabular format, where each heating

Process Integration for Improving Energy Efficiency 49

demand is referred to as a cold stream and, conversely, each

cooling demand as a hot stream.

process flowsheet involving two reactors and a distillation column.

The process already incorporates two recovery heat exchangers. The

utility heating demand of the process is H = 1760 kW, and the utility

cooling demand is C = 920 kW.

The necessary thermal data has to be extracted from the initial

flowsheet. Figure 4.4 shows the flowsheet after steps 1 through 4. The

heating and cooling demands of the streams have been consolidated

by removing the existing exchangers, and the reboiler and condenser

duties have been left out of the analysis for simplicity (although these

duties would be retained in an actual study). It is assumed that any

process cooling duty is available to match up with any heating duty.

Applying step 5 to the data in Figure 4.4 produces the data set in

Table 4.1. By convention, heating duties are positive and cooling ones

are negative. (The subscripts S and T denote supply and target

temperatures for the process streams.)

The last column of Table 4.1 gives the heat capacity flow rate (CP).

For streams that do not change phase (i.e., from liquid to gas or vice

versa), CP is defined as the product of the specific heat capacity and

the mass flow rate of the corresponding stream:

Condenser

52C

2080 kW

Column

78C 182C

Reactor 2

1760 kW

H 120C

138C

Reboiler Reactor 1

100C 120C

3240 kW

920 kW 30C

C 34C

FIGURE 4.3 Data extraction: Example process flowsheet (after CPI, 2004 and

2005).

50 Chapter Four

Condenser

52C

Column

2080 kW

Reactor 2

Heating

3840 kW 120C

138C

Reboiler Reactor 1

100C 120C

Cooling

4160 kW Heating

3240 kW

34C

30C

FIGURE 4.4 Data extraction: Heating and cooling demands (after CPI, 2004

and 2005).

H1 Hot 182 78 2080 20

H2 Hot 138 34 4160 40

C3 Cold 52 100 3840 80

C4 Cold 30 120 3240 36

H H

CP (4.5)

TT TS T

CP to calculate the stream duties. Some more problems related to

data extraction are discussed in Chapter 12.

The main strategy of Pinch-based Process Integration is to identify

the performance targets before starting the core process design

activity. Following this strategy yields important clues and design

guidelines. The most common hot utility is steam. Heating with

steam is usually approximated as a constant-temperature heating

Process Integration for Improving Energy Efficiency 51

utility. Cooling with water is nonisothermal because the cooling

effect results from sensible heat absorption into the water stream and

thus leads to increasing the temperature.

Heat Recovery between One Hot and One Cold Stream

The Second Law of thermodynamics states that heat flows from

higher-temperature to lower-temperature regions. As shown in Eq. (4.3),

in a heat exchanger the required heat transfer area is proportional to

the temperature difference between the streams. In HEN design, the

minimum allowed temperature difference (Tmin) is the lower bound

on any temperature difference to be encountered in any heat

exchanger in the network. The value of Tmin is a design parameter

determined by exploring the trade-offs between more heat recovery

and the larger heat transfer area requirement. Any given pair of hot

and cold process streams may exchange as much heat as allowed by

their temperatures and the minimum temperature difference.

Consider the two-stream example shown in Figure 4.5(a). The

amount of heat recovery is 10 MW, which is achieved by allowing

Tmin = 20C. If Tmin = 10C, as in Figure 4.5(b), then it is possible to

squeeze out one more megawatt of heat recovery. To obtain the

heat recovery targets for a practical HEN design problem, this

principle needs to be extended to handle multiple streams.

Streams: The Composite Curves

The analysis starts by combining all hot streams and all cold

streams into two Composite Curves or CCs (Linnhoff et al., 1982).

For each process there are two curves: one for the hot streams (Hot

Composite Curve, HCC) and another for the cold streams (Cold

Composite Curve, CCC). Each Composite Curve (CC) consists of a

200 Steam 200 Steam

150 150

100 100

50 20C 50 10C

CW CW

0 10 20 0 10 20

2 10 4 H [MW] 1 11 3 H [MW]

52 Chapter Four

availability in the process (the HCC) and the overall heat demands

of the process (the CCC). The procedure of HCC construction is

illustrated in Figure 4.6 on the data from Table 4.1. All temperature

intervals are formed by the starting and target temperatures of the

hot process streams. Within each temperature interval, a composite

segment is formed consisting of (1) a temperature difference equal

to that of the interval and (2) a total cooling requirement equal to

the sum of the cooling requirements of all streams within the

interval. This is achieved by summing up the heat capacity flow

rates of the streams crossing the interval. Next, the composite

segments from all temperature intervals are combined to form the

HCC. Construction of the CCC is entirely analogous.

The Composite Curves are combined in the same graph in order

to identify the maximum overlap, which represents the maximum

amount of heat that could be recovered. The HCC and CCC for the

example from Table 4.1 are shown together in Figure 4.7.

Both CCs can be moved horizontally (i.e., along the H axis), but

usually the HCC position is fixed and the CCC is shifted. This is

equivalent to varying the amount of heat recovery and (simultaneously)

the amount of required utility heating and cooling. Where the curves

overlap, heat can be recuperated between the hot and cold streams.

More overlap means more heat recovery and smaller utility

requirements, and vice versa. As the overlap increases, the temperature

differences between the overlapping curve segments decrease. Finally,

at a certain overlap, the curves reach the minimum allowed temperature

difference, Tmin. Beyond this point, no further overlap is possible. The

closest approach between the curves is termed the Pinch point (or

simply the Pinch); it is also known as the heat recovery Pinch.

It is important to note that the amount of largest overlap (and

thus the maximum heat recovery) would be different if the minimum

(a) (b)

T [C] T [C]

182

/C

kW

138

20

138 P2

=

/C

=

+C C

W CP 1 kW/

1

CP

0k 60

=4 78

78 CP

2

34 34

CP1 = 20 kW/C

2080 4160 H [kW] 1760 3600 880 H [kW]

6240 6240

The hot streams plotted separately The composite hot stream

Process Integration for Improving Energy Efficiency 53

allowed temperature difference is changed for the same set of hot

and cold streams. The larger is the value of Tmin, the smaller is the

possible maximum heat recovery. Specifying the minimum utility

heating, the minimum utility cooling, or the minimum temperature

difference fixes the relative position of the two Composite Curves

and hence the maximum possible amount of heat recovery. The

identified heat recovery targets are not absolutethey are relative

to the specified value of Tmin. If that value is increased, then the

minimum utility requirements also increase and the potential for

maximum recovery drops; see Figure 4.8.

T [C]

200

150

100 Pinch

50

Tmin = 10C

0

H [kW]

Qc,min = 328 Qrec = 5912 Qh,min = 1168

T [C]

200

150

Pinch

50

Tmin = 20C

0

H [kW]

Qc,min = 328 Qrec = 5912 Qh,min = 1168

Qc,min = 728 Qrec = 5512 Qh,min = 1568

54 Chapter Four

offs. Increasing Tmin results in larger minimum utility demands and

increased energy costs; choosing a higher value reflects the need to

reduce heat transfer area and its corresponding investment cost.

Conversely, if Tmin is reduced then utility costs go down but

investment costs go up. This trade-off is illustrated in Figure 4.9.

The heat recovery Pinch has important implications for the HEN

being designed. As illustrated in Figure 4.10, the Pinch sets the

absolute limits for heat recovery within the process.

COST

Total

Energy

Capital

1 2

Optimum TMIN

Tmin.

T QH,min

PINCH

Qc,min

FIGURE 4.10 Limits for process heat recovery set by the Pinch.

Process Integration for Improving Energy Efficiency 55

The Pinch point divides the heat recovery problem into a net heat

sink above the Pinch point and a net heat source below it (Figure 4.11).

At the Pinch point, the temperature difference between the hot and

cold streams is exactly equal to Tmin, which means that at this point

the streams are not allowed to exchange heat. As a result, the heat

sink above the Pinch is in balance with the minimum hot utility

(QH,min) and the heat source below the Pinch is in balance with the

minimum cold utility (QC,min), while no heat is transferred across the

Pinch via utilities or via process-to-process heat transfer.

No heat can be transferred from below to above the Pinch,

because this is thermodynamically infeasible. However, it is feasible

to transfer heat from hot streams above the Pinch to cold streams

below the Pinch. All cold streamseven those below the Pinch

could be heated by a hot utility; likewise, the hot streams (even above

the Pinch) could be cooled by a cold utility. Although these

arrangements are thermodynamically feasible, applying them would

cause utility use to exceed the minimum, as identified by the Pinch

Analysis. This is a fundamental relationship in the design of heat

recovery systems.

What happens if heat is transferred across the Pinch? Recall that

it is possible to transfer heat only from above to below the Pinch. If,

say, XP units of heat are transferred across the Pinch (Figure 4.12),

then QH,min and QC,min will each increase by the same amount in order

to maintain the heat balances of the two problem parts. Any extra

heat that is added to the system by the hot utility must then be taken

away by the cold utility, in addition to the minimum requirement

QC,min.

Cross-Pinch process-to-process heat transfer is not the only way

by which a problems thermodynamic Pinch partitioning can be

T QH,min

Zero cross-

pinch transfer

Pinch

Tmin

QC,min

56 Chapter Four

T

QH,min QUC,above XP

QUH,below XP

PINCH

violated. This could also happen if the external utilities are placed

incorrectly. For example, any utility heating below the Pinch will

create a need for additional utility cooling in that part of the system

(Figure 4.12). Conversely, any utility cooling above the Pinch will

create a need for additional utility heating. The implications of the

Pinch for heat recovery problems can be distilled into the following

three conditions, which must hold if the minimum energy targets for

a process are to be achieved.

2. There must be no external cooling above the Pinch.

3. There must be no external heating below the Pinch.

demands. The rules are applied explicitly in the context of HEN

synthesis by the Pinch Design Method (Linnhoff and Hindmarsh,

1983) and also before a HEN retrofit analysis to identify causes of

excessive utility demands by a process. Other HEN synthesis

methodsif they achieve the minimum utility demandsalso

conform to the Pinch rules (though sometimes only implicitly).

The Composite Curves are a useful tool for visualizing heat recovery

targets. However, they can be time-consuming to draw for problems

that involve many process streams. In addition, targeting that relies

solely on such graphical techniques cannot be very precise. The

process of identifying numerical targets is therefore usually based

on an algorithm known as the Problem Table Algorithm (PTA). Some

Process Integration for Improving Energy Efficiency 57

MPR-oriented authors employ the equivalent transshipment model

(Cerd et al., 1990). The steps are as follows:

2. Set up temperature intervals.

3. Calculate interval heat balances.

4. Assuming zero hot utility, cascade the balances as heat

flows.

5. Ensure positive heat flows by increasing the hot utility as

needed.

The algorithm will be illustrated using the sample data in Table 4.2.

Step 1

Because the PTA uses temperature intervals, it is necessary to set up

a unified temperature scale for the calculations. If the real stream

temperatures are used, then some of the heat content would be left

out of the recovery. The problem is avoided by obtaining shifted

stream temperatures (T*) for PTA calculations. The hot streams are

shifted to be colder by Tmin/2 and the cold streams are shifted to be

hotter by Tmin/2. If the shifted temperatures (T*) of a cold and a hot

stream (or their parts) are the same, then their real temperatures are

still actually Tmin apart, which allows for feasible heat transfer. This

operation is equivalent to shifting the Composite Curves toward

each other vertically, as illustrated in Figure 4.13. The last two

columns in Table 4.2 show the shifted process stream temperatures.

Step 2

Temperature intervals are formed by listing all shifted process

stream temperatures in descending order (any duplicate values are

T*

T*

T

Temperature

T

Tmin

Hot

2

Tmin

Cold

+

2

58 Chapter Four

1 Cold 20 180 20 25 185

2 Hot 250 40 15 245 35

3 Cold 140 230 30 145 235

4 Hot 200 80 25 195 75

TABLE 4.2 Problem Table Algorithm Example: Process Streams Data (Tmin = 10C)

Temp. Population [oC] [kW/oC] [kW] Deficit

[oC]

245 2

10 15 150 Surplus

235

4 40 15 600 Deficit

195 10 10 100 Surplus

CP = 30

185

CP = 15

CP = 25

40 10 400 Deficit

3

CP = 20

75

40 5 200 Deficit

35

10 20 200 Deficit

25

1

TABLE 4.3 Problem Table Algorithm for the Streams in Table 4.2

(TBs), which form the temperature intervals for the problem. For the

example in Table 4.2, the TBs are 245C, 235C, 195C, 185C, 145C,

75C, 35C, and 25C.

Step 3

The heat balance is calculated for each temperature interval. First,

the stream population of the process segments falling within each

temperature interval (the first two columns of Table 4.3) is identified.

Process Integration for Improving Energy Efficiency 59

Next, the sums of the segment CPs (heat capacity flow rates) in each

interval are calculated; then that sum is multiplied by the interval

temperature difference (i.e., the difference between the TBs that

define each interval). This calculation is also illustrated in Table 4.3.

Step 4

The Problem Heat Cascade shown in Figure 4.14 has a box allocated

to each temperature interval; each box contains the corresponding

interval enthalpy balances. The boxes are connected with heat flow

arrows in order of descending temperature. The top heat flow

represents the total hot utility provided to the cascade, and the

bottom heat flow represents the total cold utility. The hot utility flow

is initially assumed to be zero and this value is combined (summed)

with the enthalpy balance of the top cascade interval to produce the

value for the next lower cascade heat flow. This operation is repeated

for the lower temperature intervals and connecting heat flows until

the bottom heat flow is calculated, resulting in the cascade shown in

Figure 4.14(a).

Step 5

The resulting heat flow values in the cascade are examined, and a

feasible heat cascade is obtained; see Figure 4.14(b). From the

cascading heat flows, the smallest value is identified; if it is

nonnegative (i.e., positive or zero), then the heat cascade is

thermodynamically feasible. If a negative value is obtained then a

positive utility flow of the same absolute value has to be provided at

245 C 0 kW 245 C 750 kW

H = 150 kW H = 150 kW

235 C 150 kW 235 C 900 kW

H = 600 kW H = 600 kW

195 C 450 kW 195 C 300 kW

H = 100 kW H = 100 kW

185 C 350 kW 185 C 400 kW

H = 400 kW H = 400 kW

THOT PINCH = 150C

145 C 750 kW 145 C *

TPINCH 0 kW

TCOLD PINCH = 140C

H = 1400 kW H = 1400 kW

75 C 650 kW 75 C 1400 kW

H = 200 kW H = 200 kW

35 C 450 kW 35 C 1200 kW

H = 200 kW H = 200 kW

25 C 250 kW 25 C 1000 kW

HOT UTILITY HOT UTILITY

(a) Initial cascade (b) Feasible cascade

FIGURE 4.14 Heat Cascade for the process data in Table 4.2.

60 Chapter Four

the topmost heat flow, after which the cascading described in Step 4

is repeated. The resulting heat cascade is guaranteed to be feasible

and provides numerical heat recovery targets for the problem. The

topmost heat flow represents the minimum hot utility, the bottommost

heat flow represents the minimum cold utility, and the TB with zero

heat flow represents the location of the (heat recovery) Pinch. It is

often possible to obtain more than one zero-flow temperature

boundary, each representing a separate Pinch point.

Threshold problems feature only one utility typeeither hot or

cold. They are important mostly because they often result in no utility

capital trade-off below a certain value of Tmin, since the minimum

utility demand (hot or cold) becomes invariant; see Figure 4.15.

(a) T (b) T

Steam

Tmin = 14C

Tmin = 20C

CW CW

H H

Heat recovery, hot and More heat recovery, no hot utility

cold utilities

(c) T (d) T

Tmin = 10C

Generation

CW CW

CW

H H

No increase in heat recovery Utility substitution

high-temperature fuel cells, which usually have large net cooling

demands but no net heating demands (Varbanov et al., 2006;

Varbanov and Kleme, 2008). An essential feature that distinguishes

threshold problems is that, as Tmin is varied, demands for only one

Process Integration for Improving Energy Efficiency 61

utility type (hot or cold) are identified over the variation range; in

contrast, pinched problems require both hot and cold utilities over

this range. When synthesizing HENs for threshold problems, one

can distinguish between two subtypes (see Figure 4.16):

exactly as Pinch-type problems.

2. High-threshold Tmin: For these problems, it is first necessary

to satisfy the required temperature for the no-utility end

before proceeding with the remaining design (using the tick-

off heuristic; see Figure 4.53).

Utility Placement: Grand Composite Curve

In many cases, more than one hot and one cold utility are available

for providing the external heating and cooling requirements after

energy recovery and it is necessary to find and evaluate the

(a) T ST

Utilities

[MW]

CW

ST

Tmin = 10C

10C Tmin CW

H [MW]

Low Threshold Tmin

(b) T

ST

Utilities

[MW]

ST

CW Tmin = 10C

High Threshold Tmin

62 Chapter Four

(Figure 4.17).

To assist with this choice and to enhance the information

derived from the HCC and CCC, another graphical construction

has been developed, known as the Grand Composite Curve (GCC)

(Townsend and Linnhoff, 1983). The heat cascade and the PTA

(Linnhoff and Flower, 1978) offer guidelines for the optimum

placement of hot and cold utilities, and this allows one to determine

the heat loads associated with each utility. For the previous sections,

the assumption has been that only one cold and one hot utility are

availablealbeit with sufficiently low and high temperatures to

satisfy the cooling and/or heating demands of the process. However,

most industrial sites feature multiple heating and cooling utilities

at several different temperature levels (e.g., steam levels, refrigeration

levels, hot oil circuit, furnace flue gas). Each utility has a different

unit cost. Usually the higher-temperature hot utilities and the

lower-temperature cold utilities cost more than the ones with

temperatures closer to the ambient. This fact underscores the need

to choose a mix that results in the lowest utility cost. The general

objective is to maximize the use of cheaper utilities and to minimize

the use of more expensive utilities. For example, it is usually

preferable to use low-pressure (LP) instead of high-pressure (HP)

steam and to use cooling water (CW) instead of refrigeration. The

Composite Curves plot in Figure 4.7 provides a convenient view for

evaluating the process driving force and the general heat recovery

targets. However, the CCs are not useful for identifying targets

when multiple utility levels are available; the GCC is used for

this task.

Fuel

Fuel HP

Power Power

Steam MP Gas

Steam

Levels turbine

turbine

LP

BFW

preheat Heating Heating Heating Heating

Heating

(Q+W)

Processes, building complexes

Heating Heat

pump

Cooling Cooling Power

Cooling (Q) (W)

Furnace Refrigeration

Air

Fuel preheat Power

Heating

FIGURE 4.17 Choices of hot and cold utilities (amended after CPI 2004 and

2005).

Process Integration for Improving Energy Efficiency 63

The GCC is constructed using the Problem Heat Cascade (Figure 4.14).

The heat flows are plotted in the T-H space, where the heat flow at

each temperature boundary corresponds to the X coordinate and the

temperature to the Y coordinate (Figure 4.18).

The GCC can be directly related to the Shifted Composite Curves

(SCCs), which are the result of shifting the CCs toward each other by

Tmin/2 so that the curves touch each other at the Pinch; see Figure 4.19.

At each temperature boundary, the heat flow in the Problem Heat

Cascade and GCC corresponds to the horizontal distance between

the SCCs.

The GCC has several fundamental properties that facilitate an

understanding of the underlying heat recovery problem. The parts

with positive slope (i.e., running uphill from left to right) indicate

that cold streams dominate (Figures 4.18 and 4.19). Similarly, the

parts with negative slope indicate excess hot streams. The shaded

areas in the GCC plot, which signify opportunities for process-to-

process heat recovery, are referred to as heat recovery pockets.

The GCC shows the hot and cold utility requirements of the process in

terms of both enthalpy and temperature. This allows one to distinguish

between utilities at different temperature levels. There are typically

750 kW

245 C

H= 150 kW

900 kW

235 C

H= 600 kW

300 kW

195 C

H= 100 kW

400 kW

185 C

H= 400 kW PINCH

0 kW

145 C

H= 1400 kW

1400 C

75 C

H= 200 kW

1200 kW

H= 200 kW 35 C

1000 kW

25 C

Cold Utility

500 1000 1500

Q (kW)

FIGURE 4.18 Constructing the GCC for the streams in Table 4.2.

64 Chapter Four

T* [C] T* [C]

300 300

750 750

900 900

300

200 200 300

400 400

100 100

1400 1400

1200 1200

1000 1000

0 500 1000 1500 0 2000 4000 6000 8000

Q [kW] H [kW]

FIGURE 4.19 Relation between the GCC (left) and the SCC (right) for the

streams in Table 4.2.

HP steam HP steam

200 200 MP

steam

PINCH PINCH

100 100

Q [kW] Q [kW]

0 500 1000 1500 0 500 1000 1500

Single steam level Multiple steam levels

FIGURE 4.20 Using the GCC to target for single and multiple steam levels.

example, there may be a supply of both high-pressure (HP) and

medium-pressure (MP) steam. As indicated previously, it is desirable to

maximize the use of cheaper utilities and to minimize the use of more

expensive ones. Utilities at higher temperature and/or pressure are

usually more expensive; see Figure 4.20. Therefore, MP steam is used

first: ranging from the Y axis until it touches the GCC (resulting in Tmin

in this point), which maximizes its usage. Only then is HP steam used.

When a utility line or profile touches the GCC, a new Pinch point

is created, termed a Utility Pinch (the MP steam line touching the

GCC in Figure 4.20). Each additional steam level creates another

Utility Pinch and increases the complexity of the utility system.

Process Integration for Improving Energy Efficiency 65

Higher complexity has several negative consequences, including

increased capital costs, greater potential for leaks, reduced safety,

and more maintenance expenses. Therefore limits are typically

placed on the number of steam levels.

Higher-temperature heating demands are satisfied by

nonisothermal utilities. These include hot oil and hot flue gas, both

of which maintain their physical phase (liquid and gaseous) across

a wide range of temperatures. The operating costs associated with

such utilities are largely dependent on furnace efficiency and on the

intensity and efficiency of the pumping or fan blowing. When

targeting the placement of a nonisothermal hot utility, its profile is

represented by a straight line,1 which runs from the upper right to

the lower left in the graph of Figure 4.21. The lines starting point

corresponds to the utility supply temperature and also to the

rightmost point for the utilitys heating duty. The utility use endpoint

corresponds either to the zero of the H axisin which case all

utility heating is covered by the current nonisothermal utilityor

to the rightmost point on the H axis for other, cheaper hot

utilities.

FIGURE 4.21 T*

Properties of

nonisothermal hot *

Tsupply

utilities.

CP1

CP2

T*ambient

H

Hloss,2

Hloss,1

capacity of the corresponding stream.

66 Chapter Four

point corresponds to the ambient temperature. The distance from

this point to the zero of the H axis represents the thermal losses

from using the utility. The heat capacity flow rate of the nonisothermal

utility target is determined by making the utility line as steep as

possible, thereby minimizing its CP and the corresponding losses

(Figure 4.21). Its supply temperature is usually fixed at the maximum

allowed by the furnace and the heat carrier composition; the

remaining degree of freedom corresponds to the utilitys exact CP.

Smaller CP values result in steeper slopes and smaller losses.

The placement for a nonisothermal utility (e.g., hot oil) may be

constrained by two problem features: the process Pinch point and a

kink in the GCC at the top end of a heat recovery pocket; see

Figure 4.22 for an example.

When fuel is burned in a furnace or a boiler, the resulting flue

gas becomes available to heat up the corresponding cold-stream

medium (for steam generation or direct process duty). Transferring

heat to the process causes the flue gas temperature to drop as it moves

from the furnace to the stack. The stack temperature has to be above

a specified value: the minimum allowed stack temperature, which is

determined by limitations due to corrosion. If flue gas is used directly

for heating, then the Pinch point, if it is higher, may become more

limiting than the minimum allowed stack temperature. If the

analyzed process features both high-temperature and moderate-

temperature utility heating demands, then flue gas heating may not

be appropriate for satisfying all those demands. If steam is cheaper,

T

T* 290

290 Minimizing

Tsupply = 300

the CP

Minimizing

the CP

need to be limiting

n

mi

CP

in

m

CP

=130

Pinch Pinch

120 120

H H

Process Pinch limitation Heat recovery pocket limitation

Process Integration for Improving Energy Efficiency 67

then combining it with flue gas reduces the latters CP and the

corresponding stack heat losses.

Another option for utility placement is to use part of the cooling

demand of a process for generating steam. This is illustrated in

Figure 4.23, in which steam generation is placed below the Pinch.

The GCC can reveal where utility substitution may improve

energy efficiency; see Figure 4.24. The main idea is to exploit heat

recovery pockets that span two or more utility temperature levels.

The technical feasibility of this approach is determined by both the

temperature span and the heat duty within the pocket, which should

be large enough to make utility substitution worthwhile when

weighed against the required capital costs.

Utility cooling below ambient temperatures may be required, a

need that is usually met by refrigeration. Refrigerants absorb heat by

evaporation, and pure refrigerants evaporate at a constant

temperature. Therefore, refrigerants are representedon the plot of

T (or T*) versus Hby horizontal bars, similarly to the steam levels.

On the GCC, refrigeration levels are placed similarly to steam levels;

see Figure 4.25.

When the level of a placed utility is between the temperatures of

a heat recovery pocket, the Utility Pinch cannot be located by using

T* Pinch

160

140

Point of closest

approach. Not

120 necessarily at the

boiling point

100

Preheat

Superheat Evaporation

80

60

40 Cooling Water

QC,min

20

0 H

68 Chapter Four

T* T*

HP Steam HP Steam

Generation Generation

LP Steam use

CW CW

H H

T*

TCW

5C

40C

70C

the GCC. In this case, the Balanced Composite Curves (BCCs) are

used. Figure 4.26 shows how the data about the placed utilities can be

transferred from the GCC to the BCCs, enabling the correct location

of the Utility Pinch associated with LP steam.

The BCCs create a combined view in which all heat sources and

sinks (including utilities) are in balance and all Pinches are shown.

Process Integration for Improving Energy Efficiency 69

(a) T* (b) T

HP steam HP steam

LP steam LP steam

LP steam Pinch

Pinch

CW

CW

H H

Grand Composite Curve Balanced Composite Curve

obtaining targets for specific utilities, and planning HEN design

regions.

4.4.1 Heat Transfer Area, Capital Cost, and

Total Cost Targeting

In addition to maximizing heat recovery, it is also possible to estimate

the required capital cost. The expressions for obtaining these

estimates are derived from the relationship between heat transfer

area and the efficiency of a heat exchanger. Methods for targeting

capital cost and total cost were initially developed by Townsend and

Linnhoff (1984) and further elaborated by others (e.g., Ahmad,

Linnhoff, and Smith 1990; Colberg and Morari, 1990; Linnhoff and

Ahmad, 1990; Zhu et al., 1995).

The HEN capital cost depends on the heat transfer area, the

number of the heat exchangers, the number of shell-and-tube passes

in each heat exchanger, construction materials, equipment type, and

operating pressures. The heat transfer area is the most significant

factor, and assuming one-pass shell and tube exchangers it is

possible to estimate the overall minimum required heat transfer

area; this value helps establish the lower bound on the networks

capital cost. Estimating the minimum heat transfer area is based on

the concept of an enthalpy interval. As shown in Figure 4.27, an

enthalpy interval is a slice constrained by two vertical lines with fixed

values on the H axis. This interval is characterized by its H

70 Chapter Four

Enthalpy interval

1 qstream

Amin =

TLM streams hstream

boundaries, its process stream population, and the film heat transfer

coefficients of those streams.

The minimum heat transfer area target can be obtained by

estimating it within each enthalpy interval of the BCCs and then

summing up the values over all intervals (Linnhoff and Ahmad,

1990):

EI

1 NS

qs , i

AHEN, min

TLM, i s 1 hs , i

i 1

number of streams; i denotes ith enthalpy interval; s, the sth stream;

TLM,i, the log-mean temperature difference in interval i (from the

CC segments); qs, the enthalpy change of the sth stream; and hs, the

heat transfer coefficient of sth stream. The area targets can be

supplemented by targets for number of shells (Ahmad and Smith,

1989) and for the number of heat exchanger units, thus providing a

basis for estimating the HEN capital cost and the total cost. This

approach is known as supertargeting (Ahmad, Linnhoff, and Smith,

1989). With supertargeting it is also possible to optimize the value of

Tmin prior to designing the HEN. Proposed improvements to the

capital cost targeting procedure of Townsend and Linnhoff (1984)

mainly involve: (1) obtaining more accurate surface area targets for

HENs with nonuniform heat transfer coefficients (Colberg and

Morari, 1990; Jegede and Polley, 1992; Zhu et al., 1995; Serna-

Gonzlez, Jimnez-Gutirrez, and Ponce-Ortega, 2007); (2) accounting

for construction materials, pressure rating, and different heat

exchanger types (Hall, Ahmad, and Smith, 1990); or (3) accounting

for safety factors, including prohibitive distance (Santos and

Zemp, 2000). Further information can be found in the paper by Taal

and colleagues (2003), which summarizes the common methods

used to estimate the cost of heat exchange equipment and also

provides sources of projections for energy prices.

Process Integration for Improving Energy Efficiency 71

Heat Engines

Particularly important processes are the heat enginessteam and gas

turbines. They operate by drawing heat from a higher-temperature

source, converting part of it to mechanical power; then (after some

energy loss) they reject the remaining heat at a lower temperature

(see Figure 4.28). For targeting purposes the energy losses are usually

neglected.

Integrating a heat engine across the Pinch, which is equivalent

to a Cross-Pinch process-to-process heat transfer, results in a

simultaneous increase of hot and cold utility, which usually leads to

excessive capital investment for the utility exchangers. If a heat

engine is integrated across the Pinch, then the hot utility requirement

is increased by Q and the cold by Q W (in the notation of Figure 4.28).

Heat engines should be integrated in one of two ways.

1. Above the Pinch (Figure 4.29a): This increases the hot utility

for the main process by W, but this extra heat is converted

into shaftwork.

2. Below the Pinch (Figure 4.29b): This offers a double benefit. It

saves on a cold utility, and the process heat below the Pinch

supplies Q to the heat engine (instead of rejecting it to a

cooling utility).

steam turbines may be placed either below or above the Pinch because

they draw and exhaust steam. Figure 4.30 shows steam turbine

integration above the Pinch, which has the benefit of cogenerating

T*

Source

T*source

Q

QW

T*sink

Sink

72 Chapter Four

A+W

T* T* A

A(Q W) Q

100% conversion

PINCH HeatWork

Q W HE W Q

PINCH HE W

100% conversion QW

B HeatWork BQ

(a) (b)

HP Steam

T* T* QHP

QFUEL

QHP W

QLP

QLP LP Steam Boiler

Condensate

PINCH

CW

QC,min H

typically used only as a utility heat source for the processes and can

be placed only above the Pinch.

Heat Pumps

Heat pumps present another opportunity for improving the energy

performance of an industrial process. Their operation is the reverse

of heat engines. That is, heat pumps take heat from a lower-

temperature source, upgrade it by applying mechanical power, and

then deliver the combined flow to a higher-temperature heat sink

(Figure 4.31).

An important characteristic of heat pumps is their coefficient of

performance (COP). This metric for device efficiency is defined as the

Process Integration for Improving Energy Efficiency 73

configuration. Sink

Q + W = 125 kW

T*sink

W = 25 kW

Q = 100 kW

T*source

Source

ratio between the heat delivered to the heat sink and the consumed

shaftwork (mechanical power):

Q sink Q source W

COP = (4.7)

W W

between the heat sink and the heat source (Laue, 2006); this difference

is also referred to as temperature lift. Figure 4.32(a) shows the

appropriate integration of a heat pump across the Pinch, with the

heat source located below the Pinch and the heat sink above it.

The GCC facilitates sizing of the heat pump by evaluating the possible

temperatures of the heat source and heat sink, and their loads; see

Figure 4.32(b). Integrating entirely above the Pinch results in direct

conversion of mechanical power to heat. This is a waste of resources

because most of the power is generated at the expense of two to three

times the amount of fuel energy. The second alternativeplacing the

heat pump entirely below the Pinchresults in the power flow

consumed by the heat pump being added to the cooling utility

demand below the Pinch.

The procedure for sizing heat pumps to be placed across a

(process or utility) Pinch is illustrated in Figure 4.33. First, tem-

peratures are chosen for the heat source and the heat sink. Then the

horizontal projections spanning from the temperature axis to the

GCC provide the maximum values for the heat source and sink loads.

Recall that the GCC shows shifted temperatures. Because real

temperatures are used when calculating the heat-pump temperature

lift, the GCC values must be modified by subtracting or adding

Tmin/2 (see Section 4.3.3). The COP value can be derived from the

calculated Tpump and can be then used to calculate the necessary

duties.

74 Chapter Four

QHP + W

PINCH QHP +W

QHP HEAT W

PUMP W

QHP

CW

Qcmin QHP

H

Appropriate placement Load and temperature lift on the GCC

T*

QHP + W COP

5

W

6

MAX

2

QHP + W T*SINK

3

W

Tmin TPUMP

T*SOURCE 4 4

QHP

1

MAX

3

H TPUMP

an industrial process with Tmin = 20C and that a heat pump is

available, described as follows:

0.874

COP 100.18 Tpump (4.8)

integration option) allows choosing a shifted temperature for the

heat source, in this example: T*source = 85C; see Figure 4.35. Using this

value allows extracting a maximum of 6.9 MW from the GCC below

the Pinch. Setting T *sink = 100C results in an upper bound of 2.634

MW for the sink load. Transforming to real temperatures and taking

the difference yields a temperature lift of Tpump = 35C. By Eq. (4.8),

the COP is thus equal to 4.4799. Given the upper bounds on the sink

Process Integration for Improving Energy Efficiency 75

TMIN = 20C

H[MW] T*[C]

450

29.40 410

400

23.82 131 350

18.00 118 300

T*[C] 250

1.80 115

200

0.00 94 150

4.30 91 100

50

11.50 79

0

15.00 30 0 10 20 30 40

H [MW]

Heat-pump sizing

example: Attempt 1.

120

QSINK

100

Maximum

T* [C]

80

QSOURCE

60

40

20

0 5 10 15 20 25

H [MW]

and source heat loads, the smaller one is chosen as a basis. Here the

sink bound is smaller, so the sink is sized to its upper

bound: Qsink = Qsink,max = 2.634 MW. From this, the required pump

power consumption is computed to be 0.588 MW. As a result, the

actual source load for the heat pump is 2.046 MW. Comparing this

value with the upper bound of 6.9 MW, it is evident that the source

heat availability is considerably underutilized.

A different selection of source and sink temperatures is needed if

the source availability is to be better utilized. As a second attempt,

the sink temperature is increased from 100C to 110C. The maximum

source heat remains 6.9 MW, but the maximum sink capacity

increases from 2.634 to 7.024 MW. This results in the desired

76 Chapter Four

MW, and the sink load Qsink = 9.557 MW; see Figure 4.36.

However, the heat sink is oversized in this case (Figure 4.36), so

the heat source temperature has to be shifted upward. An increase of

2.28C (see Figure 4.37) yields the following values: T *source = 87.28C;

Tpump = 42.72C; a maximum source load (also taken as the selected

source load) of 5.152 MW; a maximum sink load equal to the selected

sink load of Qsink = 7.016 MW; COP = 3.7637; and W = 1.864 MW. Better

results can be obtained by optimizing the two temperatures

simultaneously while using the overall utility cost as a criterion.

140

QSINK

120

100

The sink is oversized

T* [C]

80

QSOURCE

60

40

20

0 5 10 15 20 25

H [MW]

140

QSINK

120

100

T* [C]

80

QSOURCE

60

40

20

0 5 10 15 20 25

H [MW]

Process Integration for Improving Energy Efficiency 77

Once utility levels are chosen, the heat pump can be placed also

across a Utility Pinch (Figure 4.38). A special case of integration

involves the placement of refrigeration levels (Figure 4.39).

Refrigeration facilities are heat pumps whose main value is that their

cold end absorbs heat. Utilizing their hot ends for process heating

can save considerable amounts of hot utility, especially when

relatively low-temperature heating is needed.

The simplest distillation column includes one reboiler and one

condenser, and these components account for most of the columns

T*

HP

MP

IP

LP

CW

T*

Rcond CW

HEAT 0

W

PUMP

R1

R2

78 Chapter Four

represented generally by a rectangle: the top side denotes the reboiler

as a cold stream, and the bottom side denotes the condenser as a hot

stream; see Figure 4.40.

There are three options for integrating distillation columns:

across the Pinch, as shown in Figure 4.41(a); and entirely below or

entirely above the Pinch, as shown in Figure 4.41(b). Integrating

across the Pinch only increases overall energy needs, so this option

is not useful. The other two options result in net benefits by

eliminating the need to use an external utility for supplying the

distillation reboiler (above the Pinch) or the condenser (below the

Pinch). The GCC is used to identify the appropriate column integration

options. Figure 4.42 illustrates two examples of appropriately placed

distillation columns.

T*

A Reboiler

A, B

Condenser

B

H

(a) (b) A

T* A + QDIST T* QDIST

QDIST QDIST

PINCH PINCH Q

DIST

QDIST

QDIST

B + QDIST B

Across the pinch Below or above the pinch

Process Integration for Improving Energy Efficiency 79

of distillation columns against the

GCC.

across the Pinch, there are several degrees of freedom that can be

utilized to facilitate appropriate placement. It may be possible to

change the operating pressure, which could shift the column with

respect to the temperature scale until it fits above or below the Pinch.

Varying the reflux ratio results in simultaneous changes of the column

temperature span and the duties of the reboiler and condenser.

Increasing the reflux ratio (the ratio between the reflux flowrate

returned to the column and the distillate product flowrate) yields a

smaller temperature span and larger duties, whereas reducing the

ratio has the opposite effect. It is also possible to split the column into

two parts, introducing a double effect distillation arrangement. In

this approach, one of the effects is placed below and the other above

the Pinch; this prevents internal thermal integration of the column

effects.

Additional options are available, such as interreboiling and

intercondensing. When using available degrees of freedom, it is

important to keep in mind that the energycapital trade-offs of the

column and the main process are combined and thus become more

complicated than the individual trade-offs. Another important issue

is controllability of the integrated designs: unnecessary complications

should be avoided, and disturbance propagation paths should be

discontinued. It is usually enough to integrate the reboiler or the

condenser. If inappropriate column placement cannot be avoided,

then condenser vapor recompression with a heat pump can be used

to heat up the reboiler.

Evaporators constitute another class of thermal separators.

Because they, too, feature a reboiler and a condenser, their operation

is similar to that of distillation columns and so the same integration

principles can be applied to them. Absorber-stripper loops and

80 Chapter Four

studied by Kemp (2007, sections 6.3 to 6.5).

The basic Pinch Analysis calculations assume that the core process

layers in the onion diagram (Figure 4.2) remain fixed. However, it is

possibleand in some cases beneficialto alter certain properties of

the process. Properties that can be exploited as degrees of freedom

include: (1) the pressure, temperature, or conversion rate in reactors;

(2) the pressure, reflux ratio, or pump-around flow rate in distillation

columns; and (3) the pressure of feed streams in evaporators.

Such modifications will also alter the heat capacity flow rates and

temperatures of the related process streams for Heat Integration; this

will cause further changes in the shapes of the CCs and the GCC,

thereby modifying the utility targets. The CCs are a valuable tool for

suggesting beneficial process modifications (Linnhoff et al., 1982).

Figure 4.43 illustrates this application of the CCs in terms of the plus-

minus principle. The main idea is to alter the CCs slope in the proper

direction in order to reduce the amount of utilities needed. This can

be achieved by changing CP (e.g., by mass flow variation). According

to Smith (2005), such decreases in utility requirements can be brought

about by (1) increases in the total hot stream duty above the Pinch,

(2) decreases in the total cold stream duty above the Pinch,

(3) decreases in the total hot stream duty below the Pinch, and/or

(4) increases in the total cold stream duty below the Pinch.

Another guide to modifying processes is the principle of keep hot

streams hot (KHSH) and keep cold streams cold (KCSC). As illustrated in

Figure 4.44, increasing the temperature differences by process

Process Integration for Improving Energy Efficiency 81

T T

H H

Hot streams Cold streams

(a) (b)

FIGURE 4.44 (a) Keep hot streams hot; (b) keep cold streams cold.

improved heat recovery. In particular, energy targets improve if the

heating and cooling demands can be shifted across the Pinch. The

principle suggests (1) shifting hot streams (cooling demands) from

below to above the Pinch and/or (2) shifting cold streams (heating

demands) from above to below the Pinch. See Chapter 12 (and

Figure 12.7) for more details.

Most industrial-scale methods synthesize heat recovery networks

under the assumption of a steady state.

The Pinch Design Method (Linnhoff and Hindmarsh, 1983) became

popular owing to its simplicity and efficient management of

complexity. The method has evolved into a complete suite of tools for

heat recovery and design techniques for energy efficiency, including

guidelines for changing and integrating a number of energy-intensive

processes.

HEN Representation

The representation of HENs by a general process flowsheet, as in

Figure 4.45, is not always convenient. The reason is that this

representation makes it difficult to answer a number of important

questions: Where is the Pinch? What is the degree of heat recovery?

How much cooling and heating from utilities is needed?

The so-called conventional HEN flowsheet (Figure 4.46) offers a

small improvement. It shows only heat transfer operations and is

based on a simple convention: cold streams run horizontally and hot

82 Chapter Four

Feed 2 140C 4

3

230C

Reactor 2

250C

Product 2

180C

Reactor 1 80C

Off-Gas

Feed 1 40C

1

20C

2 40C

Product 1, 40C

Conventional HEN

flowsheet. 12.5 MW

140C

3

181.7C

205C 230C

H

7.0 MW 7.5 MW

203.3C

150C

8.0 MW 180C

150C

140C

6.5 MW

52.5C

1

17.5 MW

20C 106.7C

C Above the

10.0 MW

Pinch

40C 80C

simple cases, it is still difficult to see. This representation makes it

difficult to express the proper sequencing of heat exchangers and to

represent clearly the network temperatures. Furthermore, changing

Process Integration for Improving Energy Efficiency 83

the positions of some matches often results in complicated path

representations.

The grid diagram, as shown in Figure 4.47, provides a convenient

and efficient representation of HENs by eliminating the problems

just described. The grid diagram has several advantages: the

representation of streams and heat exchangers is clearer, it is more

convenient for representating temperatures, and the Pinch location

(and its implications) are clearly visible; see Figure 4.48. As in the

conventional HEN flowsheet, only heat transfer operations are shown

in the grid diagram. Temperature increases from left to right in the

grid, which is intuitive and in line with CC diagrams, making

(re)sequencing heat exchangers are straightforward.

The procedure for designing a HEN follows several simple steps:

(a) HOT 1

HOT 2

COLD 1

COLD 2 H

Simplified view

(b) MP Steam

HOT 1

C HOT 2

COLD 1

COLD 2 H

CW

between streams

C Cold utility

84 Chapter Four

recovery Pinch, as explained in Section 4.3.

3. Synthesis

4. Evolution of the HEN topology

The first two steps were discussed in previous sections. The synthesis

step begins by dividing the problem at the Pinch and then positioning

the process streams as shown in Figure 4.49.

The engineering practice suggests starting the network design

from the Pinch (the most restricted part of the design owing to

temperature differences approaching Tmin) and then to place

heat exchanger matches while moving away from the Pinch

(Figure 4.50). When placing matches, several rules have to be followed

in order to obtain a network that minimizes utility use: (1) no

exchanger may have a temperature difference smaller than Tmin;

CP

[kW/C]

PINCH

40C 150C 250C

2 15

80C 150C 200C

4 25

1 20

140C 230C

3 30

FIGURE 4.49 Dividing at the Pinch for the streams in Table 4.2.

Process Integration for Improving Energy Efficiency 85

PINCH

and (3) no inappropriate use of utilities should occur.

At the Pinch, the enthalpy balance restrictions entail that certain

matches must be made if the design is to achieve minimum utility

usage without violating the Tmin constraint; these are referred to as

essential matches. Above the Pinch, the hot streams should be cooled

only by transferring heat to cold process streams, not to utility

cooling. Therefore, all hot streams above the Pinch have to be matched

up with cold streams. This means that all hot streams entering the

Pinch must be given priority when matches are made above the

Pinch. Conversely, cold streams entering the Pinch are given priority

when matches are made below the Pinch.

Now recall the example from Table 4.2. Figure 4.49 shows the

scaled grid diagram, indicating the hot and cold Pinch temperatures.

The part above the Pinch requires essential matches for streams 2

and 4, since they are entering the Pinch. Consider stream 4. One

possibility is to match it against stream 1, as shown in Figure 4.51.

Stream 4 is the hot stream, and its CP is greater than the CP for cold

stream 1. As shown in the figure, at the Pinch the temperature

distance between the two streams is exactly equal to Tmin. Moving

away from the Pinch results in temperature convergence because the

slope of the hot stream line is less steep owing to its larger CP. Since

Tmin is the lower bound on network temperature differences, the

proposed heat exchanger match is infeasible and thus is rejected.

Another possibility for handling the cooling demand of stream 4

is to implement a match with stream 3, as shown in Figure 4.52. The

CP of stream 3 is larger than the CP of stream 4, resulting in divergent

86 Chapter Four

CP

PINCH [kW/C] T

CPHOT > CPCOLD

150C 250C

2 15

PINCH 4

150C 200C

4 25

HEAT

Tmin

EXCHANGER

MATCH 1

140C 180C

1 20 H

Temperature difference smaller

140C 230C than Tmin

3 30

INFEASIBLE!

PINCH CP

[kW/C]

T

150C 250C

2 15

PINCH

150C 200C 4

4 25

HEAT

EXCHANGER 3

MATCH

Tmin H

140C 180C

1 20 The match is feasible

140C 230C

3 30

temperature profiles in the direction away from the Pinch. Thus, the

rule above the Pinch may be expressed as follows:

The logic below the Pinch is symmetric. This part of the design is

a net heat source, which means that the heating requirements of the

cold streams have to be satisfied by matching up with hot streams.

The same type of reasoning as before yields the following requirement:

the CP value of a cold stream must not be greater than the CP value

Process Integration for Improving Energy Efficiency 87

of a hot stream if a feasible essential match is to result. Generalizing

Eq. (4.9) shows that the CP of the stream entering the Pinch must be

less than or equal to the CP of the stream leaving the Pinch:

this stage: CP tables (Linnhoff and Hindmarsh, 1983). Here the

streams are represented by their CP values, which are sorted in

descending order. This facilitates the identification of promising

combinations of streams for candidate essential matches. Sizing the

matches follows the so-called tick-off heuristic, which stipulates that

the heat exchange match should be as large as possible so that at least

one of the involved streams will be completely satisfied and then

ticked off from the design list; see Figure 4.53.

The HEN design above the Pinch is illustrated in Figure 4.54. The

design below the Pinch follows the same basic rules, with the small

difference that here it is the cold streams that define the essential

matches.

Figure 4.55 details the design below the Pinch for the considered

example. First the match between streams 4 and 1 is placed and sized

to the duty required by stream 4. The other match, between streams 2

and 1, formally violates the Pinch rule for placing essential matches.

However, since stream 1 already has another match at the Pinch, the

current match (between 4 and 1) is not strictly termed essential. Up

CP QSTREAM

[kW/C] [kW]

150C

203.3C 250C

2 15 1500

150C 200C

4 25 1250

140C 180C

1 20 800

800 kW

3 30 2700

1250

1250 kW TX =140 + =181.7C

30

88 Chapter Four

CP QSTREAM

[kW/C] [kW]

150C

203.3C 250C

2 15 1500

150C 200C

4 25 1250

140C 180C

1 20 800

800 kW

3 H 30 2700

CP QSTREAM

[kW/C] [kW]

C 2 15 1650

1000 kW

80C 150C

4 25 1750

1 20 2400

650 kW 1750 kW

to the required duty of 650 kW, this match does not violate the Tmin

constraint, which is the relevant one. The completed HEN topology

is shown in Figure 4.56.

It is not always possible to follow the basic design rules described

previously, so in some cases it is necessary to split the streams so that

heat exchange matches can be appropriately placed. Splitting may be

required in the following situations:

than the number of cold streams (NH > NC); see Figure 4.57.

2. Below the Pinch when the number of cold streams is greater

than the number of hot streams (NC > NH); see Figure 4.58.

3. When the CP values do not suggest any feasible essential

match; see Figure 4.59.

Process Integration for Improving Energy Efficiency 89

PINCH CP

[kW/C]

40C 106.7C 150C 203.3C 250C

C 2 15

1000 kW 150C

80C 200C 4 25

1 20

650 kW 1750 kW 800 kW

140C Tx =181.7 C 205C 230C

3 H 30

1250 kW 700 kW 750 kW

(a)

CP

PINCH

[kW/C]

100C

1 12

100C

2 20

100C

3 7

I?

90C T > 90

4 15

90C T > 90

5 30

(b)

CP

PINCH

[kW/C]

100C

1 12

100C

2 20

100C

3 7

90C

4 15

90C

5 30

90 Chapter Four

CP PINCH

[kW/C]

T<100 100C

35 1

T<100 100C

60 2

?

90C

24 3

90C

30 4

90C

20 5

CP PINCH

[kW/C]

100C 1

35

100C

60 2

90C

24 3

90C

30 4

90C

20 5

determined using the tick-off heuristic. Because each stream splitter

presents an additional degree of freedom, it is necessary to decide

how to divide the overall stream CP between the branches. One

possibility is illustrated in Figure 4.60. The suggested split ratio is 4:3.

This approach would completely satisfy the heating needs of the cold

stream branches.

The arrangement in Figure 4.60 is actually trivial and can be

improved upon. Unless the stream combinations impose some severe

constraints, there is a large number of possible split ratios. This fact

can be exploited to optimize the network. In many cases it is possible

to save an extra heat exchanger unit by ticking off two steams with a

Process Integration for Improving Energy Efficiency 91

(a)

Rules: CPH CPC ; NH NC PINCH CP

[kW/C]

30C 100C

1 5

40C 100C 4

2

? ?

15C 90C

3 7

(b)

Rules: CPH CPC ; NH NC PINCH CP

[kW/C]

30C 100C

1 5

40C 100C

2 4

15C 90C

3 4

PINCH CP

[kW/C]

30C 100C

C1 1 1 5

T1 = 100 300/5 = 40C

T2 = 100 225/4

C2 2 2 4

15C 90C

3 4

3

92 Chapter Four

stream 2 and the corresponding cold branch of stream 3, thereby

eliminating the second cooler.

The complete algorithm for splitting streams above the Pinch is

given in Figure 4.62. The procedure for splitting below the Pinch is

symmetrical, with the cold and hot streams switching their roles.

Network Evolution

Any HEN obtained using the design guidelines described previously

is optimal with respect to its energy requirements, but it is usually

away from the total cost optimum. Observing the Pinch division

PINCH CP

T1 = 100 285/5 [kW/C]

C1 1 1 5

40C 1 100C

2 2 4

Step 3: Q1 = 3.8 (90 15) = 285 kW

15C 90C

3 Q1 3.8

2

Q2 3.2

Step 2: CPC2 = 240(90 15) = 3.2

Stream Data

at Pinch

Yes Yes

CPH CPC NH CPC

? ?

Place No

No

Matches

Split Cold

Stream

Split Hot

Stream

Process Integration for Improving Energy Efficiency 93

typically introduces loops into the final topology and leads to larger

number of heat exchanger units. The final step in HEN design is

evolution of the topology: identifying heat load loops and open heat

load paths; then using them to optimize the network in terms of heat

loads, heat transfer area, and topology. During this phase, formerly

rigorous requirementsfor example, that all temperature differences

exceed Tmin and that cross-Pinch heat transfers be excludedare

usually relaxed. The resulting optimization formulations are typically

nonlinear and involve structural decisions, so they are MINLP

problems. Different approximations and simplifying assumptions

can be introduced to obtain linear and/or continuous formulations.

The design evolution step can even be performed manually by

breaking the loops and reducing the number of heat exchangers.

Eliminating heat exchangers from the topology is done at the expense

of shifting heat loads (from the eliminated heat recovery exchangers)

to utility exchangers: heaters and coolers. Topology evolution

terminates when the resulting energy cost increase exceeds the

projected savings in capital costs, which corresponds to a total cost

minimum.

Network evolution is performed by shifting loads within the

network toward the end of eliminating excessive heat exchangers

and/or reducing the effective heat transfer area. To shift loads, it is

necessary to exploit the degrees of freedom provided by loops and

utility paths. In this context, a loop is a circular closed path connecting

two or more heat exchangers, and a utility path connects a hot with a

cold utility or connects two utilities of the same type. Figure 4.63

shows a HEN loop and a utility path. A network may contain many

such loops and paths.

As presented so far, the Pinch Design Method is based on a sequential

strategy for the conceptual design of HENs. It first develops an

+W +U UTILITY PATH U W

40 250

C 1 4 5 2

80 200

2 3 4

20 180

1

+U LOOP U

140 230

3 H

W +W

94 Chapter Four

of process streams, and then it exploits this knowledge to design a

highly energy-efficient HEN. However, another approach has also

been developed: the superstructure approach to HEN synthesis,

which relies on developing a reducible structure (the superstructure)

of the network under consideration. An example of the spaghetti-

type HEN superstructure fragments typically generated by such

methods (Yee et al., 1990) is shown in Figure 4.64.

Typically, this kind of superstructure is developed in stages (Yee

et al., 1990) or blocks (Zhu et al., 1995), each of which is a group of

several consecutive enthalpy intervals (discussed previously). Within

each block, each hot process stream is split into a number of branches

corresponding to the number of the cold streams presented in the

block, and all cold streams are split similarly. This is followed by

matching each hot branch with each cold branch. Once developed,

the superstructure is subjected to optimization. The set of decision

variables includes the existence of the different stream split branches

and heat exchangers, the heat duties of the exchangers, and the split

fractions or flow rates of the split streams. The objective function

involves mainly the total annualized cost of the network, although

the function may be augmented by some penalty terms for dealing

with difficult constraints. Because the optimization procedure makes

structural as well as operating decisions about the network being

designed, it is called a structure-parameter2 optimization. Depending

Enthalpy Interval

Hot

Streams

Cold

Streams

2 This should not be confused with the parameter entities from Mathematical

Programming.

Process Integration for Improving Energy Efficiency 95

on what assumptions are adopted, it is possible to obtain both MILP

and MINLP formulations. Linear formulations are usually derived

by assuming isothermal mixing of the split branches and then using

piecewise linearization on the heat exchanger capital cost functions.

With the superstructure approach it is possible to include other

heat exchange optionsfor example, direct heat transfer units (i.e.,

mixing) and different heat exchanger types (e.g., double-pipe, plate-

fin, and shell-and-tube). Sorak and Kravanja (2004) presented a

method incorporating different heat exchanger types into a

superstructure block for each potential heat exchange match. Some

other interesting works in this area are by Daichendt and Grossmann

(1997), Zamora and Grossman (1998), Bjrk and Westerlund (2002),

and Frausto-Hernndez et al. (2003).

It is clear that the superstructure methodology offers some

advantages in the synthesis of process systems and, in particular, of

HENs. Among these advantages are: (1) the capacity to evaluate a

large number of structural and operating alternatives simultaneously;

(2) the possibility of automating (to a high degree) the synthesis

procedure; and (3) the ability to deal efficiently with many additional

issues, such as different heat exchanger types and additional

constraints (e.g., forbidden matches).

However, these advantages also give rise to certain weaknesses.

First, the superstructure approaches, in general, cannot eliminate the

inherent nonlinearity of the problem. Hence they resort to

linearization and simplifying assumptions, such as allowing only

isothermal mixing of split streams. Second, the transparency and

visualization of the synthesis procedure are almost completely lost,

excluding the engineer from the process. Third, the final network is

merely given as an answer to the initial problem, and it is difficult to

assess how good a solution it represents or whether a better solution

is possible. Fourth, the difficulties of computation and interpreting

the result grow dramatically with problem size; this is a consequence

of the large number of discrete alternatives to be evaluated. Finally,

the resulting networks often contain subnetworks that exhibit a

spaghetti structure: a cluster of parallel branches on several hot and

cold streams with multiple exchangers between them. Because of

how superstructures are constructed, these subnetworks often

cannot be eliminated by the solvers.

All these considerations highlight the fundamental trade-off

between applying techniques that are based on thermodynamic

insights, such as the Pinch Design Method, and relying on a

superstructure approach. It would therefore be useful to find a

combination of both approaches and, if there is, to see whether it

offers any advantages. One such middle way is the class of hybrid

synthesis methods described next. A method of this class first

96 Chapter Four

limitations of the problem; but then, instead of continuing on to

direct synthesis, it builds a reduced superstructure. At this point

the method follows the route of the classical superstructure

approaches, including structure-parameter optimization and

topology simplifications. The cycle of optimization and simplification

is usually repeated several times before the final optimal network

is obtained. All resulting networks feature a high degree of heat

recovery, though rarely the maximum possible. A key component

of this technique is avoiding the addition of unnecessary features

to the superstructure, and this is an area where Pinch Analysis can

prove helpful. A good example of a hybrid method for HEN

synthesis is the block decomposition method (Zhu, 1997).

The networks obtained by the different synthesis methods have

distinct features, which influence their total cost and their properties

of operation and control. Because the Pinch Design Method

incorporates the tick-off heuristic rule (Figure 4.53), the networks

synthesized by this method tend to have simple topologies with

few stream splits and feature minimum number of units (Linnhoff

et al., 1982). Both the tick-off rule and the Pinch principle dictate

that utility exchangers be placed last, so they are usually located

immediately before the target temperatures of the streams.

However, the tick-off rule may also result in many process streams

not having utility exchangers assigned to them, which may reduce

control efficiency. The Pinch Design Method may reduce network

flexibility because it relies on the Pinch decomposition of the

problem (Figure 4.50) and so, to a large degree, fixes the network

behavior.

Both the pure superstructure approach and the hybrid approach

tend to produce more complex topologies. Their distinctive feature is

the greater number of heat exchangers and stream splits, a result of

how the initial superstructure is built. Spaghetti-type subnetworks

also present a significant challenge to control.

The concept of the Total Site was introduced by Dhole and Linnhoff

(1993b). Figure 4.65 shows a typical industrial Total Site. Refinery and

petrochemical processes usually operate as parts of large sites or

factories. These sites have several processes serviced by a centralized

utility system involved in steam and power generation. The two

major components of Total Site integration are closely related: heat

recovery (through the steam or utility system) and power

cogeneration.

Process Integration for Improving Energy Efficiency 97

Emissions

Emissions

Power

Fuel 1

Fuel 2 Condensing HP Steam

Emissions MP Steam

LP Steam

Cooling

The heating and cooling requirements of the individual processes

are represented by their respective GCCs (see Section 4.3.5). The

GCC represents the process-utility interface for a single process.

These individual GCCs can be used to identify the potential heat

recovery via steam mains. When a site houses several production

processes, the GCC of each process may indicate certain steam levels

that are suitable for the given process. This suggests that trade-offs

need to be made among energy demands of the various processes

on a Total Site, since each process usually needs utility heating or

cooling at levels different from the other processes.

It is possible to set utility targets for sites involving several processes

(Dhole and Linnhoff, 1993b; Linnhoff and Dhole, 1993; see also

extensions by Kleme et al., 1997). The procedure is based on

thermal profiles for the entire site that are called, naturally enough,

Total Site Profiles (TSPs). These profiles are constructed from the

GCCs of the individual processes on the site. The first step (see

Figure 4.66) is to extract the net heating and cooling demands. Two

options are possible: restricting the Heat Integration to each process

(so that heat recovery pockets are not considered for Total Site

analysis) or to allow extended integration across processes by

including the GCC segments forming the pockets. When the scope

of site integration is extended, however, some design possibilities

98 Chapter Four

Shifed T**

T**

Temperature

T*

T*

PINCH

Shifed T**

1.1. Remove the pockets

1.2. Shift

H H

1. Extract segments 2. Rotate source segments

T** T**

0 H

0 H

3. Combine segments

4. Align profiles

FIGURE 4.66 Construction of the Total Site Profiles when the heat recovery

pockets are excluded from site integration.

streams that are distant, or there may be control and start-up

problems. Note that the parts of each GCC that are directly satisfied

by local utilities (e.g., furnaces within the processes) are also

excluded from the analysis; the remaining curve parts are those

representing the net heat source and sink demands to be satisfied

by the central utility system. As shown in the figure, subseqent

steps include the rotation of heat source segments (for purely

graphical reasons), the thermal combination of stream segments

(much as in the construction of process CCs), and alignment of the

resulting Total Site source profile and Total Site sink profile.

As shown in Figure 4.66, the source and sink elements extracted

from the GCCs are shifted by Tmin/2: the temperatures of the heat

source segments are reduced while those for the sinks are increased

This operation ensures that all temperatures in the picture remain in

the scale of the true utility temperatures so that, if a utility profile

touches the process-derived profiles, then there will be just enough

temperature driving force to effect the heat transfer. The composite

of the heat sources elements is the Site Source Profile and that of the

sinks is the Site Sink Profile.

Site Source Profiles and Site Sink Profiles derive primarily from

the process GCCs. Other steam requirements (mostly for process

use not directly related to heating) are usually not represented in

the GCC; examples include steam for ejectors and reactors as well

Process Integration for Improving Energy Efficiency 99

as unaccountable steam usage. These additional requirements have

to be considered when analyzing or designing the sites utility

system. Such steam demands are considered to be sink elements

and are added to the Site Sink Profile without any temperature

correction.

The maximum possible heat recovery through a utility system can be

targeted by using the Site Sink and Source Profiles in combination

with the steam header saturation temperatures. Source CCs for utility

generation and usage are constructed that account for feasible heat

transfer from the Site Source Profile to the site source CC and from

the site sink CC to the Site Sink Profile. The site CCs are analogous to

the individual process CCs. The source CC is built starting from the

highest feasible steam level. The steam generation at each level is

maximized before the next lower levels are analyzed. This ensures

maximum utilization of the heat sources temperature potential. The

remainder of the Site Source Profile (i.e., the part that does not overlap

the newly built source CC) is served by CW. Building the sink CC

follows a symmetrical procedure only starting from the lowest

possible steam level. The use of this level is maximized before moving

up to the next higher temperature level, and so forth until steam with

the highest possible level is usedincluding boiler-generated

steam.

Figure 4.67 illustrates the procedure for building the Total Site

CCs; here, the TSPs from Figure 4.66 were reused. There are several

From boilers

4

250 VHP

3

200 HP

1

Temperature [C]

2

2 MP

150

3 1

LP

100

CW

50 4

25 20 15 10 5 0 5 10 15 20 25

Enthalpy [MW]

100 Chapter Four

saturation temperatures. The main distribution levels are high-

pressure (HP) steam at a saturation temperature of 200C , medium-

pressure (MP) steam at 170C, and LP steam at 115C. In addition,

very-high-pressure (VHP) steam at 250C is generated by the steam

boilers, and CW is also available.

The Site Source Profile in Figure 4.67 can generate at most 2 MW

of HP steam, another 2 MW of MP steam, and 15 MW of LP steam.

The rest of the heat sources will need to be served by the CW, which

completes the Site Source CC. The Site Sink CC, indicates needs for 1

MW of LP steam, 6.5 MW of MP steam, 2.5 MW HP steam, and the

remaining 5 MW demand to be satisfied by VHP steam. The two site

CCs can be overlapped in the same way as were the process CCs,

thereby illustrating the Total Site heat recovery possible through the

steam system. Figure 4.68 shows the corresponding site CCs

overlapped. The amount of heat recovery for the Total Site is indicated

by the amount of overlap between the CCs. Heat recovery is

maximized when the two Site Composite Curves touch and cannot

be shifted further. The area where the curves touch, which is usually

confined between two steam levels (here, the MP and LP levels), is

the Total Site Pinch. The steam mains located at the Site Pinch feature

opposite net steam loads; in other words, the steam main above the

Site Pinch is a net steam user while the one below the Site Pinch is a

net steam generator. Just as with the Process Pinch, the site Pinch

divides an overall heat recovery problem into a net heat source and a

net heat sink.

250 VHP

200 HP

Temperature [C]

MP

150

LP

100

CW

50

25 20 15 10 5 0 5 10 15 20 25

Enthalpy [MW]

Process Integration for Improving Energy Efficiency 101

The factors that control the economics of utility systems are fuels and

their properties, the ratio of fuel prices to power prices, efficiency of

the utility system, and the amount of power to be imported/exported.

Depending on these factors and the Total Site power demand, the site

may be a net power importer or exporter or be in power balance.

Most industrial processes require steam at different pressure

levels up to about 30 bar. Central utility boilers usually generate steam

at higher pressure (40100 bar). Back-pressure steam turbines are used

to expand the steam from higher- to lower-level steam headers, thus

generating power while delivering steam to processes. Another

method of cogeneration employs a gas turbine, which is in itself a

power generating equipment. A gas turbine produces large amounts

of waste heat along with the powerthe ratio of heat to power is

about 1.52. This is high-temperature (450600C) waste heat capable

of generating even VHP steam. The heat from a gas turbines exhaust

stream can be utilized by heat recovery steam generators with or

without supplementary firing. The generated steam is expanded

through steam turbines to produce additional power.

The work of Dhole and Linnhoff (1993b) has been further

developed by Raissi (1994) and Kleme et al. (1997). The latter paper

describes the development of a tool called the Site Utility GCC

(SUGCC). The area enclosed by this curve is proportional to the

power cogeneration potential of the site steam system; Kleme et al.

(1997) also defined a simple proportionality coefficient, whose value

is usually evaluated for each industrial site separately. This

cogeneration targeting model is referred to as the T-H model

because it is based on heat flows through the steam system.

Using SUGCCs allowed Kleme and colleagues (1997) to set

thermodynamic targets for cogeneration of power along with targets

for site-scope heat recovery that would minimize the cost of utilities.

Satisfying the goal of maximum heat recovery leads to a minimum

boiler VHP steam requirement, which in turn can be achieved by

maximizing steam recovery. Here the power generation by steam

turbines is also minimal, which has the effect of maximizing

imported power. This scenario can be represented by the Site CCs

that are shifted to a position of maximum overlap (i.e., pinched). This

target represents the thermodynamic limitation on system efficiency,

but this is not a specification that must be achieved. The case of

minimizing the cost of utilities is handled by exploring the trade-off

between steam recovery and power cogeneration by steam turbines.

If design guidelines are thus based on minimizing cost, then the

resulting network design is usually different from that produced

when aiming to minimize fuel consumption.

Mavromatis and Kokossis (1998) proposed a simple model of

back-pressure steam turbine performance. In this model, the

performance of a steam turbine is related to its size (in terms of

102 Chapter Four

power is modeled as a function of the steam mass flow that is known

as the Willan line. This model was extended to condensing steam

turbines by Shang (2000). All these works follow the same model

structure and employ the same equations; however, they use different

values for the turbine regression coefficients.

The intercept of the Willan line was mapped by Mavromatis and

Kokossis (1998) and by Shang (2000) as identical to the turbine energy

losses, also assuming a fixed loss rate. Varbanov, Doyle, and Smith

(2004) introduced improvements to those models by (1) recognizing

that the Willan line intercept has no direct physical meaning and is

simply the intercept of a linearization, and (2) accounting for both

inlet and outlet pressures of the steam turbines. These improved

steam turbine models have been incorporated into methodologies for

simulating and optimizing steam networks; they have also been

used to target heat and power cogeneration by assuming a single

large steam turbine for each expansion zone between two consecutive

steam headers.

A model for optimizing the utility system serves as a tool for reducing

site operating costs related to energy and for analyzing the

thermodynamic limitations of energy conversions. An advanced

approach to these concepts, known as top-level analysis, is one that

allows scoping which site processes to target for Heat Integration

improvement (Varbanov, Doyle, and Smith, 2004). Consider the utility

system shown in Figure 4.69 (Smith and Varbanov, 2005), whose

operating properties have been optimized for the given steam and

power demands.

Suppose it were possible to reduce the HP steam demandfor

example, by improving the energy efficiency within the processes

that use HP steam. What, then, would such a saving in steam actually

be worth? Reducing HP steam demand means that less steam needs

to be expanded from the VHP level, which could lead in turn to less

power cogeneration and increased import of power. As a support

tool for deciding how best to utilize the potential steam excess and

estimate the value of potential steam savings, Varbanov, Doyle, and

Smith (2004) introduced the concept of marginal steam price. This

characteristic captures the change in a utility systems energy cost

per unit change in steam demand, and it is specific to a given

combination of steam header and operating conditions. By

optimizing the utility system at gradual successive reductions of

potential steam demand on the headers, it is possible to obtain a

curve of the marginal steam price versus the savings that could be

obtained. The marginal price curve for the utility system in

Process Integration for Improving Energy Efficiency 103

Natural Gas

Current: 250.0 Current: 55.4

[50.0; 250.0] GT

[40.0; 180.0]

HRSG 18 MW

Boiler 1 Boiler 2 Air

[115.0; 335.0]

Current: 105.0

[60.0; 165.0] VHP (101 bar abs.)

[15.0; 75.0] [15.0; 65.0] 0.0

HP (41 bar abs.)

Current: 90.0 Current: 40.9

T5 54.5

[30.0; 9.0] [30.0; 70.0] 0.0

MP (15 bar abs.)

0.0

Current: 20.0 Current: 108.0

T7 55.9

[20.0; 20.0] [60.0; 150.0] 0.0

LP (3 bar abs.)

LP Vent CW

Current: 75.0 128.0

0.073 bar abs.

[40.0; 75.0]

All flowrates are in [t/h]. The lower and upper bounds are shown in the format [min;max].

steam prices for the 8 HP Steam MP Steam

utility system shown in 7

Figure 4.69.

6

Marginal Price, $/t

5

4

3

2

1

0

0 10 20 30 40 50 60 70

Steam Savings, t/h

Figure 4.69 is shown in Figure 4.70. The plot indicates that the most

potential for improvements is in the HP steam using processes,

followed by MP users; the potential for LP steam savings is evidently

quite modest.

This page intentionally left blank

CHAPTER 5

Mass Integration

T

he two main branches of Process Integration are energy

integration and mass integration. Mass integration is a

systematic methodology that provides a fundamental

understanding of the global flow of mass within the process and

then employs this understanding to identify performance targets

and to optimize the allocation, separation, and generation of streams

and species. In the context of wastewater minimization, a mass

integration problem involves transferring mass (contaminant load)

from rich process streams to lean process streams in order to achieve

their target outlet concentration and simultaneously minimizing

waste generation and the consumption of utilitiesincluding

freshwater and external mass separating agents (Rakovi, 2006).

Water is widely used in various industries as an important raw

material. It is also frequently used in the heating and cooling utility

systems (e.g., steam production, cooling water) and as a mass

separating agent for various mass transfer operations (e.g., washing,

extraction). Strict requirements for product quality and associated

safety issues in manufacturing contribute to large amounts of

high-quality water being consumed by the industry. Stringent

environmental regulations coupled with a growing human population

that seeks improved quality of life have led to increased demand for

quality water. These developments have increased the need for

improved water management and wastewater minimization.

Adopting techniques to minimize water usage can effectively reduce

both the demand for freshwater and the amount of effluents generated

by the industry. In addition to this environmental benefit, efficient

water management reduces the costs for acquiring freshwater and

treating effluents.

A number of different methodologies have been applied to

minimizing water use and effluents. These include

and control of process operations

105

106 Chapter Five

applying advanced optimization strategies that aim to reduce

waste

Integrating optimization and production planning techniques

in conjunction with real-time plant measurements to control

for product quality and minimize losses

Increasing the use of enhanced intelligent support to operators

by applying knowledge-based decision-making procedures

to select options that best protect the environment

Employing Process Integration techniques that are based on

Pinch Analysis

features. In all cases, however, it is advisable to progress from the

simplest measuressuch as good housekeeping based on efficient

management, control, and maintenanceto more advanced

methodologies. Some processes are continuous and run seven days a

week for the whole year; others are intermittent and/or highly

dependent on availability of the feed stock. Typical of such campaign

production are plants that process sugar, fruit juice, and cereal. In

contrast, breweries operate on a nearly continuous basis, although

the processing is performed in batches. All these factors influence

investments in processing plants and the technologies adopted,

including those that involve water.

5.2.1 Legislation

Water use and wastewater discharge are both subject to national and

international standards.

For the United States, the most significant water-related federal

legislation includes: (1) the National Pollutant Discharge Elimination

System (NPDES) permit program (1972); (2) the Clean Water Act

(Federal Water Pollution Control Amendments of 1972), as amended

by the Clean Water Act of 1977; (3) the Safe Drinking Water Act of

1974; (4) the Toxic Substances Control Act of 1976; and (5) the Water

Quality Act of 1987. Individual states also legislate regarding water

for example, the comprehensive water legislation passed in California

and signed by the governor in 2009.

Legislation in member countries of the European Union (EU)

follows the directives of the European Commission (EC), which is

the executive body of the EU. The most relevant of these directives

are published on the official EU web site (EUROPA, 2009). Selected

topics and titles include:

Mass Integration 107

Water Framework Directive (2000/60/EC)

Strategies to prevent chemical pollution of surface water

under the Water Framework Directive

Priority substances under Article 16 of the Water Framework

Directive

A European action program on flood risk management

Discharges of Dangerous Substances Directive (76/464/EEC)

Water pollution stemming from urban wastewater and

certain industrial sectors; Urban Waste Water Treatment

Directive (91/271/EEC)

Water pollution caused by nitrates from agricultural sources;

Nitrates Directive (91/676/EEC)

Bathing water quality of rivers, lakes, and coastal waters;

Bathing Water Quality Directive (76/160/EEC) and proposed

revisions

Drinking water quality; Drinking Water Directive (98/83/EC)

Most of these items directly (or indirectly) concern the water used

and wastewater discharged by processing industries.

The Integrated Pollution Prevention and Control (IPPC) Directive

(96/61/EC) introduced a framework within which EU member states

are required to issue operating permits for industrial installations

performing certain activities. These permits must prescribe

conditions that are based on best available techniques (BATs). Best

available techniques are those with the best overall environmental

performance that can be introduced at a reasonable cost, and their

purpose is to ensure a high level of protection for the environment as

a whole. A key aim of the IPPC Directive is to stimulate an intensive

exchange of information on BAT between the European member

states and affected industries. The European IPPC Bureau (eippcb.

jrc.es) organizes this exchange of information and produces BAT

reference documents (BREFs), which member states must take into

account when establishing permit conditions. The bureau carries out

its work through technical working groups (TWGs) consisting of

nominated experts from industry, EU member states, countries in

the European Free Trade Association, and nongovernmental

organizations concerned with the environment. Because the

European IPPC Bureau is located in Seville, Spain, activities carried

out within the framework of the IPPC Directive are often referred to

as the Seville process.

Several BAT-oriented studies have been made in food processing

industries. A good example comes from the Flemish Centre for Best

Available Techniques (BAT-CENTRE, 2009). This document contains

108 Chapter Five

processing industry. Using BAT as guidance, the study proposes:

which investment support may be offered because they are

less detrimental to the environment

To Flemish companies: Guidelines for implementing the

concept of BAT

industry comprises the sectors of frozen fruits and vegetables, canned

fruits and vegetables, processed potatoes, peeled potatoes, and fruit

juices. The most important environmental problems are the use of

large volumes of ground water and the production of wastewater

polluted with organic carbon, nitrogen, and phosphorus. Information

on BAT candidates was obtained mostly from expertise present in

Belgium and neighboring countries. More than a hundred different

techniques were selected and examined in terms of technical and

economical feasibility. Best available techniques in wastewater

treatment incorporate, for example, primary and aerobic wastewater

treatment for small potato-peeling enterprises as well as primary,

anaerobic, and aerobic wastewater treatment for large-scale

processing sites. The BAT concept was the basis for concluding that

Flemish wastewater discharge limits on surface water were

technologically and economically feasible, although new limits

(2550 mg/L) on total phosphorus discharge were suggested. Annual

wastewater treatment costs for an average enterprise were estimated

to be 2.53.5 million. For small potato-peeling companies,

wastewater discharge into the sewer system was found to be most

appropriate.

Water-saving measures and the reuse of water may reduce

groundwater consumption by as much as 2530 percent. A good

source for BAT practices is ENVIROWISEa U.K. government

program managed by Momenta (a division of AEA Technology Plc)

and TTI (a division of Serco Ltd.)which offers practical

environmental advice for business. Their web site (ENVIROWISE,

2009) provides a wide range of information including news and best

practice examples.

Hoekstra (2008) defined the water footprint (WFP) as an indicator of

direct and indirect water use, which is measured in terms of water

volumes consumed, evaporated, and/or polluted. The WFP includes

consumptive use of virtual green, blue, and gray water. The virtual

green water content of a product is the volume of rainwater that

evaporated during the production process. For the food industry,

this volume is consumed mainly by agricultural products and

Mass Integration 109

represents the total rainwater evaporation from the fields during a

crops growing period. The virtual blue water content of a product is

the volume of surface water or groundwater that evaporates as a

result of productionfor example, the evaporation of irrigation water

from fields, irrigation canals, and storage reservoirs. The virtual grey

water content of a product is the volume of water required to dilute

pollutants in order to meet water quality standards for reuse or

discharge to the environment. A water footprint can be calculated for

any product or activity that has a well-defined group of producers

and consumers. The water footprint is a geographically and

temporally explicit indicator: it reflects not only volumes of water

consumption and pollution but also the type of water use as well as

where and when the water was used.

The idea of water life-cycle assessment has gained more interest

since the concept of a water footprint was introduced (Hoekstra and

Chapagain, 2007; Hoekstra, 2008). In food supply chains, the actual

water content of the final product is usually negligible when

compared with the virtual water content, which is the total fresh

water used during the various steps of supply and production. Aside

from the water that appears as an ingredient in prepared foods, most

water use in the food industry consists of the virtual water described

in the previous paragraph (Casani, Rouhany, and Knchel, 2005).

The most common water-using operations are as follows:

Process water: Cooling towers

Potable uses: Offices, canteens, etc.

Washing: Equipment, bottles, floors, vehicles, etc.

Rinsing: Equipment, bottles, food materials, final products

Firefighting

Transport medium

consumption was estimated to be 347.2 Mm3 in Canada (Dupont and

Renzetti, 1998) and 455 Mm3 in Germany (Fahnich, Mavrov, and

Chmiel, 1998). Other studies (Hoekstra and Chapagain, 2007; Water

Footprint Network, 2009) have reported figures for use of virtual water

in the production of some common food products; see Table 5.1.

Just as in the case of heat recovery, for water recovery it is best to

start with simple measures based on efficient management (e.g., good

housekeeping) before moving on to more advanced methodologies

(e.g., Process Integration techniques). Industrial operations are not

always run in continuous mode, since they depend to a great extent

on the availability of feed stock and the need to control quality. As

mentioned before, sugar, fruit juices, and cereal are typically

processed intermittently, and breweries operate continuously. These

110 Chapter Five

1 glass of beer (250 mL) 75

1 glass of milk (200 mL) 200

1 cup of coffee (125 mL) 140

1 cup of tea (250 mL) 35

1 slice of bread (30 g) 40

1 slice of bread (30 g) with cheese (10 g) 90

1 potato (100 g) 25

1 apple (100 g) 70

1 glass of wine (125 mL) 120

1 glass of apple juice (200 mL) 190

1 glass of orange juice (200 mL) 170

1 bag of potato crisps (200 g) 185

1 egg (40 g) 135

1 hamburger (150 g) 2400

1 tomato (70 g) 13

1 orange (100 g) 50

Industry Products (after Hoekstra and Chapagain, 2007)

those that involve water and wastewater.

A comprehensive survey of water and wastewater applications

for the processing industry can be found in the books of Smith (2005)

and Kleme, Smith, and Kim (2008). Most of these techniques for

water minimization fall into one of two groups:

in unit operations that consume fresh water. Examples

include increasing the number of stages in an extraction

process to reduce its consumption of water, changing from

wet cooling towers to air coolers, improving energy efficiency

to reduce steam demand, increasing the condensate return

from steam systems, and improved housekeeping. Good

housekeeping practices include analyzing and measuring

water use and wastage, reducing water wastage, regular

cleaning operations, and equipment maintenance.

2. Reuse, recycling, and regeneration: These options enable the

reuse of wastewater between water-consuming operations. Of

course, the presence of pollutants in wastewater streams

Mass Integration 111

must be considered so that subsequent water-using processes

will not be adversely affected. This aspect is discussed in

more detail in Section 5.2.4.

listed in the chapter by Kleme and Perry (2007b):

Mathematical optimization techniques

Efficient management and control of process operations

Integrating optimization and production planning techniques

in conjunction with real-time plant measurements to control

for product quality and minimize losses

Overview of the Measures

The task of minimizing water usage and wastewater discharge has

received considerable attention during the last few years as water has

become more costly and an environmentally strategic concern. Smith

(2005) summed up the measures applied to minimizing water usage

and wastewater as follows:

under item 1 of the listing in Section 5.2.3. Water quality can

also be improved by reducing the use of certain processing

components, such as hazardous cleaning agents, chemicals,

and additives. Additional process changes may be driven

through inspection or through process optimization via

developed technologiesfor instance, Process Integration

(Pinch Technology). Reduced consumption and increased

efficiency may be achieved either by upgrading equipment

or adopting new technologies.

2. Reuse: This is a viable strategy when wastewater from a given

operation is used directly in other operations, provided that

the pollutants in the reused water do not disturb the processes

in the downstream operations. Methods for maximizing

water reuse are detailed in Section 5.3, along with a discussion

on the use of recycled water in food processing.

3. Regeneration reuse: This is the process of purifying wastewater

from one operation and then reusing it in another operation

or process.

4. Regeneration recycling: Here the contaminants in the

wastewater are only partly eliminated before the water is

returned for use in the same process.

112 Chapter Five

of wastewater minimization. The authors examined a large number

of preassessment and assessment techniques for respectively

identifying waste minimization focus areas (opportunities) and

options (solutions) during a waste minimization audit. Blomquist

and Brown critically reviewed these techniques and assessed their

relative merits. The preassessment techniques were analyzed in

terms of their ease and speed of implementation; the assessment

techniques were evaluated in terms of their usefulness and

applicability.

Wastewater Treatment

Methodologies for wastewater handling can be subdivided into

different stages of treatment, as follows:

sticks, plastics).

Primary treatment: Removal of suspended solids by physical

or physical-chemical treatment. This process may consist of

natural sedimentation or may be assisted via adding

coagulants and/or flocculants or via centrifugation. Primary

treatment also includes neutralization, stripping (e.g.,

elimination of ammonia, NH3), and the removal of oils and

grease by flotation.

Secondary treatment: The removal of colloids and similar

matters from the wastewater. This treatment, which may

include chemical and biological processes, minimizes the

wastewaters organic load. Processes commonly used include

activated sludge treatment and anaerobic digestion, both of

which lead to the critical removal of phosphate, ammonia,

and oxygen-depleting contaminants.

Tertiary treatment: This stage comprises physical and chemical

processes that eliminate such pollutants as phosphate,

ammonia, minerals, heavy metals, and organic compounds.

The processes are viewed as a polishing phase and are

usually more expensive than conventional techniques. The

necessity of applying this type of treatment is largely dictated

by the following two factors:

(i) Meeting discharge conditions established by environ-

mental quality standards (EQS), which may be stricter

than BAT requirements. Subject contaminants include

ammonia, so-called List I and List II (BAT-CENTRE,

2009) substances, and suspended solids.

(ii) Recycling the wastewater for further use in the factory

as either process water or washing water.

Mass Integration 113

Tertiary treatment is especially important in environmentally

sensitive areas where effluent has to have low concentration levels

and loads of nitrogen and phosphorus.

Pinch Analysis was first developed for Heat Exchanger Network

synthesis and subsequently extended to yield other energy integration

applications (El-Halwagi, 1997; Kleme et al., 1997; Smith, 2005). The

analogous characteristics of heat and mass transfer allowed the

application of Pinch Analysis to the synthesis of mass exchange

networks and a series of other mass integration problems (El-Halwagi,

1997). Water Pinch Analysis emerged as a special case of mass

integration following the seminal work of Wang and Smith (1994).

However, that papers targeting technique was limited to the fixed

load problem, where water-using processes are modeled as mass

transfer operations. Later work on Water Pinch Analysis has focused

mainly on the fixed flow-rate problem, where flow-rate requirements

are viewed as the important constraints for water-using processes

(Dhole et al., 1996; Hallale, 2002; El-Halwagi, Gabriel, and Harell, 2003;

Manan, Foo, and Tan, 2004; Prakash and Shenoy, 2005).

In the context of Water Pinch Analysis, reuse means that the

effluent from one unit is used in another unit and does not reenter

the unit where it was previously used; in contrast, recycling means

that the effluent will reenter the unit where it was previously used,

usually after certain purification. In addition, one may also use a

regeneration unit (e.g., filter, stripper) to partially purify the water

stream prior to reuse or recycling (Wang and Smith, 1994).

A typical Pinch Analysis study proceeds in two stages. The first is

targeting, whereby minimum freshwater and wastewater flow rates

are set; this is followed by network design to achieve the targeted flow

rates. It is worth emphasizing that the targeting step is the primary

focus in Water Pinch Analysis. The target is needed in order to

determine how well a reuse or recycle system can actually perform in

terms of thermodynamic constraints. Establishing targets in advance

of design provides a clear picture of the mass exchange limitations of

the design problem, indicating the smallest achievable freshwater

intake and wastewater discharge. Once the targets are established, a

water network can be designed using any network design tools.

Wang and Smith (1994) described a methodology for determining

the amount of water required by a set of operations when water is

reused. They showed that significant water savings can be achieved

compared with the case when only freshwater is used. The authors

employ a simple example that makes use of the limiting Composite

Curve (CC) and incorporates four water-using operations. The

problem data is presented in Table 5.2.

114 Chapter Five

number mass flow [ppm] [ppm] [t/h]

[kg/h]

1 2 0 100 20

2 5 50 100 100

3 30 50 800 40

4 4 400 800 10

Water-Using Operations

single contaminant for four operations. The last column gives the

limiting water flow rate, which is the flow rate required by the

operation if the contaminant mass is taken up by the water between

the inlet and outlet concentrations. Note, however, that for an

operation whose inlet concentration is greater than zero, using

uncontaminated freshwater enables a lower flow rate than the

limiting water flow rate for that operation. A straightforward analysis

of the problem data, assuming that each operation uses freshwater,

reveals that the total (uncontaminated) freshwater required by the

operations is 112.5 t/h, with the four operations requiring 20, 50, 37.5,

and 5 t/h.

However, if water reuse is allowed, then an analysis that makes

use of the limiting CC produces a target for the minimum water flow

rate of 90 t/h. The limiting CC of the four water-using operations is

plotted in Figure 5.1. The water supply linewhich satisfies the water-

using operations represented by the limiting CChas its origin at zero

concentration and lies below the curve. The slope of the line is such

that it touches the CC at one point, which is termed the Water Pinch.

Other water supply lines with the same origin could be drawn, but

these would not touch the CC and thus would indicate flow rates

greater than the (preferred) minimum. If the water supply line were

drawn with a steeper slope to indicate a smaller flow rate, then the line

would actually cross the limiting CC and so could be part of an

infeasible design.

Wang and Smith (1994) provided a methodology for calculating

the minimum flow rate of water (including reuse) required to remove

contaminants from water-using operations. In addition, this paper

provided a methodology for designing a water reuse system. Figure 5.2

displays the final system design for the water operations described

in Table 5.2. The figure shows that, of the original targeted freshwater

amount of 90 t/h, 20 t/h is fed to operation 1 and 50 t/h is fed to

operation 2. The remaining 20 t/h is fed to operation 3 along with

Mass Integration 115

C [ppm]

800

MINIMUM

WATER SUPPLY

FLOWRATE 90 t/h

400

MINIMIZE

PINCH FLOWRATE

100

0

1 9 21 41 m [kg/h]

20 t/h

20 t/h 40 t/h

Operation 1 Operation 3

Feedwater Wastewater

90 t/h 90 t/h

50 t/h 5.7 t/h

Operation 2 Operation 4

44.3 t/h

FIGURE 5.2 Water treatment system designed using Water Pinch methodology.

5.7 t/h is fed to operation 4 and the remaining 44.3 t/h goes directly

to wastewater. The authors acknowledge that this design could be

evolved further to produce alternative networks.

To set the flow-rate targets for water reuse or recycling, various

graphical and tabulated targeting techniques may be employed. In

addition to the limiting CCs (Wang and Smith, 1994) mentioned

previously, the following methods have been used: the water surplus

diagram (Hallale, 2002), the material recovery Pinch diagram

(El-Halwagi, Gabriel, and Harell, 2003; Prakash and Shenoy, 2005),

116 Chapter Five

the cascade analysis technique (Manan, Foo, and Tan, 2004), and the

source CC (Bandyopadhyay, Ghanekar, and Pillai, 2006). Once the

flow-rate targets have been identified, numerous techniques can be

used to design a water network that achieves those targets. The works

just cited were developed for continuous processes, but there have

been several reported efforts to apply Water Pinch Analysis to batch

processes; these include the works of Wang and Smith (1995), Liu,

Yuan, and Luo (2007), Foo et al. (2006), and Majozi, Brouckaert, and

Buckley (2006).

Recovery Pinch Diagram

This section illustrates the targeting technique of Material Recovery

Pinch diagramMRPD (El-Halwagi, Gabriel, and Harell, 2003;

Prakash and Shenoy, 2005). Constructing an MRPD requires knowledge

about the material flow rates and loads of each process sink and

source. Given this information, one may construct an MRPD as

follows:

(SKj) into two lists, in ascending order of concentration

level (C).

2. For each source and demand, calculate the load given by the

product of its flow rate and concentration level (i.e., F C).

3. Plot the cumulative sources and demands, on a diagram of

load versus cumulative flow rate, in ascending order of their

concentration levels to form the sink and source CCs. In

order to render the problem feasible, the cumulative water

source CC must lie below the cumulative water demand CC,

ensuring that the water purity requirements are satisfied.

4. For pure fresh resources (zero concentration of impurities),

the sink and source CCs are separated horizontally until

they barely touchwith the source composite lying below

and to the right of the sink compositeas shown in

Figure 5.3(a).

5. For impure fresh resources, the source CC is shifted along an

impure fresh locus until it lies below and to the right of the

sink CC; this is shown in Figure 5.3(b).

The overlap area of the sink and source CCs represents the

maximum recovery among all sink and source streams within the

network. The point where the two composites touch is called

the Material Recovery Pinch, which is the bottleneck for maximum

recovery. The segment where the sink CC extends to the left of the

Mass Integration 117

Load Load

Sink composite

Sink composite

Source

composite

Source

composite

Impure fresh

locus

Flowrate Flowrate

Minimum Maximum Minimum waste Minimum Maximum Minimum waste

fresh source recovery discharge fresh source recovery discharge

(a) (b)

FIGURE 5.3 MRPD for (a) pure fresh resource and (b) impure fresh resource.

(to be purchased); the region where the source CC extends to

the right of the sink CC represents the minimum waste discharge

from the network (for final treatment before releasing into the

environment).

Both the minimum fresh resources needed and the minimum

waste generated by the network are network resource targets, and

they are determined before the recovery network is designed. In the

next section, the MRPD is used to establish flow-rate targets for a

case study in the production of fruit juice.

Table 5.3 shows the limiting water data for a case study involving

the production of fruit juice (Almat, Espua, and Puigjaner, 1999;

Li and Chang, 2006). The water amounts are expressed in cubic

meters (m3) per each batch. There is a start time (tS ) and an end time

(tT ) for each water sink and source. Prior to water recovery, freshwater

[m3] [ppm] [h] [h] [m3] [ppm] [h] [h]

SK1 20 0 0.5 2.5 SR1 20 5 2.5 4.5

SK2 20 6 5.0 7.0 SR2 20 14 7.0 9.0

SK3 20 15 9.5 11.5 SR3 20 20 11.5 13.5

SK4 16 5 17.0 19.0 SR4 8 25 17.0 19.0

SK5 20 7 6.0 8.0 SR5 16 10 10.5 14.5

118 Chapter Five

respectively (these values are given by the sum of the individual

water flows).

For this case study, three assumptions are made in order to

simplify the analysis:

Therefore, the process behaves as if it is operated in

continuous mode. It has been shown elsewhere (Foo, Manan,

and Tan, 2005) that a repeated batch operation achieves the

same flow targets as for equivalent continuous operation.

2. An unlimited water storage tank is always available. Thus, water

can be stored for later use.

3. Water recovery is always carried out between two consecutive

batches. In other words, water sources in an earlier batch will

be sent to water storage tanks before being reused or recycled

to the water sinks of the next batch operation. A similar

assumption was made in the paper by Shoaib and colleagues

(2008).

continuous processes can be used to identify rigorous water targets

for the case study. For purposes of this example, the MRPD is used.

The MRPD is illustrated in Figure 5.4(a), and the network that

achieves the MRPD-derived water targets is shown in Figure 5.4(b).

As shown in the figure, the minimum freshwater demand (FFW) is

determined to be 35 m3; the wastewater flow (FWW), 23 m3. When

compared with the total water flow prior to water recovery, this

represents a significant reduction of 63.5 and 72.6 percent for

freshwater and wastewater, respectively.

5.6.1 Introduction to Mathematical Optimization

Besides Water Pinch Analysis, water minimization problems have

also been solved using mathematical optimization techniques.

Various mathematical optimization approaches have been developed

to complement Water Pinch Analysis in dealing with more complex

problemsfor example, multicontaminant systems (Takama et al.,

1980; Alva-Argez, Kokossis, and Smith, 1998; Huang et al., 1999);

complex operational constraints, which include limiting the number

of pipeline connections (Hul et al., 2007); and forbidden/compulsory

matches between water-using processes (Bagajewicz and Savelski,

2001; Kim and Smith, 2004; Li and Chang, 2006). New research results

Mass Integration 119

Fww=23m3

(a) [kg]

1200

SR4

1000

800

SR3

600

SK3

400

SK5 SR2

200

SK2 SR5

SK1 SK4 SR1

0

20 40 60 80 100 120

FFW =35m3 [m3]

MRPD

(b)

20 16

SK1 SR1 ST1

4 10

SK14

20

6 SR4

SK1 SR1 ST2

5.7 8

14.3 5

FFw 0.7 15 Fww

SK1 SR1

35m3 23m3

6

ST3

8.3

SK5 SR5 16

0 5 10 15 20 t [h]

All water amounts are in m3

Network design

FIGURE 5.4 Fruit juice production: (a) MRPD; (b) network design.

and case studies have recently been published by authors from South

Africa (Gouws and Majozi, 2008a; Gouws and Majozi, 2008b), Asia

(Chen, Chang, and Lee, 2008; Ng et al., 2008; Chen et al., 2009), and

Europe (Tokos and Novak Pintari, 2009).

120 Chapter Five

This section discusses the case study of a brewery plant (Tokos and

Novak Pintari, 2009) as a means to illustrate how mathematical

optimization is used to solve water minimization problems. In the

brewery studied, the ratio (by volume) of process water consumed to

product beer sold was 6.04 : 1; this translated into 653,300 m3/y of

water consumed. In terms of the ratio set by BREF (2006), this

freshwater consumption exceeded the upper limit by 144,900 m3. In

light of these figures, the company undertook to improve its process

by retrofitting modifications to its existing water network so that the

plants usage of freshwater would be minimized.

Production at the brewery plant involves a mixture of water-

using batch and semicontinuous processes. Water-using operations

in the packaging area are operated mainly in batch mode, with the

exception of rinsers for nonreturnable glass bottles and cans.

Wastewater stream from semicontinuous processes can be reused in

batch processes with a lower purity requirement. Hence, the basic

formulation first proposed is designed to enable the efficient

integration of semicontinuous and batch water-using processes.

The continuous wastewater streams are treated as limited

freshwater sources, and the unused wastewater is discharged. In the

next step, the model is extended by including options for installing

intermediate storage tanks for the collection of unused wastewater

streams for reuse over subsequent time intervals. This particular

design modification is motivated by differences in the operating

schedules of the filling lines. The superstructure representation for

water reuse and regeneration reuse (as defined in Section 5.2.4) is

depicted in Figure 5.5.

Opportunities for regeneration reuse were analyzed in the

brewhouse and in the cellar (see Figure 5.6), since these processes

were characterized by a high concentration of contaminants. Here

the basic model is extended by installing a local (on-site) wastewater

treatment unit that can operate in either batch or continuous mode,

thereby enabling water regeneration reuse and recycling. The

scheduling of batch wastewater treatment units is performed

simultaneously so that the treatment schedule will coincide with the

fixed schedule of the batch process. The design includes the option to

install storage tanks before and after treatment; this enables

wastewater and/or purified water to be stored until required by the

treatment schedule.

As reported in Tokos and Novak Pintari (2009), the integration

of the water network in the packaging area made it possible for

wastewater from the can rinser to be reused in the pasteurization

processes. In this way, freshwater consumption could be reduced by

23 percent and the common costs of freshwater and wastewater

treatment by 22 percent. These improvements do not require the

Mass Integration 121

Local treatment unit

Purified water

Freshwater

Purified water Freshwater

Wastewater

Storage n

Operation n Storage ww

Wastewater from central

storage tank for continuous

operation

brewery plant (Tokos and Novak Pintari, 2009).

Filtration TR

system system

Fresh water

for batch processes

Water for pouring Wort boiling

Wastewater

Cellar Brewhouse

addition of any storage tanks. The net present value of the proposed

water network reconstruction is positive (at a 15 percent discount

rate), and the payback period is 0.29 years (about 15 weeks).

In the brewhouse and the cellar, the continuous water treatment

unit (nanofiltration) was selected for wastewater purification in the

optimum water network. Purification allows the water from batch

material pouring to be reused in the clean-in-place (CIP) system, and

wastewater from filtration could be reused directly for pouring the

batch material. All told, freshwater consumption could be reduced

122 Chapter Five

treatment by 27.9 percent. Investment costs for the modifications,

which require a membrane area of 83 m2, amount to 117,205. The net

present value of the optimal water network is positive at a 15 percent

discount rate, and the payback period is 1.3 years. The price of

freshwater has a significant impact on the optimal water network.

Increasing freshwater prices result in the identification of additional

opportunities for reuse and regeneration reuse. Examples include

water reuse between the rinser for nonreturnable bottles and the

bottle washer, water reuse between the pasteurizer and the bottle

washer, and regeneration reuse between wort boiling and the CIP

system.

Complete implementation of the proposed design could allow

the brewery to reduce its current freshwater demand by about

25 percent and to reduce its costs for freshwater and for wastewater

treatment by about 27 percent. Furthermore, the brewerys ratio of

water consumed to beer sold would decrease to 4.53 : 1 (from 6.04 : 1),

which is important for cleaner production and sustainable

development within the company.

5.7 Summary

Water is used in most process industries for a wide range of

applications. Today, industrial processes and systems that use water

are subject to increasingly stringent environmental regulations

concerning the discharge of effluents. Moreover, the demand for

fresh water continues to increase.

The pace of these trends has increased the need for improved

water management and wastewater minimization. Adopting water

minimization techniques can effectively reduce overall freshwater

demand in water-using processes and also reduce the amount of

effluent generated. These reductions bring reductions also in costs

incurred to acquire freshwater and treat effluents.

The field has been developing rapidly, and every year brings a

number of new and more efficient approaches. This chapter has

reviewed and demonstrated, through selected case studies, current

methodologies that have been applied to minimize water use and

wastewater in the processing industry.

CHAPTER 6

Further Applications of

Process Integration

P

rocess Integration, also called Pinch Technology, was initially

developed for energy and specifically for Heat Integration.

Details of the origin and development of Heat Integration were

given in Chapters 2 and 4. Its further development resulted in a

methodology for integrating mass transfer and water integration in

particular; this technology was described in Chapter 5. This chapter

(chapter 6) focuses on the additional applications and especially

recent developments that have expanded the generic Process

Integration ideas in various other directions. Given the rapid

development of this methodology, it is not possible to cover all recent

achievements. Nonetheless, this chapter explores several interesting

directions that have considerable potential for future development.

The evolution of Pinch Technology has allowed mass integration to

be extended to hydrogen management systems. In one of the earliest

works in this field, Alves (1999) proposed a Pinch approach to

targeting the minimum hydrogen utility. This method was based on

an analogy with process heat recovery. Just as the distribution of

energy resources in a plant can be analyzed and designed via using

Pinch Technology, so can the distribution of hydrogen resources be

handled in refineries, which typically have several potential sources

(each capable of producing a different amount of hydrogen) and

several hydrogen sinks (with varying requirements). However, the

designer has more flexibility in determining the hydrogen loads of

individual units by varying the throughput of units and operating

many processes over a range of conditions. As a result, there is

considerable potential for optimizing refinery performance.

A liquid hydrocarbon feed stream is mixed with hydrogen-rich

gas, heated, and then fed to a reactor. Part of the hydrogen is

consumed by reaction with the feed. Light hydrocarbon compounds

(methane, ethane, and propane), hydrogen sulfide (H2S), and

123

124 Chapter Six

effluent from the reactor is then cooled down and sent to a high-

pressure flash separator. The gas released in the separator is often

treated in an amine scrubber that removes H2S. Part of the gas is

vented from the process through a high-pressure purge to prevent

any buildup of hydrocarbons in the recycle. The remaining hydrogen-

rich gas is recompressed and then returned to the reactor with a

fresh hydrogen makeup stream. The liquid stream removed from the

bottom of the high-pressure separator contains some hydrogen, light

hydrocarbon gases and H2S in the solution, which is lost from the

hydrogen system. This liquid stream is sent to a low-pressure

separator, from which off-gases are taken and typically sent to a flare

or to the fuel gas system.

A two-dimensional plot of total gas flow rate versus purity

represents the mass balance of each sink and source in the hydrogen

network. A plot that combines the profiles for hydrogen demand

(dashed line) and hydrogen supply (solid line) yields the hydrogen

Composite Curve (CC) (Figure 6.1). The sink and source profiles start

at zero flow rate and proceed to higher flow rates with decreasing

purity. The circled plus signs in the figure indicate the surplus

where sources provide more hydrogen than is required by sinks.

Where the sources do not provide enough hydrogen to the sinks, a

circled minus sign appears on CCs to indicate a deficit of supply.

The area beneath the entire Sink Curve is the flow rate of pure

hydrogen that the system should provide to all the sinks. The area

beneath the Source Curve is the total amount of pure hydrogen

available from the sources.

For the hydrogen network to be feasible, there should be no

hydrogen deficit anywhere in the network; otherwise, the sources

will not be able to provide enough hydrogen to the sinks. The

hydrogen utility can be reduced by horizontally moving the curve

toward the vertical (purity) axis until the vertical segment between

the purities of the sink and the source touches the vertical axis,

(a) 1 (b) 1

0.9 0.9

0.8 0.8

0.7 0.7

Purity []

Purity []

0.6 0.6

0.5 0.5

0.4 0.4

0.3 Sink Composite Curve 0.3

0.2 0.2

0.1 0.1

0 0

0 1 2 3 4 5 6 7 8 0 0.5 1

Gas flowrate [106 standard m3/d] Hydrogen surplus [106 standard m3/d]

FIGURE 6.1 Composite Curves and hydrogen surplus diagram (after Alves, 1999).

Further Applications of Process Integration 125

thereby forming the Hydrogen Pinch. Separating the hydrogen

source and sink parts then determines the target value for the

hydrogen utility minimum flow rate.

The procedure for calculating the supply target requires varying

the flow rate of gas supplied to the system until a Hydrogen Pinch

is found. The sources from hydrogen-consuming processes or

from processes generating hydrogen as a secondary product

(dehydrogenation plants) have flow rates that are determined by

normal process operation; these rates are assumed to be fixed for the

purposes of designing a hydrogen network. However, process

hydrogen sources with variable flow rates can be regarded as

imports from external suppliers and from processes (i.e., steam

reformers or partial oxidation units) that generate hydrogen as a

main product. Those sources are hydrogen utilities.

One approach to minimizing hydrogen utility consumption is to

increase the purity of one or more sources. A hydrogen purification

system introduces an additional sink (feedstock for purification) and

two sources (purified stream and residue stream), resulting in new

targets. By employing Hydrogen Pinch Analysis, an engineer can

make the best use of hydrogen resources in order to meet new

demands and improve profitability.

Another extension of Process Integration is Oxygen Pinch Analysis

(Zhelev and Ntlhakana, 1999). The idea is to analyze the problem so

that targets are derived prior to designing a system for minimizing

oxygen consumption of the micro-organisms used for waste

degradation. The next step is to design a flowsheet that achieves the

targets. In most cases, oxygen is supplied through agitation.

Aeration requires energy, so an analysis based on the Oxygen Pinch

eventually leads back to the original application of energy

conservation. Using the chemical oxygen demand or COD (Monod,

1949) as the baseline range for organic contaminants allows one to

set quantitative targets (for oxygen solubility, residence time, and

oxidation energy load) as well as additional qualitative targets

namely, the growth rate that is a direct indicator of the age and

health of micro-organisms (Zhelev and Bhaw, 2000). Analyzing the

information in Figure 6.2 and then matching the oxygen supply line

to the CC (so they touch at the Pinch point) yields targeting information

on the growth rate of micro-organisms, oxygen solubility, residence

time, and oxidation energy load.

In the Oxygen Pinch approach, the method recurs to energy but

also incorporates extra information concerning environmental

issues. An important contribution of this method is its ability to

targetin parallel with the concentration of oxygen and the total

126 Chapter Six

stream O2 stream

I/COD

COD

COD

Substrate

supply line

Dissolved Process Slope = m/S

O2 stream stream Intercept = I/S

D I/D I/D

m-Specific growth Slope ~ growth rate, O2 solubility,

S-Saturation residence time, oxidation energy load

D-Diluation

health as assessed by their rate of reproduction.

and Pinch-Emergy

6.3.1 Simultaneous Minimization of Energy and Water Use

Water savings can be achieved through the strategic implementation

of water reuse between water-using operations. Further minimization

of freshwater usage is possible by regenerating water, which is then

recycled. The design methodology developed by Smith and colleagues

(Wang and Smith, 1994; Kuo and Smith, 1997; Gunaratnam et al.,

2005) has proven to be effective in process industries because it

provides a systematic means of establishing realistic minimum water

requirements for a site as well as conceptual design guidelines for

de-bottlenecking water systems.

In some process industry sectors (e.g., the food industry), water

use is closely linked to energy systems. The diagram shown in

Figure 6.3 explains the basic concept and the importance of

considering water and energy systems concurrently. In Figure 6.3(a),

freshwater is supplied to two water-using operations and then is

discharged in a parallel arrangement. The necessary heating or

cooling (usually through a heat exchanger) is provided according to

process requirements. This conventional practice can be significantly

improved by implementing a design for simultaneous water reuse

and heat recovery, as shown in Figure 6.3(b). Water reuse between

operations reduces water consumption, and the proposed heat

recovery between streams will reduce the need for utilities (e.g.,

steam, cooling water).

When the problem is considered jointly, finding the best energy

recovery options and water reuse schemes is an extremely complex

task because there are strong design interactions between systems

Further Applications of Process Integration 127

(a)

Operation 1

Freshwater Wastewater

Operation 2

(b)

Freshwater Operation 1

Heat recovery

Water reuse Wastewater

Operation 2

Kim, and Smith, 2005a).

for water and energy. Both the Water Pinch and the Energy Pinch

concepts have been accommodated in separate design frameworks.

However, the methodological procedure is changed when the

interactions between water reuse and energy recovery must be

considered; see Savulescu, Kim, and Smith (2005a, 2005b). Further

interesting applications have been published (Leewongtanawit and

Kim, 2009; Manan, Tea, and Alwi, 2009).

The energy-water methodology of Savulescu and Kim (2008)

follows a two-step approach: targeting and design. During the

targeting phase, theoretical minimum requirements for freshwater

and thermal utilities (hot and cold) are obtained via graphical

manipulation of streams data (i.e., water flow rate, contaminant

levels, and temperature). The purpose of the design phase is to create

a water and heat recovery network that can achieve the established

target. A useful design tool is the two-dimensional grid diagram

(Figure 6.4), which exploits the network arrangement of water

streams subject to energy recovery constraints (Savulescu, 1999;

Leewontanawit, 2005). An industrial case study conducted recently

(Leewongtanawit and Kim, 2008) showed an 18 percent reduction in

annualized cost resulting from the integrated approach (when

128 Chapter Six

Temperature

Process Process

water users water users

Overflow

Water quality

- Water source

- Water demand/sink

Savulescu and Kim, 2008).

performed). This case study clearly demonstrated the necessity of a

holistic approach to designing water and energy systems, since there

are significant benefits to simultaneous integration.

The link between Water Pinch Analysis and Oxygen Pinch Analysis

is the use of COD as the concentration variable in Water Pinch

Analysis. Here two configurations of wastewater treatment are

investigated, centralized and distributed. The analyzed system

includes the centralized biological treatment unit and several

satellite factories (sites) in the surrounding area that send some

portion of their wastes to the centralized treatment unit. The

variables of interest when analyzing wastewater treatment are the

quantity of wastewater treated and the quantity of oxygen required.

The quantity of treated wastewater gives an indication of how

much freshwater is used and thus of the water management level.

The quantity of required oxygen gives an indication not only of the

wastewater quality but also of the energy required by the wastewater

treatment process. Both the quantity and the quality of the

wastewater treated are related to the cost of the treatment process.

Further Applications of Process Integration 129

The configuration of wastewater treatment serves as an aid in

establishing whether or not wastewater treatment costs that are

based on quality and quantity differ significantly from costs based

on quantity only.

The effluent wastewater conditions obtained via the Water Pinch

method are used to plot concentration versus COD flow rate on one

set of axes. The oxygen required is then calculated using Zhelevs

(1998) method of the limiting oxygen supply line, which is constructed

as shown in Figure 6.5. First, the CC must be constructed. This is

done by plotting all the site streams on the same set of axes. These

streams are then summed within each concentration interval, and the

resulting curve is the limiting CC. The limiting oxygen supply line is

then constructed as a line drawn between the origin and the Pinch

Point. The inverse of the gradient of the limiting oxygen supply line

is the flow rate of oxygen required.

Curve for Site 1 Curve for Site 2

12000 10000

Concentration

Concentration

[mg COD/I]

[mg COD/I]

8000

5000

4000

0 0

0 4000 8000 0 5000 10000 15000

Mass Flowrate [kg COD/d] Mass Flowrate [kg COD/d]

Curve for Site 3

10000

Concentration

[mg COD/I]

5000

0

0 4000 8000

Mass Flowrate [kg COD/d]

Oxygen Supply Line

12000

Concentration

Composite

[mg COD/I]

Site 3 Curve

8000

Site 1

4000 Minimum

Site 2 O2 Supply

0

0 5000 10000 15000 20000 25000 30000

Mass Flowrate [kg COD/d]

FIGURE 6.5 Construction of the oxygen limiting supply line (after Zhelev and

Bhaw, 2000).

130 Chapter Six

The concept of emergy (embodied energy) was first developed by

Odum in the late 1980s (Odum, 1996). Along with other definitions

referring to life cycle, it may be defined in terms of solar transformity

(Brown and Ulgiati, 2004). Solar emergy is the solar energy directly

or indirectly necessary to obtain a flux of energy in a process. The

unit of emergy is the solar emergy Joule (seJ), an extensive quantity,

which denotes the available energy of a certain type (heat, electrical,

etc.) that undergoes transformations. Transformity, an intensive

quantity, is defined as the emergy input per unit of exergy (available

energy) output: seJ/J.

The first step in the practical process of emergy analysis is

collecting information for the calculation of solar transformities

ST [seJ/unit] of the chain of activities involved in making a resource

available to the process. This is the most difficult part of the

methodology because transformity databasesalthough rapidly

growing and continuously updated by researchersare not

comprehensive. The second step is the calculation of solar emergy

SE [seJ/y] followed by calculation of the solar emergy investment

SEI [seJ/g]:

SEI [seJ/g] = SE [seJ/y]/Amount [g/y] (6.2)

conceptual design stage. The Emergy Composite Curve (ECC) is

analogous to the Pinch CCs. In the ECC, solar transformity is plotted

against solar emergy; the CC is then matched up with the total

emergy investment (TEI) supply line, which is restricted by the ECC

at the Pinch point. Analyses are based on ECC benefit from using

both emergy and Pinch features.

Each stream in the ECC carries three pieces of information:

stream

2. The market potential of the stream in terms of usability:

the heat (temperature) potential of a thermal stream or the

concentration limits of a water stream

3. The streams future in terms of further usability (regenerative

reuse)

In the case of Heat Pinch Analysis, the hot and cold streams will have

different signs for this component of final emergy investment. The

sign on the required emergy investment to heat the cold streams will

be the opposite of the one for available emergy. Hence, at this level of

Further Applications of Process Integration 131

FIGURE 6.6

Composite Curve in

emergy-transformity

coordinates (after

Zhelev and Ridolfi, Pinch

ST [seJ/J or $]

2006).

Composite Curve

Total Energy

supply line

SE [seJ]

lead to minimization of the usage of expensive hot utilities.

With Emergy-Pinch analysis, as with classical Pinch Analysis,

the processes overlap on the vertical axis (temperature range,

concentration range) or, as here, the transformities range. The emergy

loads (investments) for the different processes are characterized by

relative values, which allows their graphical representation to be

freely shifted left and right in the ST/SE plot; see Figure 6.6.

The TEI is targeted by drawing the line touching the CC and

then calculating its slope. The greater the slope of the TEI line, the

smaller the rate of TEI. This minimizes the supply of combined

resources and their corresponding costs while lifting the emergy

supply line to its maximum. This limit is represented by the point

where the supply line and the CC meetthat is, the Pinch point. The

slope and the Pinch point of the emergy supply line can be used to

help compare alternative design or operational options. Transformity

is viewed as a quality parameter; when plotted against emergy

investment, it allows targeting of TEI and determination of the

maximum total transformity needed to run a given process.

Materials Reuse-Recycling, Supply Chains,

and CO2 Emissions Targeting

6.4.1 Budget-Income-Time Pinch Analysis

There are substantial benefits to be derived from applying the process

design concept to financial management. The timing, extent, and

allocation of Process Integration for minimizing the financial risk is

132 Chapter Six

possible uncertainties in model parameters. The concept of combined

resources management can lead to more realistic design solutions

while helping decision makers account for financial investments.

Because time runs in only one direction, the direction of individual

vectors and both CCs is to the right; this is illustrated in Figure 6.7

and Figure 6.8.

Zhelev (2005a, 2005b, 2005c) reports on two aspects of this broad

area: using Pinch principles to choose alternative designs, and

amalgamating financial considerations with the management of

energy and water. Several different stages can be identified in the

processes of investment, design, commissioning, and operation. By

applying traditional targeting procedures to the management of

financial resources, the following data can be obtained prior to

design: maximum investment level, minimum payback period, and

maximum benefit.

As shown in the upper part of Figure 6.9, this targeting is

analogous to other Pinch applications, such as the Water Pinch. First

a CC is constructed, after which a capital (investment) supply line

is drawn against the CC. The steeper the investment supply line,

the shorter the payback period. The steepest slope is constrained by

the CC, which meets the supply line at the Pinch point. Lifting the

Income [$]

Income [$]

Budget [$]

Budget [$]

FIGURE 6.7 Project budget and income versus time (after Zhelev, 2007).

FIGURE 6.8

Composite Curves for

project budget and

income (after Zhelev,

Budget [$/ 5 y]

2007).

Time [y]

Further Applications of Process Integration 133

capital supply line up to the maximum allows one to target both the

investment level and the expected annual benefit (Figure 6.9).

The composition of a substance is only one of several chemical and

physical properties that are essential in a chemical process. Other

common properties include acidity and alkalinity (as measured by

pH), density, viscosity, reflectivity, turbidity, color, and solubility.

The process network synthesis associated with these chemical

properties cannot be addressed by conventional mass integration

techniques, so another generic approach has been developed to deal

with this problem (Shelley and El-Halwagi, 2000; El-Halwagi et al.,

2004). For systems that are characterized by one key property,

Kazantzi and El-Halwagi (2005) introduced a Pinch-based graphical

targeting technique that establishes rigorous targets for minimum

usage of fresh materials, maximum recycling, and minimum waste

discharge.

Foo and colleagues (2006) focused on developing an algebraic

technique to solve the problem of identifying rigorous targets for

property-based recycling and reuse of materials. A key element of

these techniques is the concept of material surplus, which generalizes

the analogous concept developed for tasks of synthesizing hydrogen

and water networks (Alves and Towler, 2002; Hallale, 2002). Foo et al.

(2006) developed an algebraic approach called property cascade analysis

(PCA) to identify various performance targets for a maximum

resource recovery (MRR) network. This paper also introduced

network design techniques for the synthesis of an MRR network as

20

Investment [$]

15

10

20 40 60 80

Savings [$]

Separate stage of design

Composite Curve

Capital (investment) Supply Line

1

2

Grid

3

4

134 Chapter Six

modification strategies. The problem of designing a property-based

material reuse network is formulated as follows (see Figure 6.10):

A process is described as having a number NSK of process sinks

(units) and a number NSR of process sources (e.g., process and/or

waste streams) that can be considered for possible reuse and/or to

replace the use of fresh material. The aim is to design a network of

interconnections among the property sinks and sources such that the

overall flow rates of fresh resource and waste discharge are

minimized without depriving the sinks of adequate quality resources.

Each sink j requires a feed with flow rate Fj as well as an inlet property

pjin that satisfies the following constraints:

p min

j d p in

j d pj

max

for j 1, 2,! , N SK (6.3)

min max

where pj and pj are the specified lower and upper bounds on

admissible properties of streams to unit j. Likewise, each source i has

a given flow rate Fi and a given property pi. Also available for service

is a fresh (external) resource, with property pF, that can be purchased

to supplement the use of process sources in sinks. Each process

source may be intercepted via design and/or operating changes in

order to modify the flow rate and property of what each sink accepts

and discharges.

The Pinch diagram shown in Figure 6.11 is a convenient tool,

developed by Kazantzi and El-Halwagi (2005), that avoids the

drawbacks of traditional iterative procedures (Alves and Towler,

2002; Hallale, 2002): low visualization insight for targeting and

Processed

sources

(back to process)

Sources Segregated

sources Sinks

j=1 i=1

i=2

j=2

Property

Interception

Network

j=NSR

i=NSK

reuse network (after Foo et al., 2006).

Further Applications of Process Integration 135

Unused Materials or

Discharged Waste

U1 + U2 + U3

Source 2

U1 + U2

Load [kg/s]

U1

Source 1

Flowrate [kg/s]

Minimum

Fresh Usage

fresh usages to determine minimum fresh consumption (after Kazantzi and El-

Halwagi, 2005).

determine minimum resource targets is the Material Surplus

Composite Curve (MSCC). The MSCC was developed by Saw et al.

(2009) based on hydrogen and water surplus diagrams (Alves and

Towler, 2002; Hallale, 2002), but it eliminates the latters iterative

steps; see Figure 6.12.

The drawbacks of the graphical approach can be resolved by

using an equivalent numerical tool, the PCA. This technique is

discussed in detail by Foo et al. (2006) for a case study on solvent

recycling in metal degreasing (Kazantzi and El-Halwagi, 2005).

Applying the PCA makes it possible to use the targeted fresh solvent

flow rate to construct a balanced material sink and source composite

diagram. Foo et al. (2006) also suggested a technique for using the

Property Pinch Analysis (graphical or algebraic) to synthesize a

property network that achieves previously established resource

targets. In addition, their paper discusses applicability of the PCA

procedure to process modifications.

136 Chapter Six

Resource Deficit

Pinch

Purity

Purity

Deficit Composite

Resource Surplus Curve

resource discharge

Flowrate Flowrate

FIGURE 6.12 Construction of (a) interval flow rate diagram and (b) the MSCC

(after Saw et al., 2009).

The power of Pinch Analysis, which combines quality (e.g.,

temperature, concentration) with quantity (e.g., heat duty, mass flow),

has been successfully applied to analyze supply chains. In this case,

(reduced) time is the quality and the amount of material (e.g.,

number of units, mass) is the quantity.

The objective of the aggregate planning is to satisfy demand in a

way that maximizes profit. Demand must be anticipated and

forecasted, and production must be planned in advance for that

demand. Aggregate planning is particularly beneficial to plants

whose products encounter significant fluctuations in demand. Such

planning determines the total production level in a plant for a given

time period, rather than the quantity of each stockkeeping unit

produced.

Singhvi and Shenoy (2002) formulated the aggregate planning

problem as follows. Given the demand forecast Dt for each period t in

a planning horizon that extends over T time periods, maximize the

profit over the specified time horizon (t = 1, . . . , T) by determining

the optimum levels of the following decision variables:

time period t

Overtime Ot = amount of overtime worked in time period t

Subcontracting Ct = number of units subcontracted

(outsourced) in time period t

Workforce Wt = number of workers needed for production in

time period t

Machine capacity Mt = number of machines needed for

production in time period t

Further Applications of Process Integration 137

Inventory It = inventory at the end of time period t

Stock out St = number of units stocked out (backlogged) at

the end of time period t

time period t.

The accumulation of material balances can be expressed

mathematically as

It 1 St 1 Pt Ct Dt It St (6.4)

or

+ Current inventory (6.5)

These equations are reflected in the Supply Chain CCs used for Pinch

analysis, as shown in Figure 6.14.

Singhvi, Madhavan, and Shenoy (2004) extended the initial

methodology to the case of planning for multiple product scenarios.

Singhvi (2002) proposed the following algorithm for minimizing

inventory cost:

and produce the products in that order.

2. For products that have the same production rate, first produce

the one whose inventory holding cost is lower.

3. For products that have the same production rate and the

same inventory holding cost, first produce the one for which

demand is lower.

ACCUMULATOR

Current Net Inventory

It St

Previous Net Inventory

Ir1

Demand Dt

IN St1 Time Period t OUT

In-house Pt Subcontract Ct

PRODUCTION

FIGURE 6.13 Material balance in aggregate planning (after Singhvi and Shenoy,

2002).

138 Chapter Six

Ending inventory

6

5

Demand

Time [months]

4 Composite Curve

Pinch point

3 Ik

dkt

Production

2 Composite Curve

1 It1

Pkt + ck

0 5000 10000 15000 20000

I0

Material quantity [units]

FIGURE 6.14 Supply chain Composite Curves (after Singhvi, Madhavan, and

Shenoy, 2004).

Emission targeting via Pinch analysis was investigated in the 1990s

by Linnhoff and Dhole (1993), Dhole and Linnhoff (1993b), and

Kleme et al. (1997). The applications, which employ the Total Site

concept, address optimization within industrial facilities, not within

extended sites as regional or national energy sectors. However, a later

work (Perry, Kleme, and Bulatov, 2008) included the regional

dimension in a Total Site Analysis of integrating renewable sources

of energy.

Tan and Foo (2007) presented a further application of Pinch

Analysis to energy-sector planning under carbon emission

constraints: the Carbon Emission Pinch Analysis (CEPA). The main

problems addressed by the proposed methodology are (1) identifying

the minimum quantity of zero-emission energy resources needed to

meet the specified energy requirements and emission limits of

different sectors or regions in a system and (2) designing an energy

allocation scheme that meets the specified emission limits while

minimizing use of the energy resources. The sequence of the

proposed Pinch Analysis is as follows (Tan and Foo, 2007):

table must contain the quantity of the energy sources (Si) and

demands (Dj) and their respective emission factors (Cout,i

and Cin,j).

Arrange the energy sources and demands in order of

increasing emission factors.

Further Applications of Process Integration 139

Calculate the emission levels (SiCout,i) and limits (DjCin,j),

respectively, of the energy sources and demands.

Plot the Demand CC with the energy quantity (Dj ) as the

horizontal axis and the emissions limit (DjCin,j ) as the

vertical axis. Hence the slope of the CC at any given point

corresponds to the emissions factor (Cin,j ).

Plot the Source Composite Curve in the same manner as with

the Demand Composite Curve, but use instead the quantities

Si and SiCout,j. In this curve, the slope at any given point

corresponds to the emissions factor SiCout,j.

Superimpose the two CCs on the same graph.

Shift the source CC horizontally to the right so that it does

not cross the demand CC. In final position, the former should

lie diagonally below and to the right of the latter. The two

curves must touch each other tangentially without crossing;

their point of contact is the Pinch point.

Note the distance from the origin of the graph to the leftmost

end of the Source Composite Curve. This distance gives the

minimum amount of zero-carbon energy needed to meet the

systems specified emissions limits.

Finding the Pinch point yields valuable insights to decision

makersin particular, it identifies the system bottleneck.

The golden rule of Pinch Analysis can then be applied to

the problem: in order to meet all the specified emission limits

for the system, the zero-carbon energy resource is supplied

only to those energy demands below the Pinch point. Any

allocation of this resource above the Pinch point will either

lead to an infeasible solution or require more zero-carbon

energy than the minimum quantity established by Pinch

Analysis.

Regional Resource Management Composite Curve

A novel approach to regional resource management has been

developed that tackles simultaneously the two most important issues

with biomass supply chains: transportation and land use. The

biomass supply chain problem is complex because of the distributed

nature of biomass resources and their low energy density,

which necessitates large transportation capacities. Growing biomass

requires considerable land areas, often leading to competition with

food production. To address these problems, a two-level approach to

biomass supply chain synthesisbased on a novel Regional Energy

Clustering (REC) approachwas proposed by Lam et al. (2009). The

140 Chapter Six

clusters of zones for biomass management. The second level involves

building the regional energy transfer cascade (RETC) and the

Regional Resources Management Composite Curve (RRMCC).

The clusters of zones formed in the first level of the methodology

are designed to minimize the environmental impact of biomass

energy exchanges among the zones within the overall supply chain

network (Lam et al., 2009). The carbon footprint (CFP) is used as a

criterion for comparing various magnitudes of this impact; cost is

another obvious criterion. The main goal of clustering is to partition

the area of the considered region into smaller subareas (the clusters)

in order to form more coherent entities. Within each cluster, stronger

and more efficient interactions and biofuel exchanges result in

minimizing the environmental impact of the whole region. Figure 6.15

presents an REC algorithm, whose steps are discussed next:

system data.

Step 2. Optimize biomass exchange flows between the zones. In this

step, linear programming is used to formulate an objective

function that minimizes total CFP within the overall region.

Flowchart of

algorithm for

Regional Energy

Step 1. Specification of energy sources

Clustering.

and demands

exchange flows between zones

biomass exchange flows

their properties

End

Further Applications of Process Integration 141

Step 3. Display the optimal biomass exchange flows. A visual mapping

of interzone biomass exchanges provides critical feedback for

the decision maker. The zone centroids are plotted in two-

dimensional Cartesian coordinates.

Step 4. Form the clusters. Mixed integer linear programming (MILP)

has proven to be a convenient tool for this task.

The formed clusters should be presented visually to help document

and explain the proposed solution. For this purpose, the use of

Regional Energy Surplus-Deficit Curves (RESDCs) (see Figure 6.16

for an example) is suggested.

The RRMCC can be developed based on results obtained from the

REC algorithm. In this graphical method, the main idea of Grand

Composite Curve has been translated to the problem of regional

resource management. Figure 6.17 illustrates two ways of presenting

the RRMCC, where panels (a) and (b) employ different directions of

cascading.

The RRMCC combines information about energy surpluses and

deficits as well as land use, allowing one to assess possible trade-offs.

The quantity of the energy demand and supply (cumulative energy

balance [PJ/y]) is shown on the X axis, and the cumulative zone area

[km2] is shown on the Y axis. The RRMCC reveals several options for

tackling the problem of resources management in a region in terms

of managing land use and energy surpluses and deficits. A

demonstration case study on constructing and using the RRMCC is

presented in Chapter 11.

10 Total

Imbalance

Cumulative Energy [PJ/y]

8 3

er

ust

Cl

6

Cumulative supply curve

ter 2

4 Clus

2 r1

ste

Clu

0 20 40 60 80

Cumulative Area [km2]

FIGURE 6.16 Regional Energy Supply-Deficit Curves (after Lam et al., 2009).

142 Chapter Six

Zone 5 H A

B

Zone 1

Zone 4 C

G

F Zone 2

Zone 3 E

D

Zone 3

Zone 2 D

Zone 4 E

F G

Zone 1 C

B Zone 5

A H

Cumulative energy balance [PJ/y] Cumulative energy balance [PJ/y]

and Low-Temperature Energy

6.5.1 Decarbonization

Conventional utility systems focus on how to produce and utilize the

steam in a steam distribution network (Varbanov, Doyle, and Smith,

2004). Unlike conventional steam-based utility systems, however,

power-dominated energy systems exhibit different characteristics

because the provision of shaft (driver) power, rather than steam, is of

paramount importance. For such power systems (e.g., in natural gas

liquefaction), a key issue is selection of the most appropriate drivers

to satisfy mechanical shaft demands. The decision factors in driver

selection include the optimal number, type, and size of the drivers,

helper motors or generators, and power plantssubject to a set of

mechanical and electricity demands and relevant economic scenarios.

Zheng, Kim, and Smith (2008a) developed a holistic approach to

account for design interactions in power systems, given that driver

selection entails unique implications for the overall design; these

factors include overall cost, fuel consumption, performance, plant

availability, carbon emissions, and so forth.

Synthesis complexity increases significantly when steam systems

are considered together with power-dominated systems. This case

arises when a process requires a large amount of heat (steam) or

when a steam turbine is preferable (as a direct driver) to a gas turbine

or electric motor. In such cases, additional information is required

about the on-site power supply and the way drivers interact with

generating facilities. Implementation of a CO2 (carbon dioxide)

capture process in the plant requires extra compression duty for the

Further Applications of Process Integration 143

CO2 separation as well as a considerable amount of steam for the

stripper.

The synthesis of power-dominated energy systems is envisaged

with the aid of superstructure-based mathematical optimization.

The proposed superstructure (Figure 6.18) includes all possible

design options, and the optimization involves (1) the systematic

screening and evaluation of possible flowsheets and (2) assessing

economic trade-offs between capital costs and operating costs. As

usual, the optimization objective is to minimize the overall cost (i.e.,

capital and operating costs) while accounting for the models

constraints (Zheng, Kim, and Smith, 2008b). This task is typically

formulated as an MILP problem in which piecewise linearization is

used to capture the capital costs.

The levels of power required for compression constitute a major

component of energy consumption when cryogenic cooling is applied

to process streams. Thus, the efficient use of such cold energy

contributes to the cost-effectiveness of low-temperature processes.

Heat Integrationin particular, one of its most powerful tools, the

Grand Composite Curvehas a long history of application to saving

energy in cryogenic plants (Linnhoff et al., 1982; Linnhoff and Dhole,

1992; Smith, 2005); see Chapter 4 for details.

Pure refrigerant systems cannot avoid some degree of

thermodynamic inefficiency, since otherwise the heat exchanger(s)

would exhibit large temperature differences and this would push

the system away from thermodynamic reversibility. However, if

mixed refrigerants are used then the refrigeration cycles structure is

simplified, considerably reducing the duty requirements for

compression. The advantage of mixed refrigerants is that they

CN

.... BO-1 BO-N

.C1

VHP

ST DR/EG

HG/HM GT-N HRSG

HP

ST DR/EG

Process Stream

EM-1 PP-1

MP

ST DR/EG

EM-N PP-N LP

ST DR/EG

ExE CON

process (after Zheng, Kim, and Smith, 2008a).

144 Chapter Six

streams. Del Nogal et al. (2008) described a methodology for mixed

refrigerant system design based on a superstructure arrangement.

The problem is highly nonlinear and features many local optima, which

behave as unwanted traps for traditional deterministic optimization

methods. This paper therefore suggests that a genetic algorithm (GA)

be used to solve the optimization problem; see Figure 6.19.

The interactions between the GA and the simulator result in a set

of the best solutions found over a discretized solution space. The

preliminary solutions so obtained then serve as starting points for

standard nonlinear programming (NLP) optimization techniques to

fine-tune the results and, finally, report the optimal solution. One of

the important aspects of this model is that it ensures the feasibility of

heat recovery in every exchanger. The design produced during

optimization is then simulated, and cold and hot Composite Curves

are produced. Finally, the CCs are rigorously checked against the

stipulated Tmin.

Maintainability into Process Design

6.6.1 Integration

Current practice often views reliability as a mere afterthought to the

design process. As a result, systems go through repeated design and

redesign in search of greater process reliability, availability, and

maintainability (RAM). An alternative approach, as described by

Yin and colleagues (2009), is illustrated in Figure 6.20. This new

design for low-temperature Algorithm

energy systems (after Del

Nogal et al., 2008). Set of operating

condition with structural

options

Refrigeration Objective

simulator function

Updated power

demands

Driver selection

Integrated design

Further Applications of Process Integration 145

Start

interval and tasks to maximize

process availability while

Process meeting risk criteria

Synthesis Stage

Analysis

minimized?

Yes

FIGURE 6.20 New process design methodology with integrated RAM stage.

process design and reliability, and optimal solutions are obtained by

Process Integration.

To integrate RAM into process conceptual design, the optimal

preventive maintenance (PM) interval must be embedded into the

superstructure. The superstructure accommodates a number of

different designs, each with a different operating mode. Each so-

called available design is linked to its associated financial penalty

through its asset usage during the designs life cycle, which is

calculated as the ratio of real throughput to ideal throughput. This

ratio is equal to the sum of the availability across all operating modes.

Thus, Yin et al. (2009) defined the superstructure system availability

(SSA) as

A x

Real throughput

SSA i i (6.6)

Ideal throughput j i

146 Chapter Six

design, and xi is the mode is ratio of actual to maximum capacity for

the jth design.

6.6.2 Optimization

Within the optimization framework for integrating RAM into process

synthesis, the mathematical model aims to minimize the life-cycle

cost. In its general form (Yin and Smith, 2008), the optimization

problem can be summarized as follows:

Subject to: process model constraints, preventive maintenance

constraints, process system availability constraints

operational cost + Annual lost production

penalty + Other costs (6.7)

Process Integration

Various factorsincluding flow rate, composition, temperature, and

phasecan affect heat capacity Cp. Another factor that should be

taken into account is pressure. Polley, Panjeh Shahi, and Jegede (1990)

extended the Heat Exchanger Network (HEN) targeting procedure

by considering pressure drop. They used the following relationship

between the pressure drop P, the heat transfer coefficient h, and the

heat transfer area A:

P KAh m (6.8)

heat exchangers tube- and shell-sidespecific coefficients. The

allowable pressure drop (rather than the heat transfer coefficient) is

specified for each stream. Then the heat transfer coefficients are

calculated iteratively to minimize the total area. Thus, when

approaching area targets the design is modified based on the fixed

pressure drops rather than fixed film coefficients.

Ciric and Floudas (1989) suggested a Mathematical Programming

based, two-stage approach to HEN retrofits that includes a match

selection stage and an optimization stage. The match selection

stage uses an MILP transshipment model to select process stream

matches and match assignments. The optimization stage uses an NLP

formulation to optimize the match order and flow configuration of

matches.

Further Applications of Process Integration 147

Nie and Zhu (1999) developed a strategy for considering pressure

drop in HEN retrofits. They assumed that any additional area would

involve only a few heat exchange units in order to minimize the

piping and civil engineering work. The optimization procedure

consists of two stages. The first stage involves selecting a small

number of units that require additional area; the second stage

considers series or parallel shell arrangements for those units. The

topology change options are initially established by applying the

Network Pinch method (Asante and Zhu, 1997). Then a two-stage

optimization procedure is used to determine area distribution and

shell arrangement under pressure-drop constraints. Area distribution

and shell arrangement are the design properties that have the greatest

effect on pressure drop.

Vclavek, Novotn, and Dedkov (2003) analyzed in more detail

the circumstances under which pressure plays a significant role in

Heat Integration. The authors formulated some heuristic heat recovery

rules for combinations of process streams (tracks), not merely

individual streams.

Aspelund, Berstad, and Gundersen (2007) described a new

methodology, called extended Pinch Analysis and Design (ExPAnD),

to account for pressure drops in process synthesis that extends the

traditional Pinch Analysis to incorporate exergy calculations. The

authors focus on the thermo-mechanical exergy, which is the sum of

pressure- and temperature-based exergy. Compared with traditional

Pinch Analysis, the problem that Aspelund, Berstad, and Gundersen

(2007) consider (a subambient process) is much more complex; there

are many alternatives for the manipulation and integration of

streams. The authors also provide a number of (general and specific)

heuristics that complement the ExPAnD methodology. In a further

development, Aspelund and Gundersen (2007) used the concept of

an attainable region in proposing a graphical representation of all

possible CCs for a pressurized, subambient cold stream along with

the cooling effect of the stream expanding to its target pressure. The

attainable region is a new tool for process synthesis, extending Pinch

Analysis by explicitly accounting for pressure and including exergy

calculations. The methodology shows great promise for minimizing

total shaft work in subambient processes.

For designing a de-bottlenecking retrofit (as distinguished from

an energy-saving retrofit), Panjeshahi and Tahouni (2008) suggested

a method for optimizing pressure drop. The technique proceeds in

two main stages as follows. (1) Simulation of the existing process

operating at the desired increased throughput: additional utility is

used to maintain required temperatures in the process. (2) Area

efficiency specification for the existing network after the areaenergy

plot is used to increase throughput: a new virtual area, a pseudonetwork,

is introduced.

Zhu, Zanfir, and Kleme (2000) suggested a heat transfer enhance-

ment procedure for HEN retrofits. The methodology features a

148 Chapter Six

targeting stage and a selection stage. The approach uses the Network

Pinch Analysis to determine whether and where enhancements

should be applied in the conceptual design. One limitation of this

technique is that heat transfer enhancement is used only for taking

the place of additional area.

Extended Total Sites

Total Site targeting has established a method for analyzing the heat

sources and sinks of multiple processes and how heat can be

transferred from one process to another via a carrying medium,

such as steam (Kleme et al., 1997). This methodology has also been

used to demonstrate the concept of a locally integrated energy sector

for distributing heat among small-scale industrial plants and

domestic, business, and social premises while integrating renewable

energy sources (Perry, Kleme, and Bulatov, 2008). A conceptual

overall design for an energy sector that involves both heat and power

is illustrated in Figure 6.21. In this scenario, demands for heating/

cooling and electricity in units (e.g., dwellings, offices, hospitals,

schools) can be met locally by renewable energy sources such as

wind, solar cells, heat pumps, and/or excess heat and power from

the local industry. Locally installed boilers that consume traditional

fossil-based fuels, biomass, or waste can also be used to help meet

these requirements when demand is high or other sources are

unavailable. Heating or cooling and power that is not required by

one unit can be fed to a grid system and then passed to another unit

Electricity Grid

turbine turbine

Fossil fuels

Bio-fuels Electricity

(including Steam

waste) Hot water

Cooling utility

fuels fuels

Bio-fuels

(including waste)

FIGURE 6.21 Locally integrated energy sector with heat and power.

Further Applications of Process Integration 149

that is unable to meet its demands locally. The grid system can

distribute power (electricity) and heating in the form of hot water or

steam. In geographic locations where air conditioning is required, a

cooling distribution main could also be provided. If local sources

are unable to provide for the demands of all units in the system,

then district renewable sources can be provided. These would

include larger-scale wind turbines, solar-cell systems, heat pumps,

and combustors fed by waste from the units or by biofuels or fossil

fuels. The sources at this level would include power-generating

equipment such as turbines driven by steam or gas.

Varbanov and Kleme (2010) presented a further extension of the

Total Sites methodology that covers industrial, residential, service,

business, and agricultural customers; incorporates renewable energy

sources; and accounts for variability on both the supply and demand

sides. The challenge of increasing the share of renewables in the

energy mix can be met by integrating solar, wind, biomass, and

geothermal energy as well as by integrating some types of waste

with the fossil fuels. The availability of renewables and the energy

demands of the considered sites all vary significantly with the time

of day, period of the year, and location. Some of these factors are

unpredictable and can change quickly. Total Site Combined Heat and

Power energy systems are optimized by minimizing heat waste and

carbon footprint while maximizing economic viability. This

methodology incorporates state-of-the-art techniques of Total Site

Integration (Kleme et al., 1997), batch Heat Integration (Kemp and

Deakin, 1989), HEN sensitivity analysis (Kotjabasakis and Linnhoff,

1986), and time Pinch Analysis (Wang and Smith, 1995); it also applies

the concept of Time Slices (see Figure 6.22) to account for the

variabilities just described.

6.9 Summary

Every attempt has been made to include in this chapter the most

recent research results, but the field is developing so rapidly that

T [C] T [C]

200 MP Steam 200 MP Steam

Source SCC

Sink SCC Embed into the source profile LP Steam

LP

Steam (shifted)

100 100 Solar

HW Solar Storage HW Storage

CW Excess

Excess heat Retrieved CW

0 heat

1000 0 1000 0

1000 0 1000

H [kW]

H [kW]

FIGURE 6.22 Time Slice and Site targets for solar heat capture and storage

(CW = cooling water, HW = hot water, SCC = Site Composite Curve).

150 Chapter Six

results being published. The most frequent publishers of novel

research results are journals covering energy and cleaner technology

for example, Chemical Engineering Transactions, Applied Thermal

Engineering, Energy, the Journal of Cleaner Production, Cleaner Technologies

and Environmental Policy, and Resources, Conservation and Recycling.

New developments in Process Integration are often published by

leading chemical engineering journals, including Computers &

Chemical Engineering, Chemical Engineering Science, and the AIChE

Journal. A conference dedicated exclusively to dealing with related

topics is the Conference on Process Integration, Modelling and

Optimisation for Energy Saving and Pollution Reduction (PRES). The

Thirteenth PRES is scheduled for autumn of 2010 in Prague; the

Fourteenth PRES will be held in Italy during 2011.

CHAPTER 7

Process

Optimization

Frameworks

Process system engineering problems, including process synthesis,

are typically considered as optimization problems. The solution or

solutions of these problems are usually generated by solving the

corresponding mathematical models. However, a review of recent

publications reveals various failures in modeling process synthesis

(Friedler and Fan, 2009). An inappropriate mathematical model may

result in a nonoptimal or even an infeasible solution or the model

may be unsolvable because of its complexity. A mathematical model

should be a valid representation of the process, taking into account

all its significant features, and still be solvable.

Process optimization problems are formulated as mathematical

models, where variables correspond to decisions (e.g., the flow rate of

a stream, the amount of heat provided by high-pressure steam) and

constraints correspond to the conceptual model of the system (e.g.,

material balance). Optimization (or Mathematical Programming)

aims to find appropriate values for the variables in such a way

that (1) constraints involving these variables are satisfied and (2) a

specific functionthat is, the objective functionof these variables

is minimized (or maximized). The constraints define the search

space, while the objective function is to determine the most favorable

point or points in this space.

Mathematical models are classified according to the types of

variables (continuous or integer) and constraints (linear or nonlinear).

Therefore, a mathematical model can be linear in constraints and

objective function with continuous variables (i.e., linear programming,

LPR). Similarly, a mathematical model is viewed as a nonlinear

151

152 Chapter Seven

function is nonlinear with continuous variables. Models that include

both continuous and integer variables are classified as mixed integer

programming ones; these include mixed-integer linear programming

(MILP) and mixed-integer nonlinear programming (MINLP).

Also, linear optimization (LO) problems are usually referred to as

linear programming or LP. Similarly, NLO, MILO, and MINLO cor-

respond to NLP, MILP, and MINLP.

Linear programming problems appear in a wide range of

applications, including transportation, distribution from sources to

sinks, and management decisions (Kleme and Vaek, 1973; Kleme

et al., 1975; Kleme, 1986; Jeowski, 1990; Williams, 1999; Jeowski,

Shethna, and Castillo, 2003; El-Halwagi, 2006). LPR problems are easily

solved by the simplex method (Dantzig, 1968) and its improvements

(see, e.g., Maros, 2003a; Maros, 2003b). In most cases, NLP is difficult

to solve, and certain limitations on the constraints and objective

function may be necessary for such problems to be practically solvable

by specific methods (Seidler, Badach, and Molisz, 1980; Banerjee and

Ierapetritou, 2003; Sieniutycz and Jeowski, 2009). A general technique

for solving NLP and mixed-integer programming problems is applied

by the branch-and-bound framework (Land and Doig, 1960), where the

original complex problem is solved via systematic generation and

solution of a set of simpler subproblems.

Process synthesis is a creative activity. In fact, it is one of the

earliest actions taken by the process designer when creating the

structure, network, or flowsheet of a process to satisfy the given

requirements in terms of constraints and specifications while attaining

the prescribed objectives.

The relationships among the mathematical model, the process

being modeled, and the solver being deployed are usually complicated,

which makes it difficult to establish the most effective and valid

model. There is only limited discussion of generating mathematical

models in the literature, and the topic is treated in only a few

publications (see, e.g., Grossmann, 1990; Kovacs et al., 2000)

concerning specific areas.

In general, a process synthesis problem is defined by specifying

the available raw materials, candidate operating units, and desired

products. Each of these is given by an individual mathematical

model. The models cannot, by themselves, directly constitute the

Mathematical Programming model for the synthesis problem.

Construction of the mathematical model from these model elements

is not evident with the risk of failure. The major steps of process

synthesis are illustrated in Figure 7.1.

The main emphasis in this chapter is on an integrated framework

for model generation and solutionthat is, the P-graph framework.

Another class of methods for process synthesis is based on

heuristic rules. Implementing heuristic methods is relatively

Process Optimization Frameworks 153

FIGURE 7.1 Major steps of Cost data and constraints for the

process synthesis. operating units.

Prices and constraints for the

products and raw materials.

(MILP, MINLP, NLP)

Solution of the

Mathematical Programming model

Optimal network

(flowsheet)

Yet by their nature, heuristics are effective only at the local level.

This is because human experiences are almost always localized:

they are gained from an often limited number of encounters with (or

observations of) specific instances. For this reason, solutions that

are globally optimal are seldom obtainable via heuristic methods

alone (Feng and Fan, 1996).

There are four good reasons to employ graph-theoretic methods:

(1) the unambiguous representation of decision alternatives, (2) the

algorithmic generation of a mathematical model, (3) the reduced

complexity of the solution procedure, and (4) the derivation of multiple

alternative solutions. The P-graph or process graph framework, as

applied by Friedler and Fan (Friedler et al., 1992a; Friedler et al., 1992b;

Friedler, Varga, and Fan, 1995) to process synthesis, involves novel

structural representations of complex processes coupled with

combinatorial algorithms for generating the superstructure, the

mathematical model, and the models optimal solution.

The P-graph framework is robust, and its algorithms have been

validated as mathematically rigorous in that they are based on a set

of axioms (Friedler et al., 1992b). These axioms express the necessary

154 Chapter Seven

algorithms are able to guarantee the resultant mathematical models

validity, reduce the search space, and generate the optimal solution.

In a P-graph, one class of nodes is assigned to operating units or

activities and the other is assigned to their inputs and outputs. Raw

materials, resources (precursors), and preconditions (activating

entities) are inputs to the operating units; products, effects (resulting

entities), and targets are outputs from the operating units. Table 7.1

shows the P-graph representation of process structure elements.

In a process network, functional units that perform operations

(e.g., mixing, reacting, separating) are termed operating units. These

operating units, which correspond to the blocks in a process

flowsheet, alter the physical and/or chemical states of materials being

processed or transported. Such transformations are carried out by

one or more unit operations, and the overall process converts raw

materials into the desired product(s). A process may also generate

by-products, which are either to be recovered for further use or to be

treated as waste.

In process network synthesis, a material is uniquely defined by

its components and their concentrationsin other words, by its

composition, which is identified by a symbol used to mark the

material. Associated with any operating unit are two classes of

materials (or material streams): input materials and output materials.

For example, operating unit O2 in Figure 7.2 consumes raw materials

E and F while producing intermediate material C and by-product B.

Note that a material may consist of more than one component.

The P-graph provides not only a formal description of the process

but also an unambiguous representation of the possibilities for

structural decisions. If an operating unit requires multiple inputs,

By-product

Operating unit

Process Optimization Frameworks 155

representing the

process structure of

three operating

units. O2

O2 O3 O3

B C D B C

O1

O1

A A

cannot be defined. In contrast, if multiple operating units are capable

of providing a particular input, then any combination of these units

may eventually be used. In Figure 7.2(a), for example, materials C and

D are necessary inputs to operating unit O1. Material C can only be

produced by operating unit O2, and material D can only be produced

by operating unit O3. For unit O1 to operate it is necessary that units

O2 and O3 both be included in the process structure. In Figure 7.2(b),

however, material C can be produced by unit O1, unit O3, or both. In

addition to unambiguous structural representation, the P-graph

framework also provides a set of rigorous and effective algorithms

for the synthesis and optimization of process networks.

The extreme complexity of process network synthesis is due mainly

to the problems combinatorial nature. This complexity grows

exponentially with the number n of candidate operating units,

because the optimal network must be found among 2n possible

combinations of the units (i.e., alternative networks) unless some

possibilities can be eliminated (e.g., by heuristics) in advance. The

factor 2n is derived by simple induction. First observe that a single

additional decision (regarding the inclusion or exclusion of an

operating unit) doubles the number of potential design alternatives:

2n 2 = 2n+1. This means that a designer contemplating a system with

a total of 35 operating units is faced with more than 34 billion

(235 = 3.436 1010) alternative arrangements!

Reducing such large numbers of alternatives requires robust

decision-making tools that are mathematically rigorous (preferably

axiomatic) and effectively implementable on computers. These ends

156 Chapter Seven

of graph theory, which can be regarded as a branch of combinatorics.

Thus was developed the graph-theoretic, algorithmic method

described in this section. The method is based on using P-graphs to

extract the universal combinatorial features (properties) inherent in

feasible processes. Such properties can be expressed mathematically

as a set of axioms that characterize the combinatorial feasibility of

processing networks.

A given process network is said to be combinatorially feasible (or to

be a solution structure) if it satisfies the following five structural

axioms:

structure.

(S2) An entity represented in the structure has no input if and

only if it represents a raw material or precursor.

(S3) Every operating unit represented in the structure is

defined in the problem.

(S4) Any operating unit represented in the structure has at

least one path leading to a final product or a final target.

(S5) An entity belongs to the structure if and only if it is either

an input entity to or an output entity from at least one

operating unit already represented in the structure.

that results from this approach. The universe of all possible networks

is reduced to a much smaller space containing only those networks

that satisfy the axiomsin other words, the combinatorially feasible

(CF) networks. Clearly this reduction will drastically reduce the

required computational effort. Search-space reductions by a factor of

nearly a billion have been reported in some of the real-life process

synthesis tasks performed to date using this axiomatic approach.

Note that each feasible network, including the optimal network, is an

element of the set of combinatorially feasible networks.

Figure 7.4 depicts two process structures that are not combina-

torially feasible. The P-graph in Figure 7.4(a) shows a process structure

in which material F is consumed as an input. Yet because material F is

not a raw material and was never produced, the structure is not

combinatorially feasible according to Axiom (S2). In the P-graph of

Figure 7.4(b), operating unit O3 produces only by-product B. Here O3

does not output any final product or material that is later used to yield

a final product, so the process structure violates Axiom (S4). In short,

the structural properties expressed by Axioms (S1)(S5) are necessary

conditions for process structures to be feasible. This means that

reducing the search space to combinatorially feasible structures does

not result in the loss of any practically feasible or optimal processes.

Process Optimization Frameworks 157

Potential networks

(search space)

CF F

Optimal network

(F = feasible networks, CF = combinatorially feasible networks).

representing process

structures that violate

(a) Axiom (S2) or

(b) Axiom (S4).

O2 O3 O2 O3

B C D C B

O1 O1

A A

SSG, and ABB

When combined with the structural axioms, P-graph representation

makes it possible to implement effective algorithms for structural

analysis, synthesis, and optimization of process structures. The

maximal structure generation (MSG) algorithm (Friedler et al., 1992a)

generates a superstructure that can be rigorously proved to

incorporate each combinatorially feasible process structure. Then the

solution structures generation (SSG) algorithm (Friedler, Varga, and Fan,

158 Chapter Seven

structures that satisfy Axioms (S1)(S5) or the accelerated branch and

bound (ABB) algorithm (Friedler et al., 1996) is used to generate the

optimal process structure together with a ranked, finite list of near-

optimal structures.

Figure 7.5 illustrates the connections among the three

algorithms. Algorithm MSG generates the maximal structure, and

it can be followed either by algorithm SSG to generate all

combinatorially feasible process structures or by algorithm ABB to

generate the optimal and near-optimal processes. Algorithms MSG

and SSG require as input the list of candidate operating units, each

defined by the set of its input materials (preconditions) and

products (effects). Algorithm ABB requires, in addition to these

data, quantitative information (e.g., prices of raw materials, costs

and capacity constraints of operating units) relevant to assessing

network optimality.

These algorithms and their description are available online

(www.p-graph.com). The methodology has been demonstrated via

typical engineering decision problems (see Chapter 8 for details). A

considerable advantage of the P-graph framework is its potential for

solving large industrial problems, as indicated by its application in

solving various process systems engineering problems (see, e.g., Liu

et al. 2004; Halasz, Povoden, and Narodoslawsky, 2005; Liu et al.,

2006; Fan et al., 2008; Varbanov and Friedler, 2008).

Algorithm MSG

Maximal Structure

Structures (n-best Processes)

(for further evaluation)

FIGURE 7.5 Inputs to and outputs from the three P-graph algorithms.

Process Optimization Frameworks 159

7.3.1 Scheduling Frameworks: Suitability and Limitations

Scheduling problems arise in many various areas, including chemical

engineering, supply chain management, operating systems design,

or train timetable design. The terminology and details vary among

the different types of applications, but the core of scheduling remains

the same: assigning tasks to resources at time intervals that satisfy

predefined conditions. The goal is usually to find a feasible schedule

that performs betterin terms of a particular objectivethan any

alternative.

One of the first contributions to scheduling of chemical processes

by Mathematical Programming was the paper by Kondili, Pantelides,

and Sargent (1993). These authors developed the state task network

(STN) representation in order to formulate production scheduling

in multipurpose plants as a MILP problem. Pantelides (1993) also

developed the resource task network (RTN) representation, which

employs a uniform treatment for all available resources. Originally,

the time representation in the formulations are discretein other

words, a certain number of predefined times on the time horizon can

be considered in the schedule. Zhang and Sargent (1996) extended the

formulation to accommodate continuous-time representation, where

the time points may vary, their number still have to be predefined.

Ierapetritou and Floudas (1998a) employed the concept of unit-specific

event points. Majozi and Zhu (2001) reformulated the problem to reduce

the number of variables in certain applications. Cerd, Henning,

and Grossmann (1997) and Mndez and Cerd (2003) developed

precedence-based models suitable for cases where sequence-dependent

changeovers must be considered.

The rest of this section addresses some issues associated with

conventional approaches to scheduling. These issues motivated the

development of a novel, graph-theoretic approach: the S-graph

framework (Sanmart, Friedler, and Puigjaner, 1998; Sanmart et al.,

2002), which is introduced in Section 7.3.2.

In most MILP formulations, the time horizon is divided into time

intervals by so-called time points, which denote the possible starting

and ending times of tasks. The number of time points is one of the

models parameters, so it must be specified prior to optimization.

Even though the quality of the solution depends strongly on this

parameter, the minimum number required for the optimal solution

is not known in advance. Therefore, an iterative approach is applied

to determining the number of time points. First the model is solved

using a small number of time points, after which each subsequent

iteration increases that number by 1 until the same objective value is

obtained for, say, two consecutive steps. However, it is not certain

that this initial convergence necessarily yields the optimal solution.

160 Chapter Seven

which the objective function value increased after this convergence,

as illustrated in Figure 7.6. A serious shortcoming of this approach is

that it may generate a suboptimal solution.

Another modeling issue may arise when one attempts to solve an

MILP model of scheduling with no intermediate storage: according

to Hegyhti and colleagues (2009), the optimal solution that is

generated may be infeasible. In particular, the Gantt chart of a

schedulewhich was reported in two independent journal articles

(Kim et al., 2000; Mndez et al., 2003) as the optimal solution for a

case studyis not a feasible solution; see Figure 7.7.

The figure reveals that, at 30 hours into production, three units

(U2, U3, and Storage) attempt to exchange materials. However, this

is infeasible because there is no intermediate storage and so the

optimal schedule cannot be implemented in practice. The true

optimal solution of the problem is obtained (Hegyhti et al., 2009)

by using the S-graph framework (Holczinger et al., 2002); see

Figure 7.8.

Objective

FIGURE 7.6 Increase in value of the objective function after initial convergence

for a maximum throughput problem.

U1 B A C

U2 B C D

U3 C D B A

U4 D B A

Storage C D A

5 10 15 20 25 30 35 40 45 50 55 60

(Kim et al., 2000).

Process Optimization Frameworks 161

U1 B A C

U2 B C D

U3 D B C A

U4 D B A

Storage D

5 10 15 20 25 30 35 40 45 50 55 60 65 70

et al., 2009).

The problems discussed in Section 7.3.1 motivated the development

of an alternative methodology, the S-graph or schedule graph

framework (Sanmart, Friedler, and Puigjaner, 1998; Sanmart et al.,

2002), which has been successfully applied to the minimization of

time required to complete all tasks (the makespan; see Sanmart

et al., 2002, and Romero et al., 2004) and also to problems of

maximizing throughput (Majozi and Friedler, 2006). Basics of the

S-graph framework are explained in this chapter; Chapter 9

describes a demonstration program for this framework that is

available online (www.s-graph.com). Once all processing tasks

have been represented in the recipe, the S-graph can be used to

generate an optimal schedule.

A recipe defines the order of tasks in the process, the material

transfers among them, and the set of plausible equipment units for

each task. The recipe is represented as an S-graph by assigning a

node to each task (task node) and one node to each product (product

node). An arc is established between nodes of the consecutive tasks

defined by the recipe, and there is an arc also from each product-

generating task node to the corresponding product node. The weight

of an arc is given as the processing time of the task that corresponds

to the arcs initial node assuming a single equipment unit is available

for the task. If more than one equipment unit can perform this task,

then the arcs weight is given as the shortest processing time of all the

feasible units. In the graph representing the recipe, the set of plausible

units capable of performing the given task is shown in the task node;

see Figure 7.9.

Suppose that two batches of product A and one batch of product

B are to be produced, where product A is produced in two consecutive

steps. Task 1 can be performed by equipment unit E1 and task 2 by

either E2 or E3. Product B is produced in three consecutive steps that

can be performed by any of the elements in sets {E1, E3}, {E1}, and {E1,

E2}, respectively. The recipe is shown in Figure 7.9.

162 Chapter Seven

performed by an equipment unit defines the solution of a

scheduling problem. The schedule is given with an S-graph

containing arcs additional to the S-graph representing the recipe.

Moreover, a single equipment unit is assigned to each task. As an

example, Figure 7.10 depicts an S-graph representing a recipe and

Figure 7.11 a corresponding solution.

1 2

{E1} {E2,E3} 8 A

6 9

3 4

{E1} {E2,E3} 9 A

6 9

5 6 7

{E1,E3} {E1} {E1,E2} 10 B

14 16 8

FIGURE 7.9 S-graph representing the recipe for two batches of product A and

one batch of product B.

{E1} 6 {E3} 9 {E2} 7

4 5 6 11

{E2} 9 {E3} 15 {E1} 17

7 8 9

{E1} 14 {E2} 16 {E3} 8 12

Schedule #1

10

the recipe under NIS policy. {E1} {E3} {E2}

4 5 6

11

{E2} {E3} {E1}

7 8 9

12

{E1} {E2} {E3}

Schedule #1

Process Optimization Frameworks 163

The algorithm generating the optimal schedule depends on the

storage policy to be considered: nonintermediate storage (NIS), finite

intermediate storage (FIS), or unlimited intermediate storage (UIS).

In this chapter we assume a NIS policy, so an equipment unit

becomes available only after finishing a task and transferring its

intermediate product to the subsequent task in the recipe. On an

S-graph representing a schedule under NIS policy, an arc leads from

the node subsequent in the recipe to the node of the task to be

performed next by the same equipment unit. For example, equipment

E1 first performs task 6, then moves to task 1 and finally to task 7,

which is represented by arcs drawn from node 11 (subsequent to

task 6) to node 1 and from node 2 (subsequent to node 1) to node 7 in

Figure 7.11.

The advantage of the S-graph framework over conventional

Mathematical Programming lies in its ability to exploit the problems

structure to effect a drastic reduction on computational intensity

without requiring unknown information, such as the number of time

points. (Visit www.s-graph.com for further information.)

This page intentionally left blank

CHAPTER 8

Combined

Process Integration

and Optimization

Process Integration (PI), as defined in Chapter 2, is a family of

optimization methodologies for reducing resource or emissions

intensity of the analyzed processes and Total Sites. As such, it is

tightly related to optimization. In fact, PI and optimization

complement each other by their functionality. First, PI sets out the

strategy for designing and/or operating industrial processes. This

gives engineers some direction regarding how processes can be

designed or changed, answering the questions of where we can go?

and what is to be done? in order to achieve the business goals at

hand. In addition, PI provides quantitative targets for designers and

engineers; it does this by exploiting the physical (in the case of Heat

Integration, thermodynamical) background to answer the question

how much is it possible to improve or achieve? The targets in most

cases are upper bounds on the process performance or lower bounds

on the extent of resource use or emissions. In many cases the targets

are practically achievable, as in the case of designing Heat Exchanger

Networks (HENs) or water networks. The most obvious example is

heat recovery targets established by the Pinch point when analyzing

HEN problems. Since they are based on the Second Law of

Thermodynamics, it is proven that better heat recovery cannot be

achieved by any feasible system. If a realistic value for Tmin is

specified for their evaluation, the targets will be also practically

achievable. Of course, the additional factor of the capital costs for

implementing the heat exchangers generally tends to shift the

economic optimum of the designed HEN away from the maximum

recovery network.

165

166 Chapter Eight

problems. As a result, optimization is used by PI to answer the

question of how should the task be performed? The general goals

and specific targets are usually achieved by employing optimization

tools at various stages. For instance, process performance targets are

typically evaluated by employing a numerical technique that involves

a cascade of some sorta heat cascade in the case of heat recovery

targeting. One way to implement such cascades is by using the

transshipment optimization formulations, where the external utility

use, resource intake, or emissions rate is set as the objective function

to be minimized. Once the PI goals are established, engineers strive

to achieve the best possible performance. In the case of grassroots

design or network synthesis, the criterion is minimization of the total

annualized cost; in the case of a retrofit, the main criterion may be

minimizing the investments necessary to achieve a certain

performance improvement or minimizing the payback period for a

given investment. For operational improvements, the criteria include

minimizing operating costs or maximizing marginal financial or

performance gains. In all cases, a certain system modelincluding

the appropriate objective functionis formulated. The model is then

subjected to optimization toward the end of achieving (or maximally

approaching) the PI targets. Another function of targets is to partition

complex optimization problems into sets of simpler problems that are

easier to solve. This approach exemplifies the problem decomposition

principle, applied for decades in the world of software development,

also known as the divide and conquer strategy.

For optimizing process models, a wide variety of linear programming

(LPR), nonlinear programming (NLP), and mixed-integer programm-

ing (MIP) methods can be used, depending on the nature of the

problem being solved. Some of these methods were described in

Chapters 3 and 7. Special tools and software (see Chapter 9)

incorporating optimization methods have been developed to exploit

PI possibilities when performing process synthesis, accounting for

the interactions between process operating conditions and the

networks for resource recovery (energy and water). There are four

main groups of optimization tools applied for PI. First, the Pinch

Analysis (Linnhoff et al., 1982) enabled industrial engineers to

obtain better results with the simple Pinch Design Method than

with Mathematical Programming methods in applications to

industrial Heat Integration; see Chapter 4. Second, the graph-

theoretic method is based on process graphs (P-graphs), which were

originally developed for Process Network Synthesis (PNS) (Friedler

et al., 1992b; Friedler, Fan, and Imreh, 1998); see Chapter 7. Third,

Papoulias and Grossmann (1983) introduced linear constraints in

Combined Process Integration and Optimization 167

their transshipment model for Heat Integration within a mixed-

integer linear programming (MILP) formulation for structural

process optimization. This work had been further extensively

developed (Duran and Grossmann, 1986; Floudas and Grossmann,

1987a; Floudas and Grossmann, 1987b). Fourth, stochastic optimization

has become popular in recent years, applying genetic optimization

(Shopova and Vaklieva-Bancheva, 2006) and especially simulated

annealing (Kirkpatrick, Gelatt, and Vecchi, 1983; Faber, Jockenhvel,

and Tsatsaronis, 2005; Hul et al., 2007; Tan, 2007).

Optimization methods can be classified according to the

characteristics of the objective function, the decision variables, and

the problem constraints (Guinand, 2001). A simplified classification

scheme for optimization methods is illustrated in Figure 8.1.

A process network uses a given set of operating units to create desired

products from specific raw materials. The objective of PNS is to

identify the most favorable (optimal) network for accomplishing the

given tasks. The P-graph methodology is a graph-theoretical approach

to solve PNS problems.

Every reaction is a material transformation, which corresponds to an

operating unit when mapped on a P-graph. Similarly, the maximal

Nonlinear

Linear Programming Integer Programming

Programming

(LPR) (IP)

(NLP)

Programming (ILP) Programming (MILP) Programming (MINLP)

Incorporation of Incorporation of

Probability Functions Time Domain

168 Chapter Eight

Generation (MSG) algorithm. From that, all feasible reaction networks

can be generated as solution structures. Historically, it has been

extremely difficult to construct an exact maximal reaction network.

The problem has become solvable only since the arrival of the

combinatorial approach based on P-graphs. Reaction network

synthesis is completely combinatorial in nature because all chemical

species participating in the reactions are defined discretely.

The application of PNS to Reaction Network Synthesis (RNS) is

illustrated by a chemical process for the manufacturing of vinyl

chloride (C2H3Cl). Because of its structural simplicity, the balanced

synthesis of vinyl chloride (C2H3Cl) leads to only a single complex

route containing one loop. It is a trivial example from the standpoint

of constructing the maximal reaction network and the feasible

reaction networks, but it will serve to illustrate the RNS methodology.

The balanced process aims to produce vinyl chloride (C2H3Cl) and

water (H2O), where the former is the desired product and the latter is

the by-product, from the three starting reactants: ethylene (C2H4),

chlorine (Cl2), and oxygen (O2). This process involves the following

three unit reactions: R1, R2, and R3.

R2: 2C2H4 + 4HCl + O2 = 2C2H4Cl2 + 2H2O

R3: 4C2H4Cl2 = 4C2H3Cl + 4HCl

Note that C2H3Cl is the final target and that C2H4, Cl2, and O2 are the

starting reactants. From the perspective of materials (M), we have:

Starting reactants Final products

Intermediates

M { C 2H 4 , Cl 2 , O 2 , C 2H 3 Cl , H N O , C 2 4 Cl 2 , HCl }

H

2

One or more of the feasible paths and valid vertices in the input

structure may disappear if some of the invalid vertices are eliminated.

Thus, the final maximal structure is composed (or reconstructed)

from the remaining skeleton of the input structure after the

elimination. This is accomplished step by step, linking alternately

the vertices of the M-type (for materials) to the vertices of the O-type

(for operating units) and vice versa. At each step, the vertices linked

are assessed in view of the appropriate axioms (see Section 7.2.2):

vertices of the M-type must satisfy axioms (S1), (S2), and (S5); and

vertices of the O-type must satisfy axioms (S3) and (S4). The execution

is initiated from the structures shallowest layerthat is, the final,

desired product end. The stepwise procedure for the composition is

illustrated in Figure 8.2.

Combined Process Integration and Optimization 169

Cl2 C2H4 O2

R1 R2

R3

C2H3Cl HCl

R1 R2 R1 R2

R3 R3

Step 3 Step 4

C2H4Cl2

R3 R3

Step 1 Step 2

FIGURE 8.2 Steps for the composition of the maximal structure representing

the maximal reaction network for the manufacture of vinyl chloride.

SSG and Accelerated Branch-and-Bound (ABB). Algorithm SSG

generates all combinatorially feasible reaction networks, and algorithm

ABB identifies a set of the most favorable reaction networks.

Synthesis of Optimal Workflow Structures

Workflow technology has become a general tool for a wide range of

business and management applications (Tick, Kovcs, and Friedler,

170 Chapter Eight

business and management systems. The mathematical foundation of

this technology has been developed mainly based on Petri net theory

(Kiepuszewski, ter Hofstede, and van der Aalst, 2003; van der Aalst

and ter Hofstede, 2005). Even though this mathematical methodology

provides a basis for determining the optimal operation of workflows,

it cannot be used to derive an optimal workflow structure.

The structural component of a workflow synthesis problem can

be identified by the sets of products, resources, and (plausible)

activities on materials. The cost of a workflow process that generates

a particular quantity of product is given as the sum of (1) the cost of

the raw materials and (2) the cost related to the activities appearing

in the synthesized workflow process. The cost of an activity is the

sum of its running cost and the investments assigned to the period of

time examined. Both the running cost and investment cost depend

on the size of the activitythat is, its output volume. The common

objective for synthesizing workflow processes is to minimize the

total cost under the assumption of unlimited intermediate storage

capacities for any activity.

Example 8.1: Workflow Synthesis (after Tick, Kovcs, and Friedler, 2006)

As an example, a set of activities is given by its inputs and outputs in Table 8.1

and represented by P-graph in Figure 8.3.

The P-graph contains the interconnections among the activities. Each feasible

activity network corresponds to a subgraph of the P-graph in Figure 8.3.

A product document represented by A and B can be generated by an appropriate

network of the activitiesprovided that the problem has at least one feasible

solution. It is important to note that a product can usually be generated by

different types and numbers of activities. When determining the optimal

network for a workflow, all possible networks of each product must be taken

into account.

1 C A, F

2 D B

3 E, F C

4 F, G C

5 G, H D

6 H B

7 J F

8 K G

9 K G

10 L H

Workflow

Combined Process Integration and Optimization 171

J K L

7 8 9 10

E F G H

3 4 5 6

C D

1 2

A B

FIGURE 8.3 P-graph, where A, . . . , L are the materials and 1, . . . , 10 are the

activities.

is needed to determine the optimal network. For this purpose, the P-graph

framework is proposed. Algorithm MSG verified that the P-graph in Figure 8.3

corresponds to the maximal structure, and algorithm ABB provided an optimal

set of several alternative structures containing the strict optimum and other

mathematically suboptimal networks. One of the 50 solution networks for this

problem is given in Figure 8.4.

P-graph is also employed to examine the contribution that process

synthesis can offer to the development of production based on

renewable resources (Halasz, Povoden, and Narodoslawsky, 2005).

These resources face formidable competition from fossil raw materials.

Because their production is usually decentralized, viable processes

can result only if the structure of the complete value chain is optimized.

In contrast to conventional chemical processes, those using renewable

resources have to account for logistic operations affecting the value

chains structure. Hence, using process synthesis in this application

calls for using the highest-quality method for synthesis. The P-graph

method, as described in Chapter 7, has proven to be extremely efficient

and flexible, so it is well suited for this purpose.

The components of this example are summarized by the flowsheet shown in

Figure 8.5. The goal of the considered system is the (decentralized) production

172 Chapter Eight

J K L

7 9 10

F G

H

4 5

C D

1 2

A B

of silage on farms. By starting from silage, two crucial steps are combined:

storage (thus enabling the downstream processes to operate continuously) and

conversion of the carbohydrates in green biomass to lactic acid. In addition,

silage production transforms many proteins into amino acids or peptides

(Povoden, 2002; Koschuh et al., 2004).

materials, intermediates, and possible products. Note that transport

is treated like a processing step: it uses trucks (or tractors), together

with the raw materials (or partially processed juice or press cake) and

available time, in order to derive a realistic logistics pattern.

Consequently, there has to be a plant-specific intermediate material

flow that leaves this process step. The steps listed in Table 8.2 reflect

the necessary logistical handling (i.e., the local and central

converters for various materials) involved in the process.

Once the cost function (including investment and operating

costs) is defined, the synthesis yields the optimal solution to process

silage as a part of the maximal structure; see Figure 8.6 (black lines

show the optimal solution, while the other options are grayed).

One major advantage of process synthesis is that it allows the

designer to apply sensitivity analysis not only to the process itself but

also to the entire value chain. In this example, sensitivity analysis

reveals a remarkable stability of the central biorefinery structure

Combined Process Integration and Optimization 173

Double Pressing

5289 t/y Water added

for second pressing

Silage Presscake

Feed Matter 35511 t/y Feed Matter 13825 t/y

Dry Matter 10000 t/y Dry Matter 6636 t/y

Dry Matter 28.16 % Dry Matter 48 %

Lactic Acid 1000 t/y Lactic Acid 215 t/y

Raw Fibre 2825.7 t/y Raw Fibre 2824 t/y

Org. Dry Matter 8992 t/y Org. Dry Matter 6483 t/y

Flux-out 1.9 t FM/h

Silage Juice

Feed Matter 26976 t/y

Dry Matter 3364 t/y

Dry Matter 13 %

Lactic Acid 785 t/y

Org. Dry Matter 2509 t/y

Flux-out 3.6 t FM/h

183442 603924 concentration: 6451609 kWh/y

kWh/y kWh/y 167033 kWh/y

Amino Acids

Conc. 80 %

Feed Matter 755 t/y Fibres

Feed Matter Product 884 t/y

Dry Matter 679 t/y Feed Matter 7373 t/y

Dry Matter 707 t/y

Dry Matter 90 % Dry Matter 6636 t/y

Org. Dry Matter 706 t/y

Lactic Acid 3.08 t/y Dry Matter 90 %

Pure Lactic Acid 706 t/y

Crude Prot. 666 t/y Lactic Acid 215 t/y

Flux-out 0.12 t/h

Flux-out 0.1 t FM/h Raw Fibre 2824 t/y

Crude Protein 98 % DM Org. Dry Matter 6483 t/y

Flux-out 1 t/h

concentrations (DM = Dry Matter; FM = Feed Matter).

for silage fractionation and biogas plants within the value chain,

including prices of key products such as electricity and press cake.

Further details on the method and the case study are provided in

(Halasz, Povoden, and Narodoslawsky, 2005).

Azeotropic distillation is common in chemical and allied industries.

Many of the existing distillation processes were designed and

developed through extensive trial-and-error efforts. The thermo-

dynamic pinches or boundaries (e.g., azeotropes), distillation

boundaries, and boundaries of liquidliquid equilibrium envelopes

are of critical importance for azeotropic distillation. Moreover, the

compositions of the feed and product streams must be specified in

order to define the synthesis problem for an azeotropic distillation

174 Chapter Eight

Mobile press MP Central converter, juice CCJ

Central press CP Local converter, rest of LCRF

fibers

Local fibers LF Central converter, rest CCRF

production of fibers

Central fibers CF Central converter, rest CCRJ

production of juice

Green biorefinery GBR Local transport, silage LTrS

Local biogas LBG Central transport, CTrS

silage

Central biogas CBG Local transport, cake LTrC

Local converter, LCS Central transport, cake CTrC

silage

Central converter, CCS Local transport, juice LTrJ

silage

Local converter, cake LCC Central transport, juice CTrJ

Central converter, CCC Central transport, rest CTrRF

cake of fibers

Local converter, juice LCJ

TABLE 8.2 List of Process Steps Incorporated into the Synthesis of the Base

Case

(RCM). The RCM of an ethanol-water-toluene system is shown in

Figure 8.7. The points E, W, and T represent the pure components

ethanol (product), water (by-product), and the entrainer, respectively,

while points F and H denote the feed and the ternary azeotrope,

respectively. The whole RCM is partitioned into materials

corresponding to these points; then the lines L1, . . . , L13 demarcate

the areas A1, . . . , A6.

A set of operating units for this process can be represented in a

P-graph; see Figure 8.8. The maximal structure and each solution

structure can be generated by algorithms MSG and SSG, respectively.

The Mathematical Programming model of a process network includes

constraints on operating units (i.e., mathematical models of those

units) as well as constraints on materials (e.g., mass balance

constraints). Mathematical models of mixers, separators, and decanters

are linear, including their mass balances, because they are based on

component flow rates. Thus, the Mathematical Programming model

of process networks involving mixers, separators, and decanters

gives rise to an LPR problem. Each solution structure satisfying

Combined Process Integration and Optimization 175

S

MP CTrS

J C SC

CTrC CP

LTrS LTrJ LTrC CTrJ

CL

LF CC

JC

RFL GBR

SL JL

F1 CTrRF CF

RJC

RFC LA AA

LCJ LCC LCRF F2

LCS CCRF CCC CCJ CCRJ CCS

GODML GODMC

CBG

LBG

E1 E2

curve map (RCM) of

the ethanol-water- L11 L1 A=Area

toluene system. L=Line

A1

L10

L13

L12 A2 L2

A6

A5 L6

F H

L8

L5

L9 A4 A3 L3

W L7 L4 T

176 Chapter Eight

representing the F

structure of an L12

azeotropic

distillation system.

Mixer

L6

Separator

W L13

L7

Mixer

L1

Separator

E L11

Decantor

evaluated by linear programming (Feng et al., 2003).

8.4.1 Simple Heat Integration

A methodology for combining PNS and HENs for integrated

synthesis has been presented by Nagy et al. (2001). It employs the

hP-graph, a special graph that incorporates both operating units and

heat exchangers. Figure 8.9 shows an example flowsheet and its

hP-graph representation.

Combined Process Integration and Optimization 177

M9

M10 M8 6

7 5

M11 M2 M6(1) M6(6) M10 M11

M4 2

M7 7

M5 M7

M1 M8

4 M6(4)

363 3 4 5

M9 328 343

6 M6 M3 1 M3(3) M3(4) M4

363

353 M3(1)

3

M5 333

1 2

Heating:

M1 M2

Cooling:

exchanging system.

solid lower half and that for a cooler by a bar with a solid upper half.

If the content of an operating unit is heated or cooled by latent heat,

then its mode is extended to the left by an appropriate heat

exchanger. For example, the flowsheet in Figure 8.9 shows that

operating unit 3 is heated with hot latent heat. Suppose that part of a

material stream fed to an operating unit requires temperature

modulation but that the remaining part, which feeds another

operating unit, does not. In this case, the former is diverted through

a heat exchanger and the latter is directed to another operating unit

(as with the two streams of M6 in the figure).

The ABB algorithm determines whether any of the operating

units should be included in the optimal structure. Applications of

the PNS algorithms extended to the synthesis of combined process

and HENs amply demonstrate the efficacy of the P-graph

framework.

A novel holistic method based on the P-graph approach has been

proposed by Halasz, Povoden, and Narodoslawsky (2002) and Liu

et al. (2006) for process retrofitting. Unlike conventional approaches,

the proposed methodology resynthesizes the entire process by

considering the enhanced performance of the operating units. As

such, this approach can account for all possible outcomes, including

the networks inevitable restructuring. The method employs a

P-graph implementation originally devised for synthesizing

grassroots processes. With the combinatorial feasibility of most

operating units largely predetermined, the approach detects any

178 Chapter Eight

retrofit of a conventional downstream process for the biochemical

production of butanol is accomplished by incorporating newly

identified adsorbing units, whose characteristics are summarized in

Table 8.3. Note that each of the processing units 24 and 25 are

represented by two operating units in the P-graphthat is, 241,

242 and 251, 252, respectively.

The P-graph of the problems maximal structure is presented in

Figure 8.10. It includes a gas stripper, an extractor, 27 simple

distillation columns, two azeotropic distillation units, a centrifuge,

and four adsorption columns.

The objective function is to minimize the operating cost (in terms

of its present values) for all operating units in the process flowsheet.

The cost for the three operating units is estimated heuristically. The

optimal flowsheet is illustrated in Figure 8.11.

This approach to retrofitting accounts for the fact that any change

implemented in downstream processing systemsespecially at an

upper segmentwill tend to propagate throughout the system as a

function of each systems unique sequence of structural features. The

approach is also applicable to many other chemical processes that

share the same structural features. With this methodology, a set of

optimal and near-optimal retrofit flowsheets can be generated and

ranked in terms of their costs. One (or more) of the options in this set

may be seen as infeasible or unsustainable when the retrofitted

system is further assessed by taking into account additional criteria

and constraints, including stability and controllability as well as

environmental, societal, and regulatory constraints.

Annual cost

Operating units and subunits Cost [103 $]

[103 $/y]

No. of No. of Equipment Type Capital Annualized Operating Total

units sub- desig- capital

units nation

23 C1 Centrifuge 9,240 3,080 1,168 4,248

24 241 B1 Adsorption 25,107 8,369 871 9,248

column

24 242 B2 Adsorption

column

25 251 B3 Adsorption 3,806 1,269 132 1,401

column

25 252 B4 Adsorption

column

P-graph Method (after Liu et al., 2006)

Combined Process Integration and Optimization 179

S00

C1 E1 G1

S05

S52 S51 S11 S16 S03

B1 B2 S1 B3 B4 D2 D1

S01

S08

S43 S46 S43

D5 D7 D9 D11 D13 D14 D21 A2 D3 A1

D15 S34 D17 S36 D19 D25 D27

S15 S17 S31 S32 S33 S35 S37 S44 S48 S49

S38 S39 S40 S45 S02

D20 D28 D29

D6 D8 D10 D12 D16 D18 D22 D26

S09

S19

S20

butanol, ethanol, and acetone with the inclusion of adsorption.

S09 S20

[A 7] [E 2]

S08 S39 S00

D21 D22

A7

E2 3(G1)

W 26 S40 S19

[B 26] S05 S01

A 11

E3 B3 B4

W 35 25-1(B3) 25-2(B4)

B 26

S08 S54

S05

S54 20-1(D21)

S00 G1

A4

S01 E1 S09 S40

W 35 S39

A 11 Legend:

E4 20-2(D22)

W 1773 E1 Acetone: A kg/h

B 27 W 1738 S20 S19

Butanol: B kg/h

B1 Ethanol: E kg/h

Water: W kg/h

Profit, and Security

8.5.1 Maximizing Throughput and Revenue

Majozi and Friedler (2006) developed an effective search algorithm

for determining the globally optimal throughput, revenue, or profit

over a predefined time horizon in multipurpose batch plants on the

basis of the S-graph framework. To demonstrate its performance, a

180 Chapter Eight

next.

The case study is taken from a multinational pharmaceuticals facility that

produces lotions, shampoos, conditioners, and various creams. The problem

features nonintermediate storage policy (NIS), and the processes involve

mixing and packaging. Mixing occurs in four mixing vessels (V1, V2, V3, and

V4), and packaging occurs in three packing lines (P1, P2, and P3). Because the

stirrers in mixing vessels are of different designs, mixing times vary according

to the vessel used. Table 8.4 shows the duration of mixing for each product

in each vessel, which have a capacity of about 3 t each. The table also lists

the economic contribution made to the company revenue or profit by selling

a unit of each product; shampoos have the highest economic contribution.

The packing duration for each product is 12 h, regardless of which packing

line is employed. The objective in this case study is to maximize the overall

economic result for a 24-h period. The S-graph for the recipe for the products

manufactured in this facility is given in Figure 8.12, where the sets of candidate

equipment units for performing tasks 1, . . . , 15 are defined by sets U1 = {V1, V2,

V4}, U2 = {P1, P2, P3}, U3 = {V1, V2, V3}, U4 = {V3}, U5 = {V2, V3}, and U6 = {V1,

V2, V4}.

The global optimal solution corresponds to two batches of Cream 2 and

one batch of Shampoo, which yields revenue of 9.5 cost units. The schedule

corresponding to the global optimum is shown in Figure 8.13.

Two advantages that this approach has over its Mathematical Programming

counterparts are: (1) it guarantees global optimality and (2) no manipulation

of the time horizon is requiredin particular, it is unnecessary to presuppose

time points that will discretize the time horizon into equal (or unequal) time

intervals. For this reason, the technique qualifies as a true continuous-time

methodology.

Many algorithmic and heuristic methods have been developed for

solving Heat Integration problems in continuous processes: Pinch

Technology (Linnhoff et al., 1982), superstructure-based mixed

contribution [cost

unit/batch] V1 V2 V3 V4

Cream 1 2 10 5 N/A 5

Cream 2 3 12 10 7 N/A

Conditioner 1 N/A N/A 12 N/A

Shampoo 3.5 N/A 8 13 N/A

Lotion 1.5 10 6 N/A 9

TABLE 8.4 Scheduling Data for the Case Study of Example 8.3

Combined Process Integration and Optimization 181

FIGURE 8.12 1 2

11 Cream 1

S-graph U1 5 U2 1

representing the

recipe for the case 3 4

study of U3 7 U2 1

12 Cream 2

Example 8.3.

5 6

13 Conditioner

U4 1 U2 1

7 8

14 Shampoo

U5 8 U2 1

9 10

15 Lotion

U6 6 U2 1

P3 Task 8 (Batch 1)

P2 Task 4 (Batch 2)

Equipment Units

P1 Task 4 (Batch 1)

V4

V3 Task 3 (Batch 1)

V2 Task 7 (Batch 1)

V1 Task 3 (Batch 2)

7 8 12 19

Time [h]

FIGURE 8.13 Globally optimal schedule for the case study of Example 8.3.

Westerberg, 1997), and integration with PNS (Nagy et al., 2001). In

contrast, scheduling and Heat Integration of batch processes are

both complex optimization problems and quite different in nature.

The problems of scheduling and Heat Integration could be solved

sequentially (in either order). Yet because the solution of one problem

will affect the other, the result of this simplistic, sequential approach

is usually poor. A better solution may result from an integrated

consideration of scheduling and Heat Integration. Since there were

not too many methods for solving this integrated problem, the

effective design and operation of integrated batch systems required

development of a new method (Adonyi et al., 2003). The goal was to

operate simultaneously those tasks that involve potential heat

182 Chapter Eight

solution.

The recipes of products A, B, and C are shown in Table 8.5, and the corresponding

S-graph of the recipe is given in Figure 8.14. The parameters of the heat streams

are listed in Table 8.6. The available hot and cold utilities are 200C and 10C.

The heat transfer coefficient between a hot and a cold stream is 1500 W/m2C,

and the area of a heat exchanger unit is 3 m2.

Task Eq. Time [h] Eq. Time [h] Eq. Time [h]

1 E1 5 E1 5 E1 6

E2 5 E2 5 E2 6

2 E3 4 E5 4 E5 3

E4 4 E6 4 E6 3

3 E5 4 E7 4 E7 4

E6 4 E8 4 E8 4

4 E1 5

E2 5

1 2 3

31 A

U1 5 U2 4 U3 4

4 5 6

32 A

U4 5 U5 4 U6 4

7 8 9

33 A

U7 5 U8 4 U9 4

10 11 12

34 A

U10 5 U11 4 U12 4

13 14 15 16 35

B

U13 5 U14 4 U15 4 U16 5 U2

17 18 19 20 36

B

U17 5 U18 4 U19 4 U20 5 U2

21 22 23 24 37

B

U21 5 U22 4 U23 4 U24 5 U2

25 26 27

38 C

U25 6 U26 3 U27 4

28 29 30

39 C

U28 6 U29 3 U30 4

FIGURE 8.14 S-graph of the recipe for the products in Example 8.4.

Combined Process Integration and Optimization 183

[C] [C] [MJ] Task

c1 Cold 40 120 400 A/2

h1 Hot 140 50 200 B/3

c2 Cold 80 130 100 B/4

h2 Hot 150 40 300 C/2

TABLE 8.6 Parameter Values for the Heating and Cooling Requirements in

Example 8.4

E1 1 4 13 21 10 20

E2 25 28 17 7 16 24

Equipment Units

E3 2 5 8

E4 11

E5 3 14 6 9

E6 26 29 18 22 12

E7 19

E8 27 30 15 23

10 20 30 36

Time [h]

Product A Product B Product C

FIGURE 8.15 Gantt chart of the optimal solution for Example 8.4.

If the heating and cooling duties are satisfied by utilities, then the minimal

makespan is 33.1 h with 3100 MJ utility. Extending the upper bound for the

makespan to 36 h reduces the required utility to 1100 MJ. Figure 8.15 displays

the Gantt chart of the optimal solution.

The task of designing a complete energy system involves significant

combinatorial complexity. For this, integer programming procedures

are not efficient. The P-graph framework and its associated algorithms

are capable of efficiently handling exactly the type of complexity that

is inherent to network optimization, and they appear to be some of

the best tools for solving this task. The P-graph approach can readily

evaluate technologies in their early stages of development, such as

fuel-cell combined cycles (FCCCs) based on molten carbonate and

solid oxide fuel cells (Varbanov and Friedler, 2008).

184 Chapter Eight

This example presents a procedure for evaluating energy conversion systems

involving FCCC subsystems that use biomass and/or fossil fuels; see Figure 8.16.

The procedure provides a tool for evaluating trends in CO2 emission levels

and the economics of such systems. The significant combinatorial complexity

involved is efficiently handled by P-graph algorithms. Promising system

components are evaluated by a methodology for synthesizing cost-optimal

FCCC configurations that account for the carbon footprint of the various

technology and fuel options using the P-graph framework.

The efficiency of an FCCC system varies with the fuel-cell (FC) operating

temperature, the type of bottoming cycle, and the degree of cycle integration

(see Varbanov et al., 2007). High-temperature fuel cells can be combined with

gas turbines, steam turbines, or both; however, combining all three yields only

marginal improvements. The main reason is that energy in the FC exhaust can

only be shared by the bottoming cycles, and typically this potential for energy

generation is most fully utilized by a steam or gas turbine alone. Hence the

involvement of more than one bottoming cycle cannot substantially increase

overall efficiency, although it can present capital cost trade-offs. Figure 8.17

shows an FCCC system represented by a conventional block-style diagram and

a P-graph fragment.

The synthesis of a processing network, such as the energy conversion system

considered here, requires that the designer choose the best solution from a

Fossil fuels

Processing: Energy Conversion:

Gasification or FCCCs

Digestion Boilers Heat

...

F

F

CO2

FCCC FCCC

W Q W Q CO2

Block-style flowsheet P-graph

Legend

F: Fuel; FCCC: Fuel Cell Combined Cycle unit; Q: Heat; W: Power

Combined Process Integration and Optimization 185

number of options. This optimization task may have several different objectives.

The most obvious are maximizing the system profit (minimizing its cost) and

minimizing the amount of CO2 emissions. Although it is mathematically

possible to define a multiobjective criterion to be optimized, using profitability

alone seems most coherent with the logic of a market economy because profit

drives the behavior of companies. Therefore, in this discussion the system profit

is used as the objective (to be maximized); CO2 emissions are then used as an

additional criterion during the analysis stage.

The materials and streams for the system considered in this example are

listed in Table 8.7. The waste products are assigned negative prices, denoting

that they generate costs for the system rather than revenue. Other performance

classification

BM Biomass Raw material Agricultural residues Varies

BG Clean biofuel Intermediate Biogas suitable for

utilization as a fuel

BR Waste / Product / Output Biomass residues 10 /t

By-product (solid remainder from

the biomass after

gasification)

CO2 Waste, Product / Output CO2 emissions Varies

greenhouse gas

FRT Useful Product / Output Fertilizer obtained as 50 /t

by-product a by-product from the

anaerobic digester

NG Fossil fuel Raw material Natural gas 36.8 /

MWh

PR Waste / Product / Output Particulates left from 10 /t

By-product cleaning the synthesis

gas

Q40 Steam Intermediate Steam at P = 40 bar(a)

Q5 Steam Product / Output Steam at P = 5 30 /

bar(a) to satisfy user MWh

demands

RSG Intermediate Intermediate Raw synthesis gas

fuel

SG Clean biofuel Intermediate Clean synthesis gas

suitable for use as a

fuel

W Power product Product / Output Electrical power to 100 /

satisfy user demands MWh

186 Chapter Eight

and economic data specifications, which provide the basis for appropriate

economic evaluation of the designs, are given in Varbanov and Friedler (2008).

The P-graph tools can be used to generate different solutions based on the

objectives of interest (operating cost, CO2 emissions) and market conditions.

The results show that systems of this type that employ renewable fuels are

economically viable for a wide range of economic conditions; this finding is due

mainly to the high energy efficiency of the FC-based systems. Figure 8.18 shows

the P-graph for the base case in which operating cost is minimized.

The significance of waste management systems has increased in

recent years because of the growing problems of waste management

chains affecting not only the environment, but also the daily lives of

millions of people. Several promising approaches have appeared,

including the RAMS (reliability, availability, maintenance, and

safety) software for modeling waste management systems. This

approach was analyzed and evaluated thoroughly by Sikos and

Kleme (2009a, 2009b, 2010).

In todays technological world, most waste management plants

depend on the continuous operation of a wide array of complex

machinery and equipment to sustain development, safety, human

health, and economic welfare. Plant operators expect a wide variety

BM

55.2 MW

BGD

4.2 t/h

1.4 t/h

32.0

CO2 BG MW FRT

FCCC_36 BLR_BG

(MCFC+ST) 12.8 MW

2.2 MW

Q40

15.0 MW

LD_40_5

W 10.0 MW Q5 15.0 MW

FIGURE 8.18 P-graph solution for the energy system of Example 8.5.

Combined Process Integration and Optimization 187

of appliances to function without unexpected problems or major

breakdowns. If these equipment units fail, the consequences can be

catastrophic: contamination, smog, acid rain, injury, loss of life,

production cutbacks, amassed garbage heaps, energy losses, and so

on. Catastrophic failures would also entail substantial added costs.

For these reasons, solid waste management is a matter of serious

concern, which in some cases (waste collection in Naples, Italy) has

even led to a change in government. The models that have been

developed to manage waste-producing processes are of two types:

optimization models deal with specific aspects of waste-related

problems; in contrast, integrated waste management models focus on

sustainability. The latter type can be subdivided into three main

subcategories: models based on costbenefit analysis, models based

on life-cycle inventory, and multicriteria models (Morrissey and

Browne, 2004).

However, there is an element of uncertainty or risk associated

with most environmental decisions. Multicriteria techniques can

be extended to consider reliability issues along the entire waste

management chain and need not be limited to comparing the environ-

mental impacts of different waste treatment methods. As the

complexity of unit arrangements increases, risk assessment becomes

more complicated. Risk is a measure of the plants ability to carry

out its specific operating mission reliably. The expected return on

related investments is a function of the plant equipments capacity,

which is defined in terms of reliability, availability, durability, and

performance.

Reliability engineering in waste management addresses all

aspects of the waste life cycle, from its collection and treatment

processes and through the energy generation lifetime, including

maintenance support and availability. The concepts of reliability,

maintainability, and availability can be quantified with the aid of

reliability engineering principles and life data analysis (Kececioglu,

2002). A significant fraction of any systems operating cost is due to

unplanned system stoppages for unscheduled repair of components

or the entire system. One method of mitigating the cost (and impact)

of such failures is to improve the systems reliability and availability.

Of course, improvements in reliability that are made by the supplier

early in the equipments life cycle may well result in additional

development cost being passed on to the customer in the form of

higher equipment acquisition cost. However, this cost increase can be

more than offset by the operational cost reduction associated with

improved reliability and increased uptime, which also improve

productivity. Note that, in the context of waste management,

reliability, availability, and maintenance have specialized meanings.

Reliability is the probability that a system will perform satisfactorily

for at least a given period of time t when used under stated conditions

(Kuo and Zuo, 2003).

188 Chapter Eight

system in a determined period of time. It can be calculated as the

ratio between an equipments lifetime and total time between failures

(de Castro and Cavalca, 2006):

Lifetime Lifetime MTBF

A (8.1)

Total time Lifetime Repair time MTBF MTTR

where MTBF is the Mean Time Between Failure (the inverse of the

failure rate) and MTTR is the Mean Time To Repair (the inverse of the

repair rate).

In addition, there are three frequently used terms defined

by Ireson, Coombs, and Moss (1996) and elsewhere: inherent,

achieved, and operational availability. The expression for inherent

availability is

MTBF

Ai (8.2)

MTBF MTTR

MTBM

Aa (8.3)

MTBM MAMT

MTBM

Ao (8.4)

MTBM MDT

Further specifications also existfor example, those given by

Hosford (1960):

able to operate within tolerances at a given instant of time.

Interval availability is the expected fraction of a given interval

of time that a system will be able to operate within

tolerances.

A special type of interval availability, called limiting interval

availability, was defined by Barlow and Proschan (1996); it is

the expected fraction of time in the long run that a system

operates satisfactorily.

There are even more specific definitions. For example, the availability

of a redundant system represented by a series of parallel systems is

formulated by de Castro and Cavalca (2006) as

n

As [1 (1 A )

i 1

i

yi

] (8.5)

Combined Process Integration and Optimization 189

where Ai is the availability of the components of subsystem i and yi is

the number of redundant components in subsystem i. Comparing

downtimes is another, intuitive way to express availability.

Maintenance covers those activities undertaken after a system is

in the field in order to keep it operational or restore it to operational

condition after a failure has occurred (Ireson, Coombs, and Moss,

1996). There are several classifications of maintenance, the most

important of which are listed as follows:

repaired each time it breaks down (Mechefske and Wang,

2003).

Condition-based maintenance (CBM): The critical components

are monitored for deterioration, and maintenance is carried

out just before the failure occurs (Mechefske and Wang,

2003).

Preventive (scheduled) maintenance: The plant is stopped at

intervals, often annually, and is partly stripped and inspected

for faults (Mechefske and Wang, 2003).

Reliability-centered maintenance (RCM): A procedure to identify

preventive maintenance requirements of complex systems

(Cheng et al., 2008).

(or restored to) a specified condition when maintenance is performed

by personnel having specified skill levels, using prescribed

procedures and resources, at each prescribed level of maintenance

and repair (Ireson, Coombs, and Moss, 1996). De Castro and Cavalca

(2006) defined it as the ability to renew a system or component in a

determined period of time, enabling it to continue performing its

designed functions.

For further information on reliability, availability, and mainten-

ance of waste management systems, see Sikos and Kleme (2010a).

Another difficulty with large systems is that troubleshooting

usually requires several problems to be solved, often simultaneously.

Data collection is also difficult because of the variety of input data;

the characteristics (e.g., type) of waste; production that changes

seasonally, weekly, and also randomly or unpredictably (due, e.g., to

weather changes, price changes that lead to different consumption

priorities, unpredictable natural disasters such as volcanic eruptions);

and changes related to a special venue, where a gathering or migration

of a mass of people can cause substantial changes as at a football

match or rock concert. The associated danger and hazards mean that

waste materials have to be appropriately cared for. A wide variety of

possible failure causes have to be identified. There are many other

issues to consider, too. Waste management systems are quite complex,

190 Chapter Eight

equipment units, each with its own degree of availability and

reliability. Scheduled outages need to be differentiated from the

downtimes caused by unexpected faults. Maintenance actions

including scheduled outagesshould be updated as needed to cope

with changes that occur within the system. The recommendations of

reliability engineers inform the designers task of effectively modeling

and optimizing all the specified objectives.

Many things need to be improved in waste management. These

include policy focus and popular opinion, which should both pay

more attention to the importance of this process and its effects.

Landfill space is decreasing while solid waste is increasing, so public

and private landfills compete for municipal clients to ensure the

capital required to extend landfill life while coping with new permits.

The situation varies in different countries and geographic regions, so

investigation is needed to devise appropriate solutions for specific

scenarios.

8.6 Summary

This chapter presented several examples of combined PI and

optimization. The main focus was on exploiting the advantages of

graph-theoretic (P-graph and/or S-graph) frameworks. These

methods are well tested and have demonstrated their efficiency

across many applications.

CHAPTER 9

Software Tools

Process Integration, modeling, and optimization problems in

chemical engineering are complex in terms of scale and relationships.

Solving these problems requires the application of information

technology and computer software, which provide fast and as much

as possible accurate solutions via a user-friendly interface. Software

tools have been widely used for process simulation, integration, and

optimization, and this has helped process industry companies

achieve their operational goals. There is a large number of efficient

tools available, each with its particular advantages. The aim of this

chapter is to describe these tools based on comprehensive experience

with them and their application to Process Integration, modeling,

and optimization.

The online encyclopedia Wikipedia (2009) presents a comprehensive

list of available software tools for the simulation of material and energy

balances of chemical processing plants. Examples include: (1) ASCEND;

(2) Aspen HYSYS by Aspen Technology; (3) ASSETT and D-SPICE

by Kongsberg Process Simulation; (4) CHEMCAD; (5) COCO simulator;

(6) Design II by Winsim; (7) EcosimPro; (8) Environment for Modeling,

Simulation and Optimization (EMSO); (9) Dymola; (10) GIBBSim;

(11) gPROMS by PSE Ltd; (12) OLGA by SPT Group (Scandpower);

(13) Omega by Yokogawa; (14) OpenModelica; (15) Petro-SIM; (16) ProMax;

(17) SimSci-Esscor DYNSIM & PRO/II by Invensys; (18) SysCAD;

(19) UniSim Design & Shadow Plant by Honeywell; and (20) VMGSim.

However, this selection cannot be fully comprehensive and is limited

just to process simulation software. It should cover a wider field and be

studied in more detail, which is the target of this chapter.

9.2.1 Process Network Synthesis Solutions

Process network synthesis (PNS) Solutions is a software package

designed to solve problems in PNS. Process synthesis involves

191

192 Chapter Nine

determining optimal types, configurations, and capacities of the

units that perform various operations within the system. The details

of the process graph (P-graph) approach were presented in Chapters 7

and 8. Process network synthesis is sometimes called flowsheeting

or flowsheet optimization because it involves the creation of a

flowsheet for the industrial process under consideration.

In order to solve a PNS problem, the designer must examine all

feasible structures and select the best among them. The structures

optimality can be assessed in terms of cost, profit, efficiency, and so

on. When designing an optimal process network, both structural

information (which processing units are connected, and how) and

sizing information (how much is produced from a given material)

are needed.

The questions addressed by PNS Solutions are as follows:

(1) How are the building blocks of a process network best represented?

(2) What are the possible solution structures of the problem? (3) What

is the maximal structure (which includes all solution structures)?

(4) What is the optimal structure?

The maximal structure comprises all the combinatorially feasible

structures capable of yielding the specified products from the

specified raw materials. Certainly, the optimal network or structure

is among these feasible structures. The maximal structure generation

(MSG) algorithm produces a P-graph (see Figure 9.1) in which each

Sof tware Tools 193

material and operating unit appears exactly once. In the composition

phase, the nodes are linked stepwise, layer by layer, starting from the

shallowest end (i.e., the final-product end) of the remaining input

structure. The algorithm proceeds by assessing whether any of the

linked nodes violates any of the axioms described in Chapter 7.

The structure generation is performed transparently, and the

maximum structure that results is the input for the Solution Structures

Generation (SSG) algorithm. If a material has to be produced, then

the SSG algorithm generates all possible ways for its production. For

example, if M1 can be produced by O2 or by O3 then the possibilities

include production by O2 alone or by O3 alone or by using both.

Once an operating unit is included, its input materials must be

produced as well, and so forth. Materials are selected in a specific

order. The parentchild relation between steps ensures that the

materials are selected by the process according to this order.

The SSG algorithm yields all the combinatorially feasible solution

structures of a given problem. Unfortunately, the number of feasible

structures at this stage is often too large to be enumerated explicitly.

Therefore, the Accelerated Branch-and-Bound (ABB) algorithm is

used to determine the optimal structure without generating all the

possible solutions. Input to this algorithm includes not only the

structural relationships between materials and operating units but

also such additional information as the costs of each raw material,

the fixed and proportional costs of operating units, and the constraints

(if any) on the quantity of materials and the capacity of operating

units.

S-Graph Studio is a software package that enables the user to design

batch processes and to optimize them via various optimization

methods (S-Graph, 2009). The program also allows scheduling

problems to be defined using graphical tools. It has a modular

architecture, so different solvers can be used with the program.

S-Graph Studio uses the industrys standard file format: BatchML (as

defined by the ISA-88 standard of the World Batch Forum), which is

used to exchange information between industry sites and plants as

well as for other purposes. The Excel file format can also be used for

both input and output. The software includes a solver that utilizes

the S-graph methodology developed at the University of Pannonia

(S-Graph, 2009).

One of the main goals of batch process optimization is to

minimize makespanthat is, finding the shortest time in which a

process can be completed using available resources. S-Graph Studio

can be used to define batch processes in terms of the tasks to be

performed, the available equipment units, and task completion times

(as a function of the equipment units used). This information is

necessary and sufficient for minimizing the makespan of a process.

194 Chapter Nine

graph) but can be exported to various formats for further use. Issues

addressed by S-Graph Studio include how to represent a scheduling

problem and how to generate the optimal schedule structure in the

cases of unlimited and no intermediate storage. The schedule

associated with the selected solution is displayed as a Gantt chart

(Figure 9.2) or a schedule graph, both of which clearly show the unit

task assignments and their associated timing.

FIGURE 9.2 The Gantt chart and schedule graph of a solution (S-Graph Studio).

Sof tware Tools 195

9.3.1 SPRINT

SPRINT is the software package used to design energy systems for

individual processes on a site (SPRINT, 2009). This software tool

provides energy targets and optimizes the choice of utilities for an

individual process. It also performs Heat Exchanger Network (HEN)

design automatically for the utilities that are selected. Both new

design and retrofit can be carried out automatically, though the

designer maintains control over network complexity. For example,

several retrofit modifications can be generated automatically and

then presented to the designer one at a time, so that the number of

modifications is minimized, and the final decision on each

modification is left to the designer. SPRINT can also be used for

HEN operational optimization tasks. The SPRINT and the STAR

programs (see Section 9.3.3) are linked by common data structures,

which facilitate their interaction (e.g., the same files can be used, and

no manual data transfer is required between the programs).

SPRINT is used in both academic and industrial settings for the

following applications: (1) optimizing the choice and load of utilities

for individual processes; (2) automatic design of new HENs;

(3) automatic retrofit design of HENs while using a minimum number

of modifications; (4) automatic design for multiple utilities (new

design and retrofit); (5) interactive network design; (6) simulation of

networks via simple models; (7) targeting minimum energy

consumption; (8) network optimization; and (9) assessing network

operability. User interaction with the network structure is through

the graphical editor shown in Figure 9.3.

9.3.2 HEAT-int

HEAT-int (2009) is a product of Process Integration Ltd. This program

is used to improve the energy performance of individual processes

on a site. It is the next-generation development of SPRINT software

(by a related team of developers) to a commercial standard, and it

offers more user-friendly interface features. The areas in which

HEAT-int is applied are similar to those listed previously for

SPRINT.

9.3.3 STAR

STAR is a software package for the design of site utility and

cogeneration systems (STAR, 2009); see Figure 9.4 for an example

user interface. This program analyzes all the interactions between

site processes and the steam system, steam turbines, gas turbines

(with auxiliary firing options), the boiler house, local fired heaters,

and cooling systems. The analysis is used to reduce energy cost and

196 Chapter Nine

Sof tware Tools 197

to plan infrastructure investment when operational changes are

anticipated, changes that may include the replacement of energy

equipment. STAR can also be used to investigate flue gas emissions,

which should often be reduced to meet tighter environmental

regulations. The STAR package incorporates several tools, as

described next.

Utility system optimization: A given utility system configuration

incorporates important degrees of freedom for optimization.

Multiple boilers with different efficiencies and different fuels in

addition to multiple back-pressure steam turbines, condensing

turbines, gas turbine heat recovery steam generators, and letdown

valves provide optional heat flow paths that can all be exploited for

significant cost reduction. STAR has a utility system optimization

facility that allows existing utility systems to be optimized. It can

also be used to plan infrastructure de-bottlenecking and investment

strategies.

Top-level analysis: When studying an existing site, it is important

to understand how its infrastructure influences the degrees of

freedom to make changes as well as the economic consequences of

those changes. These considerations are addressed by STARs top-

level analysis, whose results ensure that the designer does not waste

time and costs pursuing changes that are not viable (structurally or

economically) in the overall site context.

Process energy targets: Even though the primary function of STAR

is the analysis of utility systems, it includes tools for setting energy

targets and selecting utilities for individual processes. Using these

tools allows the picture of the Total Site to be built up from the

individual processes within STAR.

Total Sites: STAR can produce profiles that represent the heating

and cooling requirements of the Total Site. This allows targets to be

set for fuel consumption in the boilers, cogeneration potential, and

energy costs. The Site Profiles can be based either on the full heat

recovery data or, more simply, on data for the utility exchangers

only.

Boiler systems and steam turbine systems: Using STAR allows the

designer to establish optimal targets for the amount of steam

generated by boilers and gas turbines (with auxiliary firing options).

A gas turbine model enables the study of different gas turbine

arrangements. Steam turbines are a part of most utility systems,

serving to generate power or as allocated drivers for process

machines. STAR software incorporates the design of steam turbine

networks and the analysis of their operability.

Emissions: By relating process energy requirements to the supply

of utilities, it is possible to target the amount of fuel required for the

utility system. Such targets can be combined with information on the

198 Chapter Nine

of CO2, SOx, NOx, and particulates. It is then possible to explore the

various options for reducing these emissions.

9.3.4 SITE-int

Like HEAT-int (see Section 9.3.2), SITE-int (2009) is a product of

Process Integration Ltd. SITE-int is a state-of-the-art software package

for the design, optimization, and integration of site utility systems in

process industries. Its main features include methods to: (1) model

and optimize site utility systems; (2) minimize operating costs for

existing systems without modification; (3) target cogeneration

potential; (4) optimize site steam pressures and loads; (5) minimize

site energy costs through system modifications; (6) determine the

true benefit from saving energy in the individual processes; (7) reduce

greenhouse gas emissions from the site; and (8) create partial-load

models of utility system components from plant operating data using

regression functions and data reconciliation functions.

9.3.5 WORK

WORK is the software package used for the design of low-temperature

(subambient) processes (WORK, 2009). Low-temperature processes

require heat rejection to refrigeration systems. As a result, the

operating costs for such processes are usually dominated by the cost

of power to run the refrigeration system. Complex refrigeration

systemsincluding cascade and mixed refrigerant systemscan be

analyzed using WORK. For mixed refrigerants, WORK can be used

to optimize refrigerant composition. The software enables the user

to: (1) understand complex refrigeration systems; (2) target minimum

shaft work for low-temperature cooling duties; (3) optimize the

number and temperatures of refrigeration levels; (4) target minimum

shaft work for cascade refrigeration systems; (5) target minimum

shaft work for mixed refrigerant systems; and (6) determine the

optimum composition for mixed refrigeration systems. Three of

these features are discussed in more detail next.

Targeting low-temperature systems: WORK can target minimum

shaft work for simple and complex refrigeration cycles. Targets are

based on rigorous thermodynamic calculations that are highly

accurate even when compared with the results of rigorous simulation.

When multiple refrigeration levels are used, trade-offs arise between

temperature levels and shaft loads. Adjustments to each temperature

level affects not only its own shaftwork requirement but also that of

the other levels. Therefore, all levels of refrigeration must be

optimized simultaneously. This task is facilitated by WORKs

extremely accurate shaftwork predictions.

Simulating refrigeration systems: WORK enables the simulation of

simple and complex refrigeration systems (see Figure 9.5), which may

involve multiple heat levels and multiple compressors. The refrigerant

Sof tware Tools 199

(WORK user interface).

background process in order to minimize overall shaftwork.

Optimizing mixed-refrigerant systems: WORK can optimize the

composition of mixed refrigerants to minimize shaftwork

requirements. This goal is achieved by matching the composition of

the mixed refrigerant to the cooling profile (see Figure 9.5). The

software outputs a visual representation of the shaftwork losses in

refrigeration cycles, including both mechanical and thermal losses.

9.3.6 HEXTRAN

HEXTRAN is a steady-state simulator that provides a view of heat

transfer systems (IPS, 2009a). It is used to design new systems,

monitor current systems, optimize existing operations, and solve (or

prevent) heat transfer problems. The program simulates integrated

processes and allows engineers to monitor the performance of

individual exchangers or an entire heat transfer network. It also

offers superior postprocessing displays, plotting Grand Composite

Curves as well as the results from network targeting and zone

analysis. HEXTRAN provides new efficiencies in all types of design

and operational analysis work, such as individual exchanger and

network designs, Pinch Analysis, exchanger zone analysis, split

flows, area payout, and optimal cleaning cycles.

HEXTRAN analyzes factors that can make the difference between

profits and losses. These factors include: (1) improved process heat

transfer, product yield, and quality; (2) increased energy efficiency

and significantly reduced operating costs; (3) increased plant flexibility

and throughput; (4) optimized cleaning schedule for exchangers;

(5) optimal antifouling selection and usage; and (6) improved process

designs and revamps. The HEXTRAN simulator for process heat

transfer offers features that facilitate straightforward evaluations of

complex design, operational, and retrofit situations; in particular, it:

(1) enables the design of both simple and complex heat transfer

systems that result in cost-effective, flexible processes; (2) allows the

200 Chapter Nine

optimum performance; and (3) identifies cleaning incentives and

predicts future performance.

9.3.7 SuperTarget

SuperTarget is mainly used to improve Heat Integration in new

design and retrofit projects by reducing operating costs and optimally

targeting capital investment (Linnhoff March, 2009). SuperTarget is

also a tool for day-to-day application by novice or occasional users,

and it makes Pinch Analysis a routine part of process design. The

software features an intuitive user interface that makes the technology

accessible to users at all levels of expertise, and advanced tools are

available for expert applications. Many of the most time-consuming

tasks traditionally associated with Pinch Analysis have been partially

or fully automated.

SuperTarget takes data directly from most popular process

simulation programs through interfaces to Aspen Plus, HYSYS, and

PRO/II. Its automatic data extraction system converts raw process

data into Pinch data, although the user has the option of overriding

the extraction defaults. SuperTarget consists of three program

modules: (1) Process is the core program, which is used to optimize

energy use within a single process unit; (2) Column performs a

thermal analysis of the heat distribution in distillation columns; and

(3) Site is used to establish heat and power targets across a Total Site.

Pinch Analysis provides a comprehensive and systematic approach to

maximizing the plant energy efficiency and minimizing the use of

utilities. The Pinch technique is amenable to use with commercial

spreadsheets, which display a grid of rows and columns made up of

multiple cells, each containing alphanumeric text or numeric values.

Kemp (2007) developed an Excel spreadsheet for Pinch Analysis that

incorporates targeting calculations and plots (see Figure 9.6). The main

components of this spreadsheet are: (1) input of stream data; (2) calcu-

lation of Composite Curves (CCs), the problem table, energy targets, and

the Pinch temperature; (3) plots of CCs and the GCC; (4) plots of the

stream population over the temperature range of the problem and the

basic grid diagram; and (5) tables and graphs of the variation in energy

and Pinch temperature over a range of Tmin values.

Neither area targeting nor cost targeting is included in the

spreadsheet because doing so would add considerable complexity.

Suitable data on heat exchanger coefficients is often lacking and most

plots of cost against Tmin could look fairly flat, which is not always

the case when the appropriate cost scale is set. However, topology

can still be identified from the graphs of utility use and Pinch

temperature against Tmin (Kemp, 2007).

Sof tware Tools 201

FIGURE 9.6 Spreadsheet user interface for Pinch Analysis (Kemp Pinch

Spreadsheet, 2006).

WATER is a software package for the design of water systems in

process industries (WATER, 2009). Water is used for a wide variety

of operations for mass transfer, washing operations, steam systems,

cooling systems, and so forth. WATER targets and designs for

minimum water consumption by identifying opportunities for

water regeneration, reuse, and recycling. The cost of effluent

treatment systems is minimized through design methods that lead

to distributed systems. Networks involving water use and effluent

treatment are designed automatically, with the designer in full

control of network complexity. This software tool is capable of

handling multiple contaminants. The six principal issues addressed

by WATER are described next.

Water minimization: WATER minimizes freshwater use by

identifying reuse opportunities. The program works using data on

the water quality constraints of each operation of the process. In

addition to constraints for multiple contaminants, constraints on

maximum and minimum flow rate and on water losses and gains

can be specified, as well as forbidden matches.

Multiple sources of freshwater: It is often the case that there are a

number of different sources of freshwater available, each featuring

different qualities and different costs. WATER is able to optimize the

use of multiple sources of freshwater.

202 Chapter Nine

maximum and minimum flow rates, forbidden matches, and water

losses or gains in individual operations can be accounted for. The

designer also has control over network complexity.

Regeneration of water: Once water reuse has been maximized,

further reduction in water consumption can only result from

regenerating wastewater. WATER enables a network designer to

examine and compare the effects of regeneration reuse and

regeneration recycling.

Automatic design of water reuse and effluent treatment networks:

WATER can automatically generate not only water reuse networks

but also effluent treatment networks at a minimum cost. The design

engineer maintains control over network complexity and over the

relevant constraints (e.g., forbidden matches, flow-rate ranges, water

losses and gains during treatment operations).

Pipe work and sewer costs in network design: In addition to the capital

cost associated with regeneration and treatment processes, WATER

also incorporates the cost of connecting the operations that involve

running new pipes and sewers. These factors are included with the

freshwater cost and other capital cost when assessing trade-offs

between cost and overall performance. It is important to consider

pipe work and sewer cost because they have a profound effect on

network structure and complexity.

9.5.1 ASPEN

Aspen Technology, Inc., provides integrated software applications

for a variety of industriesoil and gas, petroleum, chemicals, and

pharmaceuticalsthat manufacture products using chemical

processes. Aspen ONE (AspenTech, 2009d) is an application suite

that enables process manufacturers to implement best practices for

optimizing their engineering, manufacturing, and supply chain

operations. This software package addresses inefficiencies

throughout the plant, resulting in significant cost savings. Aspen

Plus (Aspen-Tech, 2009c) is a core element of the Aspen ONE process

engineering suite. It is a process modeling tool used in conceptual

design, optimization, and performance monitoring for the chemical,

polymer, specialty chemical, metals and minerals, and coal power

industries. Aspen Plus includes a large database of pure component

and phase equilibrium data for conventional chemicals, electrolytes,

solids, and polymers. This information is updated regularly using

data from the U.S. National Institute of Standards and Technology.

Aspen Plus is well integrated with Aspens software for cost analysis

(AspenTech, 2009a) and heat exchanger design (AspenTech, 2009b).

These software applications enable rigorous sizing and rating of

Sof tware Tools 203

key equipment, such as heat exchangers and distillation columns,

within the simulation environment.

The HYSYS software was initially created by Hyprotech for

simulating both steady-state and dynamic processes. It includes tools

that can be applied to: (1) estimating physical properties, including

liquidvapor phase equilibrium; (2) establishing heat and mass

balances; (3) designing and optimizing oil and gas processes; and

(4) evaluating and selecting process equipment. HYSYS technology

was acquired and modified by Aspen (see Section 9.5.1) and later by

Honeywell, where it is known as UniSim Design (Honeywell, 2010).

Aspen HYSYS and UniSim Design are similar in terms of application

and the working interface. Both include: (1) a library of physical

properties of many chemical substances; (2) a set of subroutines for

estimating the behavior of several types of plant equipment (heat

exchangers, reactors, etc.); and (3) a graphical user interface for

inputting case specifications and displaying results.

Once the designer describes the process in terms of equipment

units interconnected by process streams, the program solves all the

equations for mass, energy, and equilibrium while taking into

consideration the units specified design parameters. The program

is built upon proven technologies that for two decades have supplied

process simulation tools to the oil and gas industry. Another

advantage of Aspen HYSYS and UniSim Design is their interactive

and flexible process modeling, which allows engineers to design,

monitor, and troubleshoot as well as to make operational

improvements and perform asset management. Employing these

features leads to decision making that enhances the productivity,

reliability, and profitability of a processing plants life cycle

(Ebenezer, 2005). The HYSYS fluid package module requires

information on the characteristics of unit components and the

physical properties of relevant streams. Also, accurate simulation of

processes requires that appropriate thermodynamic models be

selected as a framework for the simulation. Note that a process that

is otherwise fully optimized in terms of equipment selection,

configuration, and operation is of no use whatsoever if the process

simulation was based on an incomplete or inaccurate fluid package

or on an inappropriate thermodynamic model.

HYSYS requires minimal input data from the user; the most

important input parameters are the streams temperature, pressure,

and flow rate (Ebenezer, 2005). The software includes an assortment

of utilities that can be attached to process stream and unit operations.

These tools interact with the process to provide valuable additional

information. For example, the flowsheet used within the HYSYS

simulation environment can be manipulated by the designer to

estimate desired output.

204 Chapter Nine

9.5.3 gPROMS

An advanced modeling environment for process industries is offered

by gPROMS (PSE, 2009), which provides advanced custom modeling

capabilities within a flowsheeting environment combined with an

object-oriented modeling language. The process modeling, process

simulation, and optimization capabilities of gPROMS are used to

generate accurate process behavior predictions and information for

decision support in product and process innovation, design, and

operation. Because gPROMS is a flowsheeting environment, the user

can optimize complex units within the context of an entire process.

The software employs synchronized graphical and text views, which

makes it easy to develop, maintain, and assure the quality of the

models, and archive them. It is capable of assessing all phases of the

process life cycle, from laboratory experimentation to process design

and detailed engineering to online operations.

For modeling problems and deriving solutions, the environment

provided by gPROMS ModelBuilder is employed. ModelBuilder is a

flexible environment in which engineering experts can perform custom

modeling, process engineers can generate graphical flowsheets, and

process operators can run execution-only routines. Figure 9.7 shows

an example user interface. In summary, gPROMS is an equation-

oriented modeling system used for building, validating, and executing

first-principles models within a flowsheeting framework.

text views (PSE, 2009).

Sof tware Tools 205

9.5.4 CHEMCAD

The CHEMCAD software tool for simulating chemical processes

includes libraries of chemical components, thermodynamic methods,

and unit operations (Chemstations, 2009). Its purpose is to facilitate

the simulation of steady-state continuous chemical processes from

laboratory-scale experiments to full-scale operations; see Figure 9.8

for the user interface. This software package has recently been

upgraded to allow for the dynamic analysis of flowsheets. It offers

operability assessment, proportional integral derivative (PID) loop

tuning, and operator training as well as online process control and

soft-sensor functionality. Models for nonstandard unit operations

can simulate the behavior of a process under varying feed rates,

product rates, temperatures, pressures, and compositions.

The program contains the 1500-component Design Institute for

Physical Property (DIPPR) database as well as a separate, user-

defined database. Components are selected by ID number, formula,

synonym, or class. The program also includes routines for predicting

the properties of components not included in the database. Crude oil

feeds may be characterized by American Society for Testing and

Materials (ASTM) or true boiling point (TBP) curves and then

represented in the simulation by a series of pseudocomponents (cuts)

with different boiling points.

The flowsheet representing the plant layout is input graphically,

and data describing each feed stream and unit operation is entered

via pull-down menu options. Automatic error checking is used to

preclude overspecification or missing data entries. The interactive

interface also permits individual unit operations to be run separately

206 Chapter Nine

analyses (Chempute Software, 2001).

Results output by the program include a full heat and material

balance; thermodynamic and physical properties of all streams;

component flow rates; stream temperatures, pressures, and vapor

fractions; and process equipment parameters. The user may view the

results graphically or obtain a printout of either summary or detailed

results. CHEMCAD includes additional modules; some of them are

(1) CHEMCAD-THERM for the design and rating of shell-and-tube

heat exchangers (including air coolers), (2) CHEMCAD-BATCH for

simulation of batch distillation processes, and (3) CHEMCAD-REACS

for the dynamic simulation of stirred tank reactors. The software

package also includes a subroutine, called CONVERT, which

translates the process flow diagram generated by the program into a

series of drawing exchange format (DXF) files for incorporation into

AutoCAD software routines.

9.5.5 PRO/II

The PRO/II computer simulation system is used by process engineers

in the chemical, petroleum, natural gas, solids processing, and

polymer industries (IPS, 2009b). It includes a large chemical component

library and multiple thermodynamic property prediction methods as

well as advanced and flexible techniques for evaluating unit

operations. The software tool can perform mass and energy balance

calculations for modeling steady-state processes. Expert systems,

extensive input processing, and error checking are included to help

inexperienced users. Typically, PRO/II simulation is applied to the

following tasks: (1) process design; (2) evaluating alternative plant

configurations; (3) modernizing and revamping existing plants;

(4) assessing, documenting, and complying with environmental

regulations; (5) troubleshooting and de-bottlenecking plant processes;

and (6) monitoring and improving plant yields and profitability.

9.6.1 GAMS

The General Algebraic Modeling System (GAMS, 2009) is a high-level

modeling system for Mathematical Programming and optimization.

GAMS is designed for modeling linear, nonlinear, and mixed integer

optimization problems. The system is well suited to complex, large-

scale modeling applications, and it allows the user to build and

archive large models that can later be adapted to new situations. A

particular advantage of GAMS is its ability to handle large, complex,

and/or unique designs that may require many revisions before an

accurate model is established.

Sof tware Tools 207

The system models problems in a natural and highly compact

way. The package includes an integrated development environment

and a group of integrated solvers. GAMS was the first algebraic

modeling language, and it is formally similar to several common

programming languages. Models are described in algebraic

statements that are easy for humans and machines to read. The

system is capable of handling models of many different types, so

switching between model types can be done with a minimum of

effort. For instance, the same data, variables, and equations can be

reused for a linear and a nonlinear model by simply converting a

small number of parameters to variables. GAMS also includes a

variety of solvers for different classes of models.

9.6.2 MIPSYN

The MIPSYN (short for Mixed Integer Process SYNthesizer) is a

user-friendly computer package for the integrated synthesis of new

plants and for the innovative reconstruction of existing plants at

different levels of complexity. The tasks to which it can be applied

range from simple nonlinear programming (NLP) solutions for

plant optimization problems to the mixed integer nonlinear

programming (MINLP) optimization of heat-integrated, flexible

plants. MIPSYN is the successor to the PROSYN synthesizer

(Kravanja and Grossmann, 1990; Kravanja and Grossmann, 1994).

As such, it is based on the most advanced modeling and optimization

techniques: those rooted in disjunctive MINLP. The MIPSYN

software can simultaneously address both discrete optimization

(e.g., selection of process units, their operating status, their

connectivity, and ranges of operation) and continuous optimization

(of temperatures, flows, pressures, etc.). The package integrates the

following methods and related components: (1) GAMS with a variety

of different NLP and MILP solvers; (2) different versions of the outer

approximation (OA) algorithmincluding the modified OA/ER

algorithm (ER denotes equality relaxation) and a new logic-

based OA/ER algorithmwhich are supervised by MIPSYN

command files; (3) a simple simulator that serves as an initializer to

provide NLP subproblems with feasible (or nearly feasible) starting

points; (4) a library of models pertaining to process units,

interconnected nodes, and the simultaneous integration of heat and

mass; (5) a database of the physical properties of the most common

chemical components; and (6) a hybrid modeling environment, with

a link to external FORTRAN routines, for solving the implicit part of

synthesis models.

Execution of the NLP and MILP steps in OA algorithms is

performed through the use of GAMS saving and restart capabilities,

which enable the user to execute MIPSYN in automated or interactive

modes of operation. The synthesizer features many important

208 Chapter Nine

and MILP solvers in a sequence with different option files (text files

containing specifications of solver options to be applied); the efficient

modeling of different formulations and strategies (e.g., multilevel

MINLP); the capacity to solve feasibility problems whose objective

functions are augmented by penalties; multiobjective optimization;

integer-infeasible path optimization; multiperiod optimization; and

flexible synthesis for cases where the true parameters are uncertain.

Some of these applications were described in Kravanja (2009).

MIPSYN can be comprehended and used at different levels of

problem abstraction because it includes: (1) an MINLP solver for

problems of a general nature; (2) a process synthesizer for generating

process flowsheets; and (3) a synthesizer shell for accommodating

applications from different engineering domains.

A number of case studies have been performed using MIPSYN.

In these studies, the synthesis was applied to all basic process

systems and subsystems. Examples include: (1) heat-integrated

reactor networks in overall process schemes; (2) heat-integrated and

flexible separator networks; (3) Heat Exchanger Networks, including

retrofits and networks that use more than one exchanger type;

(4) mass exchanger networks; (5) heat-integrated overall process

schemes based on a sustainable, multiobjective approach; and

(6) flexible and heat-integrated flowsheets, together with their

HENs, for cases involving as many as 30 uncertain parameters.

Note that the MIPSYN synthesizer shell also enables applications

in the area of mechanics (Kravanja, Kravanja, and Bedenik, 1998a;

Kravanja, Kravanja, and Bedenik, 1998b; Kravanja, ilih, and

Kravanja, 2005). These mechanical applications range from simple

NLP optimizations to complex, multilevel MINLP syntheses of

structures in which topology, material use, and dimensions are

optimized simultaneously.

9.6.3 LINDO

LINDO is a tool for solving linear, integer, and quadratic program-

ming problems (Lindo Systems, 2009). It provides an interactive

modeling environment that facilitates the simulation and solution of

optimization problems. LINDO has the speed and capacity to solve

large-scale linear and integer models. The dynamic link library

(DLL) version of LINDO allows users to seamlessly integrate the

LINDO solver into Microsoft Windows applications that are written

in Visual Basic, C/C++, or any language that supports DLL calls.

Workstation users can exploit the linkable object libraries to hook

the solver engine to applications written in FORTRAN or C. The

latest LINDO version (ODC, 2009) offers a number of enhancements,

including: (1) significantly expanded nonlinear capabilities; (2) global

optimization tools; (3) improved performance on linear and integer

Sof tware Tools 209

problems; and (4) enhanced interfaces to other systems, such as

MATLAB and Java.

LINDO API was the first full-featured solver with a callable

library to offer general nonlinear and integer nonlinear capabilities.

This feature allows developers to incorporate a single, general-

purpose solver into their custom applications. The softwares linear

and integer capabilities provide the user with a comprehensive set

of routines for formulating, solving, and modifying nonlinear

models (although a separate, nonlinear license must be purchased to

access these nonlinear capabilities). The global solver combines

techniques for range bounding (e.g., interval analysis, convex

analysis) and range reduction (e.g., linear programming, constraint

propagation) within a branch-and-bound framework to find proven

global solutions to nonconvex nonlinear programs or mixed-integer

nonlinear programs.

Solvers (optimizers) are software tools that help users find the best

way to allocate scarce resources. Frontline Systems (Solver.com,

2009) offers products that cover several problem types and that

allow model definition via an Excel spreadsheet or via a program

written in any of several common programming languages and

environments (e.g., Visual Basic, C/C++/C#, VB.NET, Java, MATLAB).

Models designed using these solver products include decision

variables for quantities of resources as well as calculated results

(constraints) that are subject to limits (e.g., budget, capacity, and/or

time constraints).

The ILOG system is geared toward optimizing decisions and thereby

finding the best solutions to complex planning and scheduling

problems. It provides an application environment that supports the

flexible exploration of all the trade-offs and sensitivities. Thus, the

ILOG optimization decision manager (ODM) makes optimization

easier (ILOG, 2006); see Figure 9.9. A unique aspect of ODM-based

applications is that they provide all the features that business people

need to take full advantage of optimization technology. Applications

built using ILOG ODM allow users to create, visualize, and compare

planning or scheduling scenarios, to adjust any of the models inputs

or goals, and to comprehend the relevant binding constraints, trade-

offs, sensitivities, and business options.

With ODM-based applications, overconstrained problems are

automatically relaxed during runtime by ILOG CPLEX, which is

programmed to relax the least important (and fewest possible)

constraints. This ensures that a solution will always be found, and

solutions are presented along with clear information about any

210 Chapter Nine

FIGURE 9.9 User interface for an ILOG ODM-based application (ILOG, 2006).

includes a recommended plan or schedule and the attendant

metricscan easily be further explored, allowing users to

understand the optimization models dynamics and perhaps

identify better solution scenarios.

9.7.1 MATLAB

MATLAB (short for matrix laboratory) is an interpreted language

for numerical computation (MathWorks, 2009). It allows performing

numerical calculations and then visualizing the results without

the need for complicated and time-consuming programming.

MATLAB allows users to solve problems accurately, to produce

graphics easily, and to generate code efficiently. It also enables

matrix manipulation, the plotting of functions and data,

implementation of algorithms, creation of user interfaces, and

interfacing with programs written in other languages. For technical

problem solving, MATLAB has many advantages over conventional

computer languages, as described next (see also Chapman, 2009).

Ease of use: Programs can be easily written and modified under

the built-in integrated development environment, and they can be

debugged using the MATLAB debugger.

Sof tware Tools 211

Platform independence: MATLAB is supported on many different

computer platforms, which provides a large measure of platform

independence. The language is supported on Windows, Linux,

several versions of UNIX, and the Macintosh. This means that

programs written in MATLAB can migrate to different platforms.

Predefined functions: MATLAB comes complete with an extensive

library of well-tested, predefined functions that generate solutions

to many basic technical tasks. The arithmetic mean, standard

deviation, median, and hundreds of other mathematical functions

are built in to the MATLAB language, which makes the users job

much easier.

Device-independent plotting: Unlike most other computer languages,

MATLAB has many commands for imaging and integral plotting.

The images and plots can be displayed on any graphical output device

supported by the computer that is hosting MATLAB.

Graphical user interface: MATLAB includes tools with which a

programmer can interactively construct a graphical user interface for

any program. Given this capability, programmers can design

sophisticated data-analysis programs that can be operated by

relatively inexperienced users.

The built-in functions of MATLAB allow users to perform basic

minimization and maximization routines. However, compiling and

executing a proper optimization program may require the use of

add-on packages. One popular add-on is TOMLAB (TOMLAB, 2010),

a powerful optimization platform and modeling language for

solving applied optimization problems in MATLAB. The TOMLAB

environment includes a wide range of features, tools, and services

for optimization analyses.

There are two free alternatives to MATLAB software: SCILAB

(Scilab, 2009) and OCTAVE (Octave, 2009). Both provide number-

crunching power similar to MATLABs but at an advantageous

cost/performance ratio (since they are free). In essence, SCILAB and

OCTAVE are interpreted, matrix-based programming languages.

They have strong similarities to MATLAB: (1) the use of matrices as

a fundamental data type; (2) built-in support for complex numbers;

(3) powerful built-in math functions and extensive function libraries;

and (4) extensibility in the form of user-defined functions and macro

languages.

9.8.1 Modelica

Modelica is an object-oriented, declarative, multidomain language

for the component-oriented modeling of complex systemsthat is,

212 Chapter Nine

thermal, control, power, and/or process-oriented subcomponents

(Modelica, 2009a; Modelica, 2009b). The free Modelica language is

developed by the nonprofit Modelica Association. In this language,

models are described by classes, which may contain differential,

algebraic, and discrete equations alongside properties and algorithms.

The language can be used for hardware in the loop simulations

and for embedded control systems (Modelica, 2009a). Modelica

supports high-level modeling by composing complex models from

detailed, component models. Models of standard components are

typically available in model libraries. A model can be defined by

using a graphical model editor offered by the various language

implementations (Modelica, 2010) to draw a composition diagram:

positioning icons that represent the model components, drawing

connections between the components; and providing parameter

values in dialogue boxes (Modelica, 2009b). Constructs for including

graphical annotations in Modelica render the icons and composition

diagrams portable between different platforms. Typical composition

diagrams from various domains are shown in Figure 9.10.

In addition to the basic language elements mentioned previously,

Modelica also supports arrays (via a MATLAB-like syntax). Array

elements may consist of basic data types (e.g., real, integer, boolean,

string) or, more generally, of component models. This flexibility

allows for convenient descriptions of complex models containing

repetitive elements.

Lead times for the development of new energy technologiesfrom

initial idea to commercial applicationcan run into years. Reducing

b3={0,0,0}

Inertial

y

3D mechanics

x

b1={0,0,0} b4={0,0.73,0}

load

r2={1,0,0}

r3={1,0,0}

r1={0,1,0}

r5={1,0,0}

r4={0,1,0}

b0 r6={0,1,0}

state machines

b2={0,0.5,0} b5={0,0,0}

time>3

T2

R=50

control systems

Rp2

control

motor gear load

R=250

VolP

Ra

1 hydraulics

P4

C Rp1

J=1 q

time>5

C=C R=200

T4

Rel

L=L

La

power trains

power systems

T1 Line1

p1 p2 p3 k=k 3rd

Order

electrical systems Line2

emf

G2

Sof tware Tools 213

this lead time is a primary objective of the European Commissions

Directorate-General for Transport and Energy (DGTREN), which has

funded two related projects, EMINENT (Early Market Introduction

of New Energy Technologies) and EMINENT2 (Kleme et al., 2005b).

The principal features of these projects are a software tool and an

integrated database of new technologies and sectoral energy supplies

and demands. The software tool is for analyzing the potential

impact of new and underdeveloped energy technologies emerging in

different sectors from different countries. This tool has also been

used in case studies that illustrate the new technologies.

The aim of the EMINENT software is to assess the market

potential of early-stage technologies (ESTs) in various energy supply

chains by evaluating their performance in terms of: (1) CO2 emissions,

(2) costs of energy supply, (3) use of primary fossil energy, and

(4) effects on different subsectors of society. Technology developers

and financial supporters are frequently not aware of all the potential

applications or the relative market attractiveness of such technology

across different countries and sectors of society. Thus, the EMINENT

project provides insight into the market potential that can accelerate

the development of technologies; this benefits research and development

efforts by targeting them more effectively.

The EMINENT tool that evaluates ESTs makes use of two databases:

(1) national energy infrastructures, which contain information on the

number of consumers per sector, type of demand, typical quality of

the energy required, and consumption and installed capacity per end

user and (2) other ESTs and technologies that are already commercial

including key information on new energy technologies currently

under development and proven energy technologies now available

and in use. The availability, price, and geographical conditions of

primary energy resources differ significantly worldwide, so the impact

of ESTs can be evaluated only within the context of a particular

(national) energy supply system (Kleme et al., 2007b). The EMINENT

package consists of an integrated resource manager, a demand

manager, and an EST manager as well as databases on resources and

demand. The methodology, as shown in Figure 9.11, can be briefly

summarized as follows:

resources (electrical, fuel-based, geothermal, hydro, ocean tidal,

wave, and wind energy).

Demand manager describes energy demands per subsector in a

given country; it selects data for the technology assessment, enters

new data, and modifies old data.

Technology manager stores key data on existing technologies and

ESTs.

User input: (1) The sectoral energy demands whose potential supply

by EST is being evaluated; (2) any other peripheral technologies

214 Chapter Nine

Solar Coal

Hydro Gas/Oil

Hydrogen

Wind

Ground

Biomass Heat

new technology potential

Intermediate products

Biomass/waste

Pre-

Coal treat- Electricity

ment- Transpor- Conver-

Oil sys- tation and sion

tems storage processes

Natural gas Distribu- Heat

tion and

EST storage

Hydro

Transportation

Wind to be fuel

evaluated

Solar energy

Geothermal

2007b).

the EST would require to satisfy the full energy supply chain.

Output: (1) aggregate numbers; (2) application potential of ESTs per

(sub)sector; (3) annual cost of energy delivery per consumer and

per (sub)sector; and (4) annual CO2 emissions.

Performance indicators: (1) efficiency of energy supply chain;

(2) usage of primary fossil energy; (3) CO2 emissions per MWh;

and (4) costs of delivered energy [/MWh].

Most of the ESTs analyzed were not yet able to achieve the cost

levels of existing technologies. Some of the ESTsfor instance,

molten carbonate fuel cells (MCFCs)could become competitive

with relatively small efforts aimed at cost reduction. Various

promising future trends located with the help of EMINENT have

been supported by policy makers. Some near-term prospects are:

(1) diverse energy systems that encompass supplies, management,

and control of demand; (2) market-based grids with large power

stations, including wind farms of different types; (3) local distributed

generation, including biomass, waste, and wind; (4) micro generation,

Sof tware Tools 215

including Combined Heat and Power (CHP), fuel cells, and

photovoltaic technology; new homes built with nearly zero carbon

emissions; (6) natural gas; (7) coal-fired generation combined with

carbon capture and storage; and (8) mixed fuels (e.g., coal mixed

with biomass, natural gas mixed with hydrogen). Longer-term

prospects include nuclear power stations, fuel cells, hydrogen

obtained from nonelectricity sources (e.g., biomass, low-carbon

biofuels), and nuclear fusion.

Energy-Saving Analysis

Tools for balancing, reconciliation, and flowsheeting simulation are

frequently used for energy-saving analysis and have become an

essential item in the process engineers toolbox. These tools help

designers develop complete mass and energy models based on actual

measurements and/or design values and mathematical models. As a

result, these simulation tools play an important role in the technical

and economic decision making related to the planning and design

stages of processes under development and to the operation of

existing equipment. Several computer-based systems have been

developed over the years in order to assist process engineers with

energy and mass balance calculations. However, ongoing development

costs have resulted in a limited number remaining on the market,

and these have been secured only by a substantial volume of sales.

An early overview of the field was presented by Kleme (1977).

The technology used for balancing and for data validation and

reconciliation consists of a set of procedures incorporated into a

software tool. Process data reconciliation has become the main

method for monitoring and optimizing industrial processes as well as

for performing component diagnosis, condition-based maintenance,

and online calibration of instrumentation. According to Heyen and

Kalitventzeff (2007), reconciliation has three main goals: (1) detect

and correct deviations and errors of measurement data so that all

balance constraints are satisfied; (2) exploit the structure of and

knowledge about the process system by using measured data to

estimate unmeasured data whenever possible, in particular as

regards key performance indicators (KPIs); and (3) determine the

postprocessing accuracy of measured and unmeasured data,

including KPIs.

A comprehensive system called DEBILwhich included

balancing, flowsheeting simulation, and optimizationwas created

several decades ago (see Kleme, Lucha, and Vaek, 1979) and has

since been further developed by Belsim into the balancing and

reconciliation tool VALI (VALI III User Guide, 2003). This tool has been

applied to various energy-efficiency tasks, as reported with respect

to nuclear power plants (Langenstein, Jansky, and Laipple, 2004) and

regenerative heat exchangers (Minet et al., 2001).

216 Chapter Nine

Other Energy Systems

There are many different computer tools now available that can be

used to analyze the integration of renewable energy. Connolly and

colleagues (2010) have reviewed these tools in considerable detail.

Their study analyzed 37 computer tools designed for this purpose, as

summarized (in alphabetical order) in Table 9.1.

AEOLIUS (commercial) Power plant dispatch Universitt Karlsruhe

simulation tool (2009)

BALMOREL (free Open-source electricity Balmorel (2009)

download) and district heating tool

BCHP Screening Tool Assesses CHP in buildings Oak Ridge National

(free download) Laboratory (2009a)

COMPOSE (free Single-project techno- Aalborg University

download) economic assessments (2009a)

E4cast (commercial) Tool for energy projection, ABARE (2009)

production, and trade

EMCAS (commercial) Creates techno-economic Argonne National

models of the electricity Laboratory (2009a)

sector

EMINENT (registration Early-stage technologies EMINENT2 (2009)

required) assessment

EMPS (commercial) Electricity systems with SINTEF (2009b)

thermal/hydro generators

EnergyPLAN (free User-friendly analysis of Aalborg University

download) national energy systems (2009b)

energyPRO Single-project techno- EMD International A/S

(commercial) economic assessments (2009)

ENPEP-BALANCE (free Market-based energy- Argonne National

download) system tool Laboratory (2009b)

GTMax (commercial) Simulates electricity Argonne National

generation and flows Laboratory (2009c)

H2RES (internal use Energy-balancing models Instituto Superior

only) for island energy systems Tcnico (2009)

HOMER (free Techno-economic HOMER (2009)

download) optimization for stand-

alone systems

TABLE 9.1 Software Tools for Analyzing the Integration of Renewable Energy into

Other Energy Systems (after Connolly et al., 2010)

Sof tware Tools 217

HYDROGEMS (free for Renewable and H2 stand- Institute for Energy

TRNSYS users) alone systems Technology (2009)

IKARUS (earlier Bottom-up cost Forschungszentrum

versions are free) optimization tool for Jlich Institute for

national systems Energy Research (2009)

INFORSE (free to Energy-balancing models INFORSE-Europe (2009)

nongovernmental for national energy

organizations) systems

Invert (free download) Simulates promotion Vienna University of

schemes for renewable Technology (2009)

energy

LEAP (free for User-friendly analysis for Stockholm Environment

developing countries national energy systems Institute (2009)

and for students)

MARKAL/TIMES Energyeconomic tools for International Energy

(commercial) national energy systems Agency (2009)

MESAP PlaNet Linear network models of Schlenzig (2009)

(commercial) national energy systems

MESSAGE (free except Medium- and long-term International Institute

for the simulators) assessment of national or for Applied Systems

global energy systems Analysis (2009)

MiniCAM (free Simulates long-term, MiniCAM (2009)

download) large-scale global changes

NEMS (free except for Simulates the U.S. energy Energy Information

the simulators) market Administration (2009)

ORCED (free Simulates regional Oak Ridge National

download) electricity dispatch Laboratory (2009b)

PERSEUS (sold to large Family of energy and Universitt Karlsruhe

European utilities) material flow tools (2009)

PRIMES (projects are Market equilibrium tool National Technical

undertaken for a fee) for energy supply and University of Athens

demand (2009)

ProdRisk (commercial) Optimizes operation of SINTEF (2009a)

hydro power

RAMSES (projects are Simulates the electricity Danish Energy Agency

undertaken for a fee) and district heating sector (2009)

RETScreen (free Renewable analysis for National Resources

download) electricity and/or heat in Canada (2009)

systems of any size

TABLE 9.1 Software Tools for Analyzing the Integration of Renewable Energy into

Other Energy Systems (Continued)

218 Chapter Nine

SimREN (projects are Bottom-up supply and Institute for Sustainable

undertaken for a fee) demand for national Solutions and

energy systems Innovations (2009)

SIVAEL (free download) Electricity and district Energinet.dk (2009)

heating sector tool

STREAM (free Overview of national Ea Energy Analyses

download) energy systems to create (2009)

scenarios

TRNSYS16 Modular structured TRNSYS (2009)

(commercial) models for community

energy systems

UniSyD3.0 (contact National energy systems Unitec New Zealand

jleaver@unitec.ac.nz) scenario tool (2009)

WASP (free to IAEA Identifies the least-cost IAEA (2009)

member states) expansion of power plants

WILMAR Planning Tool Increasing the use of Ris National

(commercial) wind in national energy Laboratory (2009)

systems

TABLE 9.1 Software Tools for Analyzing the Integration of Renewable Energy into

Other Energy Systems (Continued)

CHAPTER 10

Examples and

Case Studies

T

his chapter provides a selection of problems that have been

collected by the authors over twenty years of teaching and

consulting. The problems include step-by-step solutions,

which provide guidance for mastering the methodology. Space

limitations make it impossible to cover all aspects or to provide fully

comprehensive solutions, for which an entire book would be required.

The representative problems selected for presentation here are

divided into five groups: Heat Pinch Technology, Total Sites,

integrated placement of processing units, utility placement, and

Water Pinch Technology.

The first group of problems involves heat Process Integration, known

also as Heat (and, in Australia and South Africa, as Thermal) Pinch

Technology.

Problem 1: Task Assignment

A chemical process is served by a Heat Exchanger Network (HEN),

as displayed in Figure 10.1. The figure shows the supply and the

target temperatures at the start and end of each stream. For this

problem, the minimum allowed temperature difference Tmin = 10C.

The following tasks should be performed:

representation to a grid diagram. (2) Calculate the missing

temperatures and heat loads for the recovery heat exchangers,

heaters, and coolers.

(b) Identification of the Network Pinch. (1) Find and write down a

path that connects a heater and a cooler through a recovery

heat exchanger. Derive expressions for the heat loads and

219

220 Chapter Ten

H H

CP [kW/C]

337C 190C 40C

100 H1 C

160C

160 H2 220C

220C 60C

50 H3 C

160C 45C

190 H4 C

C1 C2 C3 C4 C5

X [kW] through the path in order to increase the heat

recovery. (2) Shift the maximum amount of load through the

path while accounting for the minimum allowed temperature

difference. Write down the maximum shifted load, the

pinching exchanger, and the stream temperatures for the

exchanger. Using the Pinch method yields the Maximum

Energy Recovery (MER) targets listed in Table 10.1 and the

Grand Composite Curve (GCC) shown in Figure 10.2.

(c) Identification of the scope for improvement. (1) Calculate the

scope for improvement in heat recovery in terms of the

networks total heating requirement. (2) Find the heat

exchangers, implementing cross-Pinch heat transfer, and

write them down.

Problem 1: Solutions

Answer to (a)(1) and (a)(2). The HEN is represented as a grid diagram in

Figure 10.3, which also shows the missing HEN parameters (i.e., temperatures

and loads).

Answer to (b)(1). A path between a cold and a hot utility is called a utility path.

Heat duty can be shifted along a utility path, which provides a degree of freedom

in the HEN retrofit. Figure 10.4 shows a utility path connecting a heater and a

Examples and Case Studies 221

1 332 17,780

2 305 20,480

3 215 2,480

4 175 2,880

5 169 2,580

6 155 900

7 145 0 (Pinch)

8 130 1,650

9 105 25

10 85 725

11 65 4,925

12 55 7,625

13 40 10,925

14 35 11,425

for Problem 1

Tmin = 10C

350

300

250

200

T* [C]

150

100

50

0

0.0 5.0 10.0 15.0 20.0

H [MW]

222 Chapter Ten

CP [kW/C]

40.0 99.7 190.0 337.0

C3 E2 E1 H1 100

5970 kW

160.0 220.0

E4 H2 160

60.0 88.0 220.0

C6 E5 H3 50

C10 E7 H4 190

100.0 300.0

247.0

C1 H8 100

35.0 14700 kW 5300 kW 164.0

C2 70

C3 175

7875 kW 170.0

60.0 60

C4

6600 kW 300.0

140.0 188.0 200

C5 H9

9600 kW 22400 kW

Load-shift path

CP [kW/C]

C3 E2 E1 H1 100

QC3

160.0 220.0

E4 H2 160

88.0

60.0 220.0

C6 E5 H3 50

1400 kW 118.55

45.0 160.0 190

C10 E7 H4

13984 kW

100.0 TC11 300.0 100

C1 H8

QH8

QE1

TC2,in=35.0 TC2,out=164.0 70

C2

80.0 9030 kW 125.0 175

C3

60.0 7875 kW 170.0

C4 60

140.0 6600 kW

188.0 300.0

C5 H9 200

9600 kW 22400 kW

Examples and Case Studies 223

cooler through a heat recovery exchanger. The temperatures, which will vary

with the amount of heat shifted, are also indicated in the figure.

The expressions for the variation in the affected heat loads and temperatures

in response to a shifted load of X kW are displayed in Eqs. (10.1)(10.8) (notation

as in Figure 10.4):

heat exchangers, which are E1 and E2. As can be seen, the heat capacity flow

rate (CP) values for both streams in E1 are equal to 100 kW/C. Exchanger E2s

hot stream has a higher CP value than that for the cold stream, so the smaller

temperature difference will be at its hot end. This allows one to calculate the

maximum amount of the load X to shift by solving a few inequalities. For

process exchanger E2, the temperature difference should not be less than

Tmin, and this determines the value of the maximum heat load that can be

shifted:

X 2600 kW (10.11)

For this value of load shift, the exchanger E2 will be pinched at its hot end as

follows: TH11 = 174.0C and TC2,out = 164.0C. The temperatures at the cold end

will be TH12 = 83.7C and TC2, in = 35.0C.

Answer to (c)(1). Total heating requirement for the existing network:

Q H = Q H8 + Q H9 = 27,700 kW. Total cooling requirement for the network:

Q C = Q C3 + Q C6 + Q C10 = 21,345 kW. There is no inappropriate use of utilities.

Compared with the targets, both total utility heating and total utility cooling

are higher by 9920 kW, which is due to cross-Pinch heat transfer.

Answer to (c)(2). The Process Pinch is at 150C for hot streams and at 140C for

cold streams. Comparing the network temperatures with the Pinch location

yields a list of exchangers that violate the Pinch; see Table 10.2.

224 Chapter Ten

E1 No Yes

E2 Yes Yes

E4 No No

E5 Yes Yes

E7 Yes No

1 Cold 38 205 11

2 Cold 66 182 13

3 Cold 93 205 13

4 Hot 249 121 17

5 Hot 305 66 13

Problem 2: Task Assignment

The stream data for a process are as listed in Table 10.3. The hot utility

is steam at 200C, the cold utility is water at 38C, and Tmin = 24C.

(b) Determine Q H,min, QC,min, and the Pinch temperatures.

(c) Assuming that the cost of cooling water and steam is 18.13

and 37.78 $ kW1 y1, plot the minimum annual cost for the

utilities as a function of Tmin in the range 5154C (in 1

increments).

(d) Design a network that features a minimum number of units

and maximum energy recovery for Tmin = 24C.

Problem 2: Solutions

Answer to (a). The Composite Curves for the process stream data are shown in

Figure 10.5.

Answer to (b). The position of the CCs in Figure 10.5 indicates that, for

Tmin = 24C, this is a threshold problem at the cold end: Q H,min = 0 kW. On the

cold end there is excess of hot streams, which means that some cold utility is

required: Q C,min = 482 kW.

Examples and Case Studies 225

Curves for Problem 2. T [C] Tmin = 24C

No

heating utility

200

100

482 kW

0

0 2000 4000 H [kW]

Utility

Cost [$/y]

12000

11000

10000

9000

8000

51 52 53 54 Tmin [C]

utility cost.

Tmin [C]

51 52 53 54

QH [kW] 0 0 27 57

Cost of heating [$/y] 0 0 1,020 2,153

QC [kW] 482 482 509 539

Cost of cooling [$/y] 8,739 8,739 9,228 9,772

Total utility cost [$/y] 8,739 8,739 10,248 11,926

TABLE 10.4 Utility Requirements and Cost for Various Tmin Values

Answer to (c). The utility cost plot, which is given in Figure 10.6, is based on the

data in Table 10.4.

Answer to (d). A network that features the minimum number of units and

maximum energy recovery is shown in Figure 10.7.

226 Chapter Ten

(a) CP [kW/C]

121C 163.4C 249C

E4 E1 4 17

CP=9.1 kW/C

E2

66C 103.1C 305C

C E3 5 13

CP=3.9 kW/C

482 kW

38C 205C 11

1 E2

1837 kW

66C 121.43C 182C 13

2 E4 E3

720 kW 788 kW

93C 205C

3 E1 13

1456 kW

(b) CP [kW/C]

CP=5.22 kW/C

E3

E1 17

121C 249C

E2 4

CP=11.78 kW/C

C E4 E1 5 13

482 kW

38C 144.3C 205C 11

1 E4 E3

1169 kW 668.16 kW

66C 182C 13

2 E2

1508 kW

93C 205C

3 E1

13

1456 kW

10.2.1 Total Sites: First Problem

Problem 3: Task Assignment

Suppose a Total Site incorporates two processes, A and B. The stream

data for process A (assuming TA,min = 10C) are given in Table 10.5,

and the GCC for this process is shown in Figure 10.8. The stream

data for process B (assuming TB,min = 10C) are shown in Table 10.6.

Examples and Case Studies 227

A1 50 140 450

A2 100 30 420

A3 100 140 80

Tmin = 10C

(330;145)

150

Sink A1

Sink A2

T* [C]

100 (59;105)

(0;95)

Source A1

50

(40;55) Source A2

(220;25)

20

0 100 200 300

H [kW]

B1 190 120 6

B2 100 240 4

B3 80 60 2

saturation temperature, and cooling water is available at 2030C.

It is necessary to analyze these two processes (and hence the

Total Site) as follows:

for processes A and B.

(b) Construct the GCC for process B.

228 Chapter Ten

(d) Assuming that the site manager is willing to accept just one

more steam main in addition to the HP, identify this steam

level on the Site Profiles. What saturation temperature will

yield the most energy savings?

(e) Determine the load target for this steam level.

Problem 3: Solutions

Answer to (a). For process A, Tmin = 10C, the minimum hot utility is 330 kW,

and the minimum cold utility is 220 kW (see Table 10.7).

For process B, Tmin = 10oC, the minimum hot utility is 240 kW, and the minimum

cold utility is 140 kW (see Table 10.8).

Answer to (b). Given the data in Table 10.8, the GCC for process B is drawn as

shown in Figure 10.9.

Answer to (c). Using the GCC plots from Figures 10.8 and 10.9, the heat source

and heat sink segments have been extracted. These segments are listed in

Tables 10.9 and 10.10 for process A and process B, respectively.

The data reported in Tables 10.9 and 10.10 have been combined into

composite heat source and sink profiles. Figures 10.10 and 10.11 illustrate the

composition procedure, which is analogous to the one for constructing the

process CCs. Figure 10.12 shows the resulting TSPs.

1 145 330

2 105 50

3 95 0 (Pinch)

4 55 40

5 25 220

1 245 240

2 185 0 (Pinch)

3 115 140

4 105 100

5 75 100

6 55 140

Examples and Case Studies 229

Tmin = 10C

(240;245)

Sink B1

(0;185)

200

Source B1

T* [C]

(100;135)

(140;115)

(100;105)

100 Source B2

(100;75)

(140;55)

40

0 50 100 150 200 250

H [kW]

Segment T *start [C] T *end [C] H [kW] T **start [C] T **end [C]

Sink A1 105 145 280 110 150

Sink A2 95 105 50 100 110

Source A1 95 55 40 90 50

Source A2 55 25 180 50 20

TABLE 10.9 Heat Source and Sink Segments from the GCC for Process A

[Problem 3(c)]

Segment T *start [C] T *end [C] H [kW] T **start [C] T **end [C]

Sink B1 185 245 240 190 250

Source B1 185 135 100 180 130

Source B2 75 55 40 70 50

TABLE 10.10 Heat Source and Sink Segments from the GCC for Process B

[Problem 3(c)]

Answers to (d) and( e). By analyzing the TSPs (Figure 10.12), one can see that

there is an opportunity to match heat source and heat sink requirements

in the steam temperature interval between 100C and 180C. The proposed

steam saturation temperature is in the interval from 117.1C to 130C. Within

this interval, the maximum heat recovery through the steam system (equal to

100 kW) is achieved. This results from matching the values for steam generation

and steam use (here both are equal to 100 kW).

230 Chapter Ten

TB [C]

A1 A2 B1 B2

180

CP = 2 H = 100

130

CP = 0 H = 0

90

CP = 1 H = 20

70

CP = 3 H = 60

50

CP = 6 H = 180

20

CP [kW/C] 1 6 2 2

Combined Site Sink A1 Sink A2 Sink B1

heat sinks for 250

Problem 3(c). CP = 4 H = 240

190

CP = 0 H = 0

150

CP = 7 H = 280

110

CP = 5 H = 50

100

H [kW] 280 50 240

CP [kW/C] 7 5 4

Temperature

250

100 kW heat recovery via

the steam system

150

T = 130C

T = 117.1C

50

H [kW]

Examples and Case Studies 231

Problem 4: Task Assignment

Two processes (A and B) are operating on a site. The stream parameters

for process A and process B are given in Tables 10.11 and 10.12,

respectively. Medium-pressure (MP) steam from a boiler is available

at 220C saturation temperature, and Tmin = 10C for both processes.

The cooling water supply temperature is 18C and the return

temperature is at most 35C.

minimum heating/cooling duty for each process, and plot

the GCCs.

(b) Create the TSPs including the GCC pockets.

(c) Select the saturation temperature for a second (low-pressure)

steam level, and construct the Total Site composites, so as to

maximize heat recovery via the steam system.

Problem 4: Solutions

Answer to (a). Pinch Analysis has been applied to the two processes just

described, and the targets thus identified are given in Table 10.13.

The GCCs for process A and process B are shown in Figures 10.13 and 10.14,

respectively.

A1 143 160 0.15 2.55

A2 75 116.01 0.15 6.16

A3 124 90 0.412 14.0

A4 90 84 0.35 2.1

A5 84 37.5 0.101 4.7

B1 158.3 159.9 1.33 2.08

B2 70 158.3 0.05 4.42

B3 10 70 0.03 1.8

B4 180 135.6 0.075 3.33

B5 135.6 105 0.105 3.21

232 Chapter Ten

Process A 2.852 14.955 124 / 114

Process B 1.800 0 None (threshold problem)

T [C]

(2.9; 165.0)

160

140 (0.3; 148.0)

(0.3; 121.0)

120

(8.9; 85.0)

100 (0; 119.0) (10.3; 79.0)

80

(9.9; 80.0)

60

40

20 (15.0; 32.5)

0

0 2 4 6 8 10 12 14 H [MW]

T* [C]

180 (1.8; 175.0)

(0.5; 163.3)

160 (2.6; 164.9)

140

120 (1.4; 130.6)

100

(3.1; 100.0)

80

60 (1.8; 75.0)

40

20

(0; 15.0)

0

0 0.5 1 1.5 2 2.5 3 H [MW]

Answer to (b). The TSPs for the combination of the two processes are plotted in

Figure 10.15. The apparent temperature overlap of the profiles indicates a good

opportunity for heat recovery.

Answer to (c). The heat recovery on the site can only take place via the steam

system. Any steam raised from process cooling has to be utilized to the greatest

Examples and Case Studies 233

extent possible in order to maximize the recovery. Usage may exceed generation;

in this case, the difference would be made up by MP steam. The profiles in

Figure 10.15 were used to balance steam generation and use by the processes

by varying the LP level saturation temperature. This has been achieved for LP

generation and use steam loads of 6.9 MW at 109C saturation temperature,

which corresponds to 1.39 bar(a) steam pressure. The Total Site composites are

shown in Figure 10.16

T** [C]

200

100

50

30 25 20 15 10 5 0 5 10 15 20

H [MW]

T** [C]

200

6.9 MW

100

50

30 25 20 15 10 5 0 5 10 15 20

H [MW]

234 Chapter Ten

Units and Data Extraction

Problem 5: Task Assignment

The process flow sheet for a base-case design is shown in Figure 10.17

with some additional data. The CCs and GCC of the process (with

appropriate data) for Tmin = 20C are shown in Figures 10.18 and

10.19, respectively.

purge

120C

FR m=0.4

H

C 900

m=1 1300 505 180C CA

50C 130C

F1 H 180C

R1 A

50C 180C T=180C Column A

F2

m=0.8 HM

220C

H m=1.4 220C

RA

1050 H

CB 200C

1000

980 C

200C

70C C Heat exchanged

P2

with cold utility

m=1.3 H Heat exchanged

B

with hot utility

Column B Q

RB Heat exchanged [kW]

1928 m=[kg/s]

m=0.5 270C

70C 150C 270C

P1 C H

FIGURE 10.17 Process flowsheet for the base-case design (Problem 5).

250

Pinch

200

Th = 240 C

Tc = 220 C

150

100

50

1958.6 10316.6

0

0 2000 4000 6000 8000 10000 H [kW]

Examples and Case Studies 235

T (C)

250

200

150

100

50

0

0 500 1000 1500 2000 2500 Q (kW)

Analysis. Tabulate the extracted data where appropriate. If a

stream involves phase changes (several segments), then give

the relevant details (e.g., stream F1 has three segments:

liquid phase, vaporization, and vapor phase).

(b) In the base-case design, what is the hot and cold utility

consumption and how much is the process heat recovery?

Compare the utility usage in the base case with the targets

shown on the CCs. Why are they different?

(c) Explain how the heat recovery system can be improved over

the base case without modifying any conditions (e.g., reactor

pressure, column pressure) of the main process.

(d) Refer to the GCC in Figure 10.19.

(1) If a heat pump is proposed, what operating temperature

levels will best fit the heating and cooling needs of the

process?

(2) What qualitative suggestions can be made concerning

alternatives for utility saving and appropriate utility

placement?

(e) If process modifications are allowed, comment on whether

they could help to improve the process Heat Integration.

Your answer should address the reactors, the distillation

columns, feed vaporization, and soft data.

generated can be sold, and the unit cost of high-temperature utilities

236 Chapter Ten

flue gas for direct heating ranges from 800C to 160C. Steam usage

and generation are isothermal utilities (up to four levels between

100C and 340C). A heat pump or heat engine can be introduced.

Additional data for the problem:

involve only phase changes.

In the condenser of distillation column B, total condensation

is assumed (i.e., the outlet is in the liquid phase only).

An endothermic reaction takes place in reactor R1. A heating

medium (HM) is necessary to keep the isothermal condition

in R1.

Cp,vap = 4.5 kJ/kgC

F2: Boiling point = 125C, Hvap = 1250 kJ/kg,

Cp,vap = 7.5 kJ/kgC,

Cp,liq = 15 kJ/kgC

FR: Cp,vap = 5.0 kJ/kgC

HM: Cp,liq = 25 kJ/kgC

P1: Boiling point = 270C, Hvap = 1556 kJ/kg

P2: Boiling point = 200C

Problem 5: Solutions

Answer to (a). Stream data needed to perform a Pinch Analysis is reported in

Table 10.14.

Answer to (b). For the base case, Q H = 4828 kW, Q C = 3366.4 kW, and process heat

recovery = 3530 kW. The targets are Q H = 2642.6 kW and Q C = 1958.6 kW, and

process heat recovery = 5715.8 kW. The base-case design requires much more

utility consumption. This is because some heat is transferred across the Pinch

and also because there is some unnecessary utility use. For example, some

amount of heat is needed in the column B reboiler for heating up P1, and this

amount must be taken out again when P1 needs cooling.

Answer to (c).

The cold streams below the Pinch are F1, F2, and FR. These streams can

be heated up by P1 (parts below the Pinch), P2, and by the condensers of

columns A and B. In the base case, stream P1 is used to supply heat to

stream F2, resulting in heat transfer across the Pinch. Instead, to heat up

stream F2 at its higher temperature range, heat exchange with stream P2

can be employed; for the lower-temperature part of F2, the heat from the

condenser of column A can be recovered.

Above the Pinch, P1 can be used to heat up parts of the column A

reboiler or of the heating media of reactor R1.

Examples and Case Studies 237

[kg/s] [kJ/kgC] [kJ/kg] [kW/C] [kW]

F1-liq Cold 50 120 1 11 11 770

F1-evap old 120 120 1 990 990

F1-vap old 120 180 1 4.5 4.5 270

F2-liq Cold 50 125 0.8 15 12 900

F2-evap Cold 125 125 0.8 1250 1000

F2-vap Cold 125 180 0.8 7.5 6 330

HM Cold 220 250 1.4 25 35 1050

CA Hot 120 120 900

RA Cold 220 220 1000

FR Cold 120 180 0.4 5 2 120

CB Hot 200 200 980

RB Cold 270 270 1928

P1-cond Hot 270 270 0.5 1556 778

P1-liq Hot 270 70 0.5 37.17 18.58 3716

P2 Hot 200 70 0.8 12.5 10 1300

Answer to (d)(1). Using the GCC in Figure 10.19, from the viewpoint of a heat

pump, two good candidates for its integration are as follows:

Across the Process Pinch. Heat rejection from the heat pump to the

process T* = 230C (real temperature T = 240C) and heat absorption

from the process at T* = 190C (T = 180C), would result in temperature

lift Tlift = 60C. The main challenge is that the heat available for

absorption by the heat pump is relatively smaller than what can be

delivered by the heat pump above the Process Pinch.

Across a potential Utility Pinch. Heat rejection level T* = 130C and

absorbtion heat level T* = 110C. The corresponding real heat pump

temperatures are: heat rejection level T = 140C and heat absorption

level, T = 100C. This results in temperature lift Tlift = 40C. At this

level, a large amount of heat is available to be absorbed from the process

and an even larger amount can be rejected to the process.

Answer to (d)(2).

From the GCC, the heat pump integration can be placed at the levels of

140C and 100C with only a small amount of shaft work required. This

could save a large part of the cold utility required.

Above the Pinch, a large and quite flat pocket is present in the CCC,

providing an opportunity to generate steam at the maximum level

of 180C.

Above the Pinch, steam is preferred over flue gas for a hot utility because

the process stream temperatures are not too high.

238 Chapter Ten

GCC below the Pinch. However, since T is smallabout 80C, or

[(190 + 10) (130 10)]only a small amount of work may be produced.

Answer to (e).

Reactors. The reaction is endothermic and, to save utility, its heating

medium heat demand should be placed below the Pinch. It is now placed

above the Pinch. If the pressure and temperature of reactor R1 are

reduced, then an alternative heating media with lower temperature levels

could be used. As a result of decreased R1 pressure, the boiling points of

feeds F1 and F2 would also decrease. This would result in different

targets. For instance the GCC could be shifted more to the left and the

Pinch point could be changed, resulting in more process heat recovery.

Distillation columns. The reboilers of columns A and B currently

evaporate the entire flow from the column bases. The appropriate

arrangement would be to move the reboilers to evaporate only the

branches of these flows that are intended for the columns.

Feed vaporization. In the base case, the vaporization of F1 and F2 are both

placed below the Pinch. However, as mentioned, the current vaporization

of F2 results in cross-Pinch heat transfer. Also, the vaporization of F1

employs utility heating below the Pinch, which should be eliminated

and substituted for process-to-process heat recovery.

Soft data. The P1 and P2 target temperatures are soft data that can be

changed. The temperatures could be increased to reduce cold utility

required, but this would not change the CC much.

10.4.1 Utility Placement: First Problem

Problem 6: Task Assignment

A set of data for a part of a crude distillation process has been

extracted, and the stream data are shown in Table 10.15. For this

problem, the minimum temperature difference is Tmin = 10C. The

available utilities are presented in Table 10.16.

A targeting procedure was performed for the initial data

summarized in Tables 10.15 and 10.16. Figure 10.20, which exhibits

two Pinch points, shows the balanced CCs for the problem.

(b) Identify the Pinches and explain their significance.

(c) Draw the design grid for the problem, including the hot utilities.

Position the streams relative to the location of the Pinches.

(d) Design the MER HEN below the Process Pinch.

(e) Design the MER HEN above the Process Pinch.

(f) Design the MER HEN above the Utility Pinch.

Examples and Case Studies 239

[kW/C]

1 OH_Naphta_1 100 41 27.5113

2 OH_Naphta_2 132 60 23.1850

3 OH_HKD 224 65 11.2400

4 LCT 268 30 363.12 1.5257

5 Residue 283 45 336.25 1.4128

6 Crude 30 146 2031.50 17.5132

7 Denaphta_1 117 204 977.88

8 Denaphta_2 176 305 1641.45 12.7244

[/kWy]

HP steam 320.1 320.0 100

MP steam 260.1 260 60

Cooling water 20.0 40.0 8

HP steam = 654.2 kW

320.0C 320.1C

MP steam = 874.4 kW

300 305.0C

260.1C

260.0C

250.1C

224.0C

200

T* [C]

204.0C

176.0C

132.0C 146.0C

100.0C 122.0C

100

C

5 .0

41.0C 4 60.0C

40.0C

30.0C 20.0C Cooling water = 2657 kW

H [kW]

FIGURE 10.20 Balanced Composite Curves and heat recovery targets for

Problem 6.

240 Chapter Ten

Problem 6: Solutions

Answer to (a). Table 10.17 shows the missing parameters of process streams

described in Table 10.15.

Answer to (b). There are two Pinches: the Process Pinch (located at 132C/122C)

and the Utility Pinch (located at 260.1C/250.1C; see Figure 10.21. The Utility

Pinch is caused by the placement of the MP steam. More significant is the

Process Pinch, which divides the design area into two different parts: a net

heat sink above the Pinch temperatures and a net heat source below the Pinch

temperatures.

The Utility Pinch further subdivides the area above the Process Pinch. In

the interval between the Process Pinch and the Utility Pinch, the only utility

allowed is the MP steam. Above the Utility Pinch, the only utility available

remains the HP steam.

Answer to (c). See Figure 10.22.

1 OH_Naphta_1 100 41 1623.17 27.5113

2 OH_Naphta_2 132 60 1669.32 23.1850

3 OH_HKD 224 65 1787.16 11.2400

4 LCT 268 30 363.12 1.5257

5 Residue 283 45 336.25 1.4128

6 Crude 30 146 2031.50 17.5132

7 Denaphta_1 117 204 977.88 11.2400

8 Denaphta_2 176 305 1641.45 12.7244

TABLE 10.17 Process Streams for Problem 6(a) with Missing Data Filled In

HP steam = 654.2 kW

320.0C 320.1C

MP steam = 874.4 kW 305.0C

300

260.1C

260.0C

Utility Pinch

224.0C 250.1C

200 204.0C

T [C]

176.0C

132.0C

146.0C

100.0C 122.0C

100

Process Pinch

C

5 .0

41.0C 4 60.0C

30.0C 20.0C Cooling water = 2657 kW

H [kW]

Examples and Case Studies 241

132.0 260.1

320.0 320.1

10 HP_steam

CP = 6541.63 H = 654.163

260.0 260.1

11 MP_steam

CP = 8743.73 H = 874.373

41.0 100.0

1 OH_Nafta_1

CP = 27.51 H = 1623.167

60.0 132.0

2 OH_Nafta_2

CP = 23.18 H = 1669.32

65.0 224.0

3 OH_HKD

CP = 11.24 H = 753.08 CP = 11.24 H = 1034.08

30.0 268.0

4 LCT

CP = 1.63 H = 165.621 CP = 1.63 H = 195.442 CP = 1.53 H = 12.053

45.0 283.0

5 Residue

CP = 1.41 H = 122.913 CP = 1.41 H = 180.979 CP = 1.41 H = 32.353

30.0 146.0

6 Crude

CP = 17.51 H = 1611.214 CP = 17.51 H = 420.316

117.0 204.0

7 Denafta_1

CP = 11.24 H = 56.2 CP = 11.24 H = 921.68

305.0

8 Denafta_2

CP = 12.72 H = 942.878 CP = 12.72 H = 698.589

20.0 40.0

9 Cooling water

CP = 132.83 H = 2656.687

122.0 250.1

Tmin = 10C; All temperatures are in [C], H are in [kW], CP are in [kW/C]

1623.17 kW 132.0

41.0 100.0

C1 1

57.71 kW

60.0 62.51 132.0

C3 2 2

696.88 kW

65.0 127.3 132.0

C5 4 3

155.62 kW

30.0 132.0

C6 4

122.91 kW

45.0 132.0

C7 5

30.0 122.0

6 2

1611.21 kW

117.0 122.0

7 4

56.2 kW

20.0

9 C1 C7 C6 C5 C3

122.0

FIGURE 10.23 HEN design below the Process Pinch [Problem 6(d)].

242 Chapter Ten

132.0 260.1

260.0 260.1

12 13 11 MP steam

214.0

224.0

CP = 11.24 8 9 3

CP = 1.53 10 4

CP = 1.41 11 5

144.32

CP = 8.755 10

142.67 195.44 kW 143.49 146.04

CP = 17.51 6 11 13

180.98 kW 43.90 kW

204.0

CP = 11.24 7 8

921.68 kW

176.0 185.32

CP = 12.72 8 9 12

112.4 kW 830.48 kW

122.0 250.1

FIGURE 10.24 HEN design above the Process Pinch [Problem 6(e)].

260.1

320.0 320.1

16 10

CP = 6541.86

12.05 kW

268.0

15 4

CP = 1.53

283.0

14 5

CP = 1.41

CP = 3.36 251.99

15

8 14 16

CP = 3.36 CP = 6.72

32.35 kW 654.16 kW

250.1

FIGURE 10.25 HEN design above the Utility Pinch [Problem 6(f)].

Examples and Case Studies 243

Problem 7: Task Assignment

A process involves the set of process streams described in Table 10.18.

The utilities available to satisfy its heating and cooling requirements

are given in Table 10.19.

For the HEN to be designed, assume Tmin = 30C.

duty, and the Process Pinch location.

(b) Plot the GCC.

(c) Perform an appropriate placement of the utilities against the

GCC so as to achieve minimum total cost of the utilities.

(1) Draft the placement on the GCC.

(2) Calculate the total duty for each of the utilities.

(3) Calculate the total utility cost for the problem.

(d) Consider the following ways to reduce the utility costs

further, and comment on their suitability for application to

H1 175 75 3.5

H2 100 40 2.4

H3 180 130 1.5

H4 195.1 195 100

C1 50 150 2.0

C2 140 140.1 150

C3 80 140 1.5

C4 20 80 3.0

HP steam 300.1 300 65000

MP steam 200.1 200 50000

Cooling water 15.0 20.0 7000

Chilled water 5.0 10.0 30000

244 Chapter Ten

an option can be exploited.

(1) Flue gas heating

(2) Heat pumping

(3) Introduction of a new steam level

Problem 7: Solutions

Answer to (a). The heat recovery targets and the Pinch location can be obtained

with the help of the PTA. The heat cascade intervals and the corresponding

stream population are shown in Table 10.20, and the computed problem table

is given in Table 10.21.

180.1

1 H4

180.0

2

165.0

3 H3, C1

160.0

3 H1, H3, C1

155.1

4 H1, H3, C1, C2

155.0

5 H1, H3, C1, C3

115

6 H1, C1, C3

95.0

7 H1, C1, C4

85.0

8 H1, H2, C1, C4

65.0

9 H1, H2, C4

60.0

10 H2, C4

35.0

11 H2

25.0

TABLE 10.20 Heat Cascade Intervals and Stream Population for Problem 7(a)

Examples and Case Studies 245

Alternatively, the heat recovery targets can be obtained by plotting the CCs,

as shown in Figure 10.26.

Answer to (b). See Figure 10.27.

Answer to (c)(1) and (c)(2). See Figure 10.28.

Answer to (c)(3).

Total utility cost = 37.5 [MW MP steam] 50,000 [$/MWy]

+ 62.5 [MW cooling water] 7,000 [$/MWy]

+ 24 [MW chilled water] 30,000 [$/MWy]

= 3,032,500 [$/y]

[C] [MW]

1 180.1 37.5

2 180 137.5

3 165 137.5

4 160 135.0

5 155.1 149.7

6 (Pinch) 155 0.0

7 115 60.0

8 95 60.0

9 85 45.0

10 65 63.0

11 60 77.5

12 35 62.5

13 25 86.5

Composite Curves for 250

Problem 7(a), QH,min = 37.5 MW

Tmin = 30C.

200

150

100

140C (cold)

50

QCmin=86.5 MW

0

0 200 400 600 800 H [MW]

246 Chapter Ten

Grand Composite T [C]

Curve for

Problem 7(b),

150

Tmin = 30C.

100

50

QCmin=86.5 MW

0

0 50 100 H [MW]

Appropriate

placement of

utilities for 150

Problem 7(c).

100

Cooling Water

50 62.5 MW Chilled Water 24 MW

0

0 50 100 H [MW]

Answer to (d)(1). The flue gas heating option will probably be more expensive

than heating with steam. The temperature levels in the problem do not suggest

exploitation of furnace heat.

Answer to (d)(2). The appropriate placement of a heat pump is across the Process

Pinch, so the pump will transfer heat from below to above the Pinch. In this

problem it is technically impractical to install a heat pump with water or steam

as a working fluid, since this would require operation at a slight vacuum (the

Pinch is at 170C/140C). The relatively small duties that could be achieved

with this technique indicate that investment in a heat pump here is likely to be

economically unattractive.

Answer to (d)(3). Introducing a new steam level at a steam temperature of 165C

is a good option that can be exploited by passing the steam (taken at 200C)

through a steam turbine. This has the potential of generating additional power

on the site. The steam would be returned to the process at the level of 165C

(the process heating requirement is at 150C, which makes this arrangement

feasible). If the site could use additional power, then this option may reduce the

overall utility cost by providing on-site power generation.

Examples and Case Studies 247

10.5.1 Water Pinch Technology: First Problem

Problem 8: Task Assignment

A certain process contains several water-processing operations, as

shown in Table 10.22.

tration, C, against contaminant load, m [kg/h], [ppm].

(b) Calculate the minimum water flow rate for maximum reuse.

Problem 8: Solutions

Answer to (a). The individual limiting water profiles are graphed in Figure 10.29.

Figure 10.30 shows the result when these water profiles are combined.

rate [kg/h] rate [t/h]

1 2 0 100 20

2 5 50 100 100

3 30 50 800 40

4 4 400 800 10

800

4

C [ppm]

400

3

100 2

1

50

0 2 7 37 41 40

m [kg/h]

248 Chapter Ten

Minimum flow rate

Pinch concentration

9 [kg/h] 9 10 6 kg

100 [ppm] 100 h

9 10 3 t

h 90 [t/h]

100

800

C [ppm]

400

ly L t

Pinch

S upp ate 90 flowrate

t e r w r

Wa m flo

i n imu

100 (M

0

1 9 21 41

m [kg/h]

Examples and Case Studies 249

Problem 9: Task Assignment

Using the stream data from Problem 8, suggest a design of a network

for maximum reuse of water. The limiting CC is shown in

Figure 10.32.

Problem 9: Solution

Cut off the pockets of the limiting CC. The minimum water requirement for

each pocket can then be defined for each region (Figure 10.33). This enables one

to propose the design strategy shown in Figure 10.34.

Following the design strategy, set up a design grid as shown in Figure 10.35.

Connect streams to water mains and merge operations that cross boundaries;

see Figure 10.36.

Remove intermediate water mains where this is appropriate (Figure 10.37).

Then connect operations directly, as shown in Figure 10.38. Finally, Figure 10.39

illustrates (in conventional flowsheet form) one of the possible resulting designs

of the water system.

800

C [ppm]

400

100

0

1 9 21 41

m [kg/h]

250 Chapter Ten

800

C [ppm]

45.7 t/h

400

100

50

90 t/h

m [kg/h]

FIGURE 10.33 Water Pinch diagram used to target the minimum water flow

rate in Problem 9.

800 ppm

C [ppm]

400

100

45.7 t/h 100 ppm

50 44.3 t/h100 ppm

90 t/h

90 t/h

m [kg/h]

Examples and Case Studies 251

Flowrate 90 t/h 45.7 t/h 0 t/h

required for the F.W. 100 ppm 800 ppm

interval

20 t/h 1

100 t/h 2

Limiting 50 ppm

Flowrates

40 t/h 3 3

50 ppm

10 t/h 4

400 ppm

Wastewater (90-90) (90-45.7) (45.7-0)

0 t/h 44.3 t/h 45.7 t/h

100 ppm 800 ppm

Water mains

F.W. 100 ppm 800 ppm

20 t/h 20 t/h

1

50 t/h

100 t/h 2

20 t/h 40 t/h

40 t/h 3

20 t/h

5.7 t/h

10 t/h 4

FIGURE 10.36 Streams are connected with the water mains (Problem 9).

252 Chapter Ten

F.W. 100 ppm 800 ppm

20 t/h 20 t/h

1

2

20 t/h 40 t/h

40 t/h 3

20 t/h

5.7 t/h

10 t/h 4

sources and sinks (Problem 9).

90 t/h

F.W.

20 t/h 20 t/h

1

50 t/h

100 t/h 2

20 t/h

40 t/h 3

20 t/h

5.7 t/h

10 t/h 4

44.3 t/h

20 t/h

20 t/h 40 t/h

Operation 1 Operation 3

F.W. Wastewater

90 t/h 90 t/h

50 t/h 5.7 t/h

Operation 2 Operation 4

44.3 t/h

CHAPTER 11

Industrial Applications

and Case Studies

T

his chapter provides an overview of selected implementations

of the Process Integration methodology for various industrial

case studies. The presentation is somewhat condensed because

of space limitations; more information is available in the works

cited.

In this case study, the Heat Exchanger Network (HEN) of a Fluid

Catalytic Cracking (FCC) unit process consisted of a main column

and a gas concentration section (Al Riyami, Kleme, and Perry, 2001).

The stream data was made up of 23 hot streams and 11 cold streams.

The associated cost and economic data required for the analysis were

specified by the refinery owners. Incremental area efficiency was

used for the targeting stage of the retrofit design. This was carried

out using the Network Pinch method (Asante, 1996; Asante and Zhu,

1997), which consists of a diagnosis stage and an optimization stage.

In the diagnosis stage, a few promising retrofit steps were generated

using the UMIST (now the University of Manchester) software

package SPRINT (2009). This software was also used to optimize the

initial design by trading off capital cost against energy savings. The

design options were then compared and evaluated, followed by

the final retrofit design proposed for final inspection.

The existing Tmin of the process was identified as 24C, and the

hot utility consumption of the process was 46.055 MW; the area

efficiency of the existing design was 0.805. The potential for energy

saving was then derived from the resultant Composite Curves, which

are shown in Figure 11.1. As seen in the figure, the Composite Curves

are relatively wide apart except in the area around the Pinch.

The capital cost was estimated under the assumption that the

retrofit distribution of area would be the same as that for the

existing network. The resulting optimum minimum temperature

approach was found to be about 11.5C for incremental and about

253

254 Chapter Eleven

400

Tmin = 11.50C

Temperature [C]

300

200

100

0

1 2

Enthalpy [MW]

FIGURE 11.1 Composite Curves of FCC process with optimum Tmin (after

Al Riyami, Kleme, and Perry, 2001).

was found to be 0.804. This value indicated that the existing design

was using the area reasonably efficiently. Even so, there was still room

for improvement. Since the constant- targeting produced a

conservative estimate, an incremental value of 1.0 was used to set

the retrofit target, which yielded potential for energy savings of

about 12.117 MW. Analysis of the existing design revealed that there

were four process-to-process heat exchangers that transferred heat

across the Process Pinch (from above to below the Pinch). It was

also found that some heaters supplied utility heat to process streams

below the Pinch and that some coolers removed heat from process

streams above the Pinch. These energy violations of the established

Pinch rules generated the scope of the projects possible energy

savings.

The retrofit design using the Network Pinch method allowed a

limit to be set on the structures energy recovery. The next stage

consisted of testing a set of modifications that would result in higher

levels of energy recovery in the process. The increase in energy

recovery would come at the expense of increased heat exchange area.

Therefore, any benefit in energy cost reduction had to be weighed

against the additional capital cost associated with increasing that

area. A number of promising design solutions were generated, which

were then optimized for minimum total cost. The four designs

identified each involved a payback period of less than two years, yet

increases in energy prices rendered the actual payback period closer

to one year. The final design chosen for the retrofit situation was the

one with the shortest payback period and the least additional area

required; it is shown in Figure 11.2.

In this design option, four new heat exchangers were added and

one existing exchanger (number 1 in Figure 11.2) was removed, since

its duty approached zero. In reality the exchanger equipment item

Industrial Applications and Case Studies 255

4

5

224 366

6 H1

60 150.8 366

C 3 H2

238 760.5 kW 293

12 H3

52 123 293

C 43 H4

38 481 kW 207.8 210.5 238

54.9 74

C 10 8 7 41 H5

1247 kW

50 109 224

C 2 H6

108 1235 kW 158

42 H7

40 104

C H8

294.25 kW 83.3

38 97.17

C 40 H9

12459 kW

38 62.87 68.11

C 44 H10

3324 kW

38 62.05 63.11

C C H11

1902 kW

38 2704 kW 52.5

C H12

38 244.5 kW 51.5

C H13

46 73.9 kW

96.41

C H14

53267 kW

38 46

C H15

359.6 kW 74.3 96.3

38 48.92 193

C C 11 9 H16

563 kW

46 653.3 kW 53

C H17

14125 kW

38 46

C H18

195.83 kW

38 46.44

C H19

856.4 kW

38 53

C H20

32.3 kW

60 67.24

C H21

7336 kW

38 60

C H22

206 kW

38 102

C H23

1815.6 kW 88.2 110 134.8 171.7

C1

60

40 2 3 4 H 385

22841 kW

100 3802 kW 2381 kW 1803 kW 4282.8 kW

C2 195

5

58.32 3697.5 kW 192.6

C3 8

38 10198.9 kW 55

C4 10

38 1413.6 kW 50

C5 11

8787 kW

184.5 872.25 kW 193

C6 12

6

78.56 95 41479.7 kW 101.7

C7 9 7

118.4 119.3 13291 kW

118 4258.8 kW 213.7 kW 124

C8 43 41 H

C9 44

1051.9 kW

51.16 54.3

C10 H

101.78 763.9 kW

95.40 102

C11 42 H

5669.4 kW 200 kW

FIGURE 11.2 Grid diagram of the chosen retrofit option (after Al Riyami,

Kleme, and Perry, 2001). Exchanger number 1 from the original network is

removed and now used repiped as new added area (exchanger number 42);

exchanger number 10 is only repiped.

additional exchangers called for by the new design by repiping either

the hot or cold stream (so in this case, exchanger number 1 becomes

exchanger 42 in Figure 11.2), thus further reducing the additional area

256 Chapter Eleven

required. All the modifications carried out in the diagnosis stage had

the effect of moving heat from below to above the Network Pinch. The

final retrofit design produced energy savings of 8.955 MW, about

74 percent of the potential of the design. The annual utility cost savings

amounted to $2,388,600, a 27 percent decrease in the utility bill. Since

the modified HEN required an investment of $3,758,420, the payback

period was less than 19 months.

The study demonstrated that a combination of targeting and

Pinch Technology (process Pinch and Network Pinch) can yield

substantial improvements in an existing HEN and thereby reduce the

total network cost. The employed method can recognize bottlenecks

in an existing system, and then generate a series of potential

improvements by searching for modifications capable of shifting heat

from below to above the Network Pinch. It was found that targeting

for maximum energy savings at each potential modification usually

produces a good trade-off between area and energy cost.

Distillation System

A crude-oil preheating system retrofit problem was studied by

Seikova, Varbanov, and Ivanova (1999). The objective of the study was

to develop a topology modification proposal for the refinery owners

that would feature better energy efficiency than the existing network

under a flexibility requirementnamely, that the distillation plant

be able to work with several alternative feedstocks. The preheating

system of the distillation plant was analyzed, and several topology

modifications were evaluated.

The retrofit aspect of Heat Integration was first considered

systematically by Tjoe and Linnhoff (1986), who proposed a framework

for retrofit assessment and targeting. Later, Asante and Zhu (1997)

developed the Network Pinch methodology. This hybrid approach,

which combines heuristics and Mathematical Programming (MPR),

was implemented through an iterative procedure whose two main

stages are to identify the heat transfer bottleneck of the given HEN

and then to implement modifications for overcoming it.

Another important Process Integration issue involves accounting

for different plant operating conditions stemming from the market

availability of various feedstocks. The case study reported in Saboo,

Morari, and Woodcock (1985) was based on the HEN retrofit

methodologies just described and combined the energy recovery

improvement with handling of the flexibility requirements that are

common to real-life processing plants.

The crude-oil distillation unit (Figure 11.3) consists of a main

distillation column (with 6 m3/h capacity) for fractionating crude oil

into four products: light distillate, heavy gasoline, diesel fuel, and

atmospheric residue. The preheating HEN recovers heat from the

Industrial Applications and Case Studies 257

Gases

Water

Light Distillate Steam

Steam

Crude Oil

Steam

Atmospheric residue

Diesel Oil

Lower pumparound

Heavy Distillate

bring the final crude oil temperature to the columns required target,

the remaining heat duty is provided by the combustion of

noncondensed light gases and of additional natural gas.

The process features the following characteristics: (1) temperature-

dependent heat capacities; (2) continuous partial phase change of the

crude oil; (3) temperature variations, which are small owing to

specifications of the distillation process; and (4) large variations of

the hot stream flow rates due to changing stockfeed composition.

The three utility sources available for HEN operation are light

gases (as a furnace fuel for higher temperature levels), steam at 1 atm

(1.01325 bar), and cooling water at 1835C. The data used to estimate

the utility cost are $68.74/kWy for furnace heating, $103.10/kWy for

steam, and $30.00/kWy for cooling water. The following area cost

law was used:

crude-oil types were selected to represent this range. Feed 1 is a light

crude oil; it contains a significant amount of the lightest fractions,

which require a large amount of cooling in the condensers. However,

the relatively small amount of the heavy fraction means that less

reboiler heating is required. Feed 2 is medium crude oil and feed 3 is

heavy crude oil. These two feedstocks are characterized by a relatively

greater amount of the heavy fractions (diesel and atmospheric

residue), which means that more heat is available for recovery in the

higher temperature range. It was assumed that, over a years worth of

operation, the plant uses equal amounts of the three feedstocks. Each

alternative feedstock is associated with a unique operating point.

258 Chapter Eleven

large temperature intervalfrom 20C to 310C. The preheating

process consists of three phases. In the first phase, the crude oil is

heated from the starting temperature to its bubble point. The next

phase involves continuous partial evaporation in heat exchangers.

The third and hottest phase is heating to the specified column entry

temperature of 310C (this heating is performed in the furnace). The

upper pump-around features a large enthalpy change split between

condensation and subcooling segments. None of the other process

streams has phase transitions.

The existing HEN for the preheating train is shown in Figure 11.4.

In order to initialize the retrofit procedure, a Pinch Analysis of the

stream data for Tmin = 10C was carried out. The results of this

analysis are given in Table 11.1.

For the first two feeds, the Pinch is located at the temperature

boundary of the crude oil bubble point. For feed 3, the Pinch is

located at a temperature that is close to the beginning of the phase

change. Note that each operating point offers a different potential

for heat recovery.

After looking at the existing network and thermodynamic

targets, one can suggest several retrofit modifications. First, the hot

utility usage below the Pinch leads to extremely poor heat recovery

as well as to substantially increased utility cooling and thus to using

considerable amounts (3038 m3/h) of cooling water. Taking a look

at the coolers in the temperature interval of heater H1 (Figure 11.4),

one can see that the sum of the cooler loads at the outlets of streams

H3, H4, and H5 offers enough heat supply to satisfy the heating

demand in the interval from 20C to 60C. This analysis implies that

steam heating can be eliminated through repiping of the coolers as

recovery matches. It can also be seen that the availability of heat

recovery varies in response to the different temperature levels

characteristic of each type of crude-oil feed. This effect is most

significant for the atmospheric residue stream H5, whose heat

capacity flow rate varies from 0.98 to 1.47 kW/C. To handle this

variation in heat availability temperatures, a cascade of recovery

location hot utility cold utility heat recovery

[C] [kW] [kW] [kW]

Feed 1 110120 556.25 560.26 610.64

Feed 2 110120 481.31 428.24 664.33

Feed 3 86.1996.19 507.34 395.66 696.45

TABLE 11.1 Pinch Analysis Results for Operating Points of the Three

Preheating Phases

Industrial Applications and Case Studies 259

CP [kW/C] CPH11 CPH12

CPH11 = 17.68 28 65.5 94.2

C2 C1 H1

CPH12 = 17.68 101.9 kW 507.5 kW

40.0 70.0 111.3

0.38 C3 1 H2

11.3 kW

60.0 75.3 186.0

1.66 2 H3

25.4 kW C4

40.0 129.8 197.0

0.84 C5 3 H4

75.8 kW

100.0 140.3 296.3

0.98 C6 4 H5

39.7 kW

CPC11 = 2.92 20.0 60.0 65.3 119.8 130.3 158.7 310.0

CPC12 = 5.40 C1 H1 1 2 3 4 H2

CPC13 = 4.23 CPC11 CPC1 CPC1

1 2

101.7 kW 11.9 kW 183.6 kW 56.7 kW 153.5 kW 677.3 kW 3

(a) Initial HEN Feed 1

C2 C1 H1

CPH12 = 2.086 55.7 kW 436.7 kW

40.0 70.0 111.6

0.28 C3 1 H2

8.6 kW

60.0 74.2 186.3

1.25 C4 2 H3

17.70 kW

40.0 122.1 196.7

0.87 C5 3 H4

21.4 kW

294.6

H5

54.8 kW

CPC11 = 2.87 20.0 60.0 64.2 112.1 127.3 180.5 310.0

CPC12 = 4.25 C1 H1 1 2 3 4 H2

CPC13 = 4.49 114.8 kW 11.9 kW 140.4 kW 64.7 kW 231.1 kW 582.8 kW

(b) Initial HEN Feed 2

CPH12 = 2.27 C2 C1 H1

46.6 kW 469.9 kW

0.27 40.0 70.0 111.2

C3 1 H2

8.0 kW

C4 2 H3

15.1 kW

0.88 40.0 109.4 197.1

C5 3 H4

61.2 kW

1.41 100.0 127.6 294.8

C6 4 H5

38.7 kW

CPC11 = 3.43 63.2 99.4 117.6 172.2

20.0 60.0 310.0

CPC12 = 4.25 C1 H1 1 2 3 4 H2

CPC13 = 4.46

137.0 kW 10.9 kW 129.2 kW 77.7 kW 235.1 kW 614.1 kW

mentioned previously, which would form three loops for internal

heat transfer.

Another important observation concerns the amount of recovery

from streams H2 to H5. Streams H3 and H4 feature close supply

temperatures. For all three operating points, and especially for that

of feed 1, H3 has substantially greater heat capacity flow rate than

260 Chapter Eleven

This fact and the heat exchanger cascade already planned suggest

another topology modification: resequence matches 2 and 3 on

stream C1. Such resequencing would yield a larger driving force in

both matches, although some load would then be shifted to the lower

temperature segment of stream C1.

The network was modified according to the strategy and changes

suggested by the analysis. The result is shown in Figure 11.5. Cooler

C3, on the hot stream with the lowest target temperature, is left at its

old location. This, together with the introduced three loops, will

account for variations in composition of the feedstock. When there is

a larger total heat supply from streams H2 to H5, any heat surplus is

removed from the system through cooler C3.

Comparing the initial and new topologies reveals that a

significant improvement in energy efficiency has been achieved.

The heat recovery fraction is increased from 67.03 to 91.22 percent

for feed 1, from 73.37 to 89.77 percent for feed 2, and from 74.12 to

77.86 percent for feed 3. Although the relative increase for feed 3 is

lower, its net increase in heat recovery is the same as that of

feed 2.

Based on the simulations for the initial and modified networks,

the heat transfer area for each match is determined as recorded in

Table 11.2 (where H1 is 4 6.26 m 2). The area requirement of the

recovery matches EC4 (new), 3, EC5, and EC6 increase after the

topology change. Note that only the increase of EC4 is significant,

which is due to the heat load shifted from match 2. Another effect

of the modification is that the four eliminated steam heating

Match Initial Retrofit Initial Retrofit Initial Retrofit

C1 67.69 67.69 61.62 61.62 75.56 75.56

C2 43.42 43.42 30.51 30.51 31.30 31.30

1 10.83 4.74 8.12 3.44 7.39 4.80

C3 3.45 2.00 2.62 1.61 2.42 0.00

2 98.04 38.43 69.49 27.21 57.75 31.57

C4EC4 4.95 59.41 3.49 39.97 3.02 12.99

3 10.67 11.83 11.85 12.30 12.97 15.64

C5EC5 7.99 11.43 2.37 11.17 7.47 5.64

4 14.58 14.44 24.98 23.84 24.20 24.42

C6EC6 1.81 9.68 2.54 12.01 1.89 2.48

H1 21.33 20.98 25.04

Industrial Applications and Case Studies 261

28.0 101.85 kW 65.5 94.2

C2 C1 H1

52.1 507.47 kW 111.3

40.0

C3 1 H2

60.0

EC4 2 H3

EC5 3 H4

EC6 4 H5

27.7 49.4 89.4 105.3 120.0 137.0 165.0

20.0 310.0

C1 1 EC5 EC4 EC6 3 2 4 H2

C2 C1 H1

53.2 436.7 kW

40.0 111.6

C3 1 H2

EC4 2 H3

EC5 3 H4

EC6 4 H5

20.0 25.8 49.1 81.6 107.0 124.0 138.0 187.0 310.0

C1 1 EC5 EC4 EC6 3 2 4 H2

C2 C1 H1

39.8 436.72 kW

40.0 111.2

C3 1 H2

0 kW 99.4 186.0

60.0

EC4 2 H3

EC5 3 H4

EC6 4 H5

25.5 31.9 45.2 54.5 89.4 112.7 169.1 310.0

20.0

C1 1 EC5 EC4 EC6 3 2 4 H2

matches 3, EC5, and EC6. Thus the investment, which is estimated

to be $42,330, is only for the additional area of EC4. Overall, the

modified heat recovery system yields a significant reduction in

energy cost. The total sum of savings after the retrofit is estimated

262 Chapter Eleven

of about 2.4 years (or less if energy prices increase).

This case study of a combined problem in heat recovery and

process flexibility within a distillation units preheating system

combined advanced Pinch-based retrofit methodologies with

additional heuristic rules. The result was that the flexibility goals

were better met by the introduction of loops in the network topology.

In short, the new and modified paths enabled a significant improvement

in heat recovery. Another benefit was that the more expensive steam

heating was partially replaced by slightly increased furnace duty in

the case of higher heat demand. This interesting trade-off between

two hot utilities yields an economic benefit owing to the lower cost

of furnace fuel.

a Citrus Juice Plant

This case study describes a water and wastewater minimization

project designed for a citrus plant located in Argentina (Thevendiraraj

et al., 2003); the study proceeded by applying the Water Pinch

technology (Wang and Smith, 1994). Citrus juiceprocessing plants

consume large quantities of freshwater. The principal objective of

this study was to reduce both the freshwater consumed and the

wastewater produced by the plant. The citrus processing plant housed

the following processes: selection and cleaning, juice extraction, juice

treatment, emulsion treatment, and peel treatment.

Water minimization was achieved by maximizing water reuse

and identifying regeneration opportunities. Water-using operations

were represented by the maximum inlet and outlet contaminant

concentrations, which are functions of equipment corrosion,

fouling limitations, minimum mass transfer driving forces, and

limiting water flow rate through an operation. Targets determined

the minimum freshwater requirement using the Limiting

Composite Curve for the water design network. The graphical

Pinch methods that are based on single contaminants can be

extended to cover multiple contaminants. When dealing with a

number of operations, multiple contaminants, and multiple water

sources, the problem becomes more complex and so algorithms

using the basic Pinch principles have been developed for solving

by MPR-based methods (see Smith, 2005).

The study began with data extraction: each stream had to be

characterized by its contaminant concentration, inlet and outlet

concentration levels, and limiting flow rate through each operation.

The data were provided in a schematic flow diagram of the citrus

plant that incorporated a simplified water distribution network and

the mass balance of the plants water streams. Eleven freshwater-

using operations were identified. Chemical oxygen demand (COD)

Industrial Applications and Case Studies 263

was chosen as a component to proxy for all contaminants for two

reasons: first, COD measures the most significant contaminant

load in the majority of water streams; and second, COD exhibits

significantly high values.

The overall mass balance is closed by using the assumption

that evaporation losses from the steam system amount to 1 t/h

(otherwise, there is an inconsistency of 1 t/h of water that must

be accounted for). The extracted data on contaminant concentrations

and water flow rate made it possible to establish the amount of

water gained or lost by each operation in the process. The total

mass load picked up by the freshwater through each operation was

then calculated. The eleven water-using operationstogether with

the water flow rates entering and leaving each operationcan be

represented in the form of a simplified water network, as shown in

Figure 11.6.

The freshwater COD concentration level for the plant is 30 ppm.

There was an existing water reuse between processes currently in

the plant, and these reuse streams were left unchanged. The

simplified water network presented in Figure 11.6 shows the

freshwater-using operations with existing water reuse streams built

in to each identified operation. The current total freshwater consumed

and wastewater generated by this citrus plant were, respectively,

240.3 and 246.1 t/h.

The existing water network provided a base starting point for the

Water Pinch Analysis. The freshwater target was evaluated by using

the Composite Curves. The maximum concentration levels were

based on the constraints and limitations dictated by process

conditions and requirements. This data were represented in the

WATER software tool (2005) with identified constraints. The process

restrictions on water type permissible for each operation indicated

that operations 2, 4, 5, and 10 can use only freshwater as input. Hence,

the minimum freshwater required by plant operations was

164.4 t/h.

Operation 1 is a batch process for which the analysis assumed

freshwater must always be available. This increases the plants total

minimum freshwater requirement [t/h] to 164.9 = 164.4 + 0.5 for

operation 1. The current total freshwater feed to the plant is 240.3 t/h.

These figures can be used to calculate the maximum theoretical

freshwater reduction (MTFWR) that is achievable:

MTFWR 100 31.4 percent

240.3

The Water Pinch Analysis was then carried out for the existing

water network with maximum concentration levels. The overall

freshwater target was calculated using the maximum reuse analysis.

Figure 11.7 shows the modified water network represented as a

conventional diagram, and the Limiting Composite Curve is plotted

264 Chapter Eleven

Packing (1)

Loss to

Loss to Process 6.265 t/h 1 t/h Environment

7.265 t/h 0 t/h

Treatment Plant

Boiler

(Potable) (2) Blowdown+

1.000 t/h Condensate

Lossess Neglected

26.15 t/h

26.15 t/h

Selection / Cleaning (3)

APV Condenser & Screen 2&3 &

Green Tank (4) Vincent Press

90.0 t/h

240.318 t/h 3.385 t/h 3.385 t/h 246.054 t/h

Vacuum Pump (5)

Waste Water

Fresh Water to

1.62 t/h from plant

Plant 1.62 t/h

Screen Taca-Taca (6)

4.8 t/h

Finisher (7)

Condensate from

1.5 t/h

distiller stream

7.322 t/h Condensate from

juice vapour

Gain from Process

8.822 t/h

Distiller (8)

39.6 t/h 28.65 t/h

Screen 1 (9)

Distiller Condenser &

Washing Spiral 1 (10)

3.6 t/h 4.076 t/h

Centrifugal (11)

Total Raw Water 240.318 t/h Total Gain from Process 37.391 t/h

Total Waste Water 246.054 t/h Total Losses to Process 31.655 t/h

Difference 5.736 t/h Difference 5.736 t/h

Error % 0.00

in Figure 11.8. The freshwater target was compared with the current

freshwater consumption in order to assess the plants overall potential

for minimizing water and wastewater. The redesigned network

yielded a freshwater demand of 169.3 t/h and a wastewater flow rate

of 175.1 t/h. Although this amounts to a substantial reduction in both

water and wastewater, the design includes reuse of certain streams

that will require treatment. As already described, the analysis was

based on COD as the pseudocontaminant in reuse streams. However,

these streams may well contain other contaminants (e.g., solid waste,

Industrial Applications and Case Studies 265

0.5 t/h 0.5 t/h

Operation 1

8.96 t/h

Loss to Process, 6.265 t/h

Recycle, 23.7 t/h

Operation 2 Operation 3

Operation 7 Operation 6

0.48 t/h 4.01 t/h

Operation 11

3.53 t/h

Loss to Process,

10.95 t/h

47.56 t/h 31.83 t/h 39.60 t/h 28.60 t/h

Operation 4 Operation 9

169.3 t/h 28.09 t/h 4.38 t/h 175.05 t/h

90.00 t/h 113.72 t/h

Operation 4a

Operation 5

28.09 t/h

1.00 t/h 8.82 t/h

Operation 2a/8

Loss to Process

8.64 t/h

15.84 t/h 7.20 t/h

Operation 10

being used for other processes. Hence additional design options

must be developed for dealing with specific process requirements,

operating conditions, and suitability standards for water reuse.

The necessary design refinements may be achieved by introducing

further constraints to potential reuse streams and by utilizing the

maximum water reuse analysis to obtain optimal designs that use

even less freshwater while meeting all process operating conditions

and restrictions. The regeneration reuse analysis can also be used to

explore additional design options incorporating the reuse of

regenerated water in some operations. Such analysis is based on

installation of a treatment unit to regenerate wastewater (by gravity

settling, filtration, membranes, activated carbon, biological agents,

etc.). This modification would further reduce plant levels of

freshwater and wastewater, and it was also evaluated with the

WATER software.

266 Chapter Eleven

Limiting Composite Curves

30.0

25.0

Concentration [ppb]

20.0

15.0

10.0

5.0

Mass [t/h]

Limiting CC Water Supply Line

both the maximum reuse analysis and regeneration reuse analysis

(Figure 11.9). Design options A and B were based on the maximum

reuse analysis, and both achieve a freshwater consumption of 188 t/h;

this is a reduction of about 22 percent from the actual freshwater

consumption of 240 t/h. Because of process limitations and

restrictions, there was no further scope for water reuse. Yet results

from the diagnostic stage indicated that a 31 percent reduction in

freshwater and wastewater was possible. This additional extent is

achievable by regenerating wastewater and then reusing it in other

operations. Design options C and D are based on the regeneration

reuse analysis; both result in total freshwater consumption of 169

t/h, which amounts to a 30 percent reductionthat is, nearly all of

the theoretical maximum predicted by Eq. (11.2), the expression for

MTFWR.

The reduction in freshwater use was achieved by rerouting water

streams, a modification that requires new pipes. Design option A

requires five new pipes; this is fewer than options B or D, which

require seven pipes, and option C, which requires nine new pipes.

New piping will affect investment cost. The five new pipes identified

for design option A are also required in the other options for the

same function and at similar flow rates. Therefore, this option

requires the least investment and option C the most. Furthermore,

design options C and D each require additional investment cost for a

Industrial Applications and Case Studies 267

regeneration unit that is not part of options A and B. These results

are summarized in Figure 11.9.

Design options A and D are the most attractive ones in terms of

maximizing both reuse and regeneration reuse. Option A results in a

smaller reduction in freshwater use but at a lower investment cost

than option D, which results in a larger reduction in freshwater use

but at a much higher investment cost. Wastewater reductions are

proportional to freshwater reductions, with corresponding reductions

in wastewater treatment costs for each of the options. The cost

analysis carried out for design option A indicates attractive financial

returns for this low-investment option, whose payback period is only

0.14 years. The outlet water quality of operation 3 required further

analysis, so a complete cost evaluation of design option D was not

possible. Fully evaluating this option would require additional

detailed studies to identify the regeneration process type required

and its associated costs.

The heat energy of the reuse water streams proposed in the

design options was reviewed to ensure that stream temperatures at

the inlet of operations remained unchanged. Citrus plant managers

reported that nearly all of the process operations occur at ambient

temperatures; the only exception was operation 8, which produces

wastewater at 90C. (This particular waste stream is highly

contaminated, which imposes some limitations.) All the water reuse

streams proposed by the four design options are at appropriate

temperatures, so they should not have a thermal effect on operations.

The overall hot and cold utility requirements of the plant would not

be affected by the changes proposed in the design options.

With its existing water network, the plant consumes 240.3 t/h of

freshwater and generates 246.1 t/h of wastewater. The proposed

design options offer a 30 percent and a 22 percent reduction in

freshwater consumption and wastewater generation. For a practical

project, the number of modifications is limited. The maximum water

FRESHWATER SAVINGS

Number of

New Pipes % Feed Water

Theoretical Freshwater Reduction Limit, 31% Reduction

Required

30

30 30

25

22 22 20

10 9

5 7 7

5

A B C D

Design Options

268 Chapter Eleven

regeneration reuse design requires seven new pipes. The reuse

analysis predicted a short payback period, but the regeneration

analysis was not definitive for the reasons described previously. In

sum, a Water Pinch Analysis for this citrus plant demonstrated that

the consumption of freshwater could be reduced by as much as 30

percent with low investment and few changes to the existing

plant.

Drink Industries

Many studies have employed Pinch Technology (and its associated

Heat Integration Analysis) in the food-processing industry. This

industry has a far different thermodynamic profile than that of the

refining and petrochemical industries. The food-processing industry

is characterized by process streams of relatively low temperature

(normally 120140C), a small number of hot streams, low boiling-

point evaporation of food solutions, considerable deposition of scale

in evaporators and heat recovery systems, and seasonal operation.

However, a number of studies have also found that the application

of Pinch Technology and Heat Integration is hampered by particular

aspects of the food-processing systems. These aspects include direct

steam heating, difficulties in cleaning heat exchanger surfaces, and

high utility temperatures. Despite these drawbacks, the benefits that

can be obtained by applying the Pinch Technology (e.g., optimized

heat recovery systems and reduced energy consumption) far outweigh

the difficulties of performing the studies. There are also other

advantages that can be realized by technological improvements:

reduced deposition of scale due to reduced utility temperatures, self-

regulation of heat processes, and reduced emissions.

A case study of the production of refined sunflower oil (Kleme,

Kimenov, and Nenov, 1998) exemplifies the benefits of process

analysis based on Pinch Technology and Heat Integration. The

process studied operated with a minimal temperature difference of

65C at the Process Pinch. The external heating required by the

system was provided by two types of hot utilitiesDowtherm steam

and water steam; the required external cooling was provided by two

cold utilitiescooling water and chilled water. The analysis proposed

that heat recovery be increased and that the minimum temperature

difference be reduced to 814C. The increase in heat recovery

(provided by a reduction in the minimum driving force for the

process) entailed an increase in the heat transfer area, but this was

more than offset by reductions in the hot and cold utility requirements.

A further benefit of the analysis was a reduction in the number of

utilities needed: eliminating water steam and cooling water

considerably simplified the overall design.

Industrial Applications and Case Studies 269

A case study of a whisky distillery by Smith and Linnhoff (1988;

see also Caddet, 1994) provides another example of how Pinch

Technology and Heat Integration can reduce energy use and increase

energy efficiency. In this case it was found that steam was being used

below the Process Pinch, resulting in an overall increase in utility

usage. The steam was related to use of a heat pump, so the steam

used below the Process Pinch was eliminated by reducing the size of

that heat pump. Although the steam now had to be used for process

heating above the Process Pinch, the overall energy costs were

reduced owing to the reduction in compressor duty.

Another study involving a whisky distillery was made by Kemp

(2007). The hot utility requirement for the process examined was

8 MW, and the Process Pinch was at 95C. The main hot utility

requirements were steam for the distillation system and hot air for

the drying system. Kemp showed that the form of the Grand

Composite Curve and the temperature of the Pinch could be

exploited for heat pumping, and he also suggested that the process

would benefit from the introduction of a Combined Heat and Power

(CHP) scheme for improved Process Integration and energy

efficiency. The sites power demand was 12 MW, so there were two

possibilities for providing both the power and the heat demands

from the same utility system. First, a gas turbine that produced

12 MW of power would also supply about 30 MW of high-grade heat

from the exhaust. The second option involved the use of back-

pressure steam turbines, but these would produce nearly 100 MW,

much more than what was required. Another advantage of a gas

turbine was that its exhaust could be used for drying purposes.

Figure 11.10 shows the final configuration of the utility system

matched to the GCC. Most of the necessary heat was provided by the

gas turbine exhaust and the existing thermo compressors. The

existing package boilers were used to provide steam for the thermo

compressors. The efficiency of this part of the system was increased

by using waste heat from below the Pinch to preheat the boiler feed

water. Waste heat boilers driven by the exhaust from the gas turbine

provided additional steam.

Many processes in the food and drink industry make use of

chilling and refrigeration systems. Pinch Technology and Heat

Integration have also been used to increase the efficiency of these

systems. For example, Fritzson and Berntsson (2006) studied a

Swedish slaughtering and meat-processing plant. Their analysis of

the plants subambient temperature section employed the method

proposed by Linnhoff and Dhole (1992) for low-temperature process

changes that involve shaftwork targeting. Figure 11.11 shows the

plants Exergy Grand Composite Curve (EGCC). There is a large gap

between the EGCC and the utility curve, which indicates low

efficiency in the use of available shaftwork. Improvements in this

area could result in a 15 percent reduction in energy demand.

270 Chapter Eleven

500

Shifted temperature [C]

st

400

au

xh

ee

bin

tur

300

s

Ga

200 Waste

Site steam heat

Driver steam boiler

ex rbi s

ha ne

tu Ga

t

us

Heat Cooling

pump water

10 0 10 20 30 40 50 60

Net heat flow [MW]

process GCC (after Kemp, 2007).

0

0.05 EGCC

Utility curve

0.10

0.15

0.20

0.25

c [1]

FIGURE 11.11 Exergy Grand Composite Curve for meat-processing plant (after

Fritzson and Berntsson, 2006).

process by adjusting the loads on the subambient utilities, first

maximizing the load on the highest-temperature (10C) subambient

utility. After this, the load on each lower-temperature utility was

Industrial Applications and Case Studies 271

maximized. The reduction process began at the highest level (the

highest temperature), since this utility requirement can be satisfied

at a lower cost than refrigeration at lower temperatures. The modified

system was then modeled and simulated in HYSYS, which showed a

5 percent reduction in the shaftwork required.

However, these results still made for a relatively poor fit between

the EGCC and the utility curve. To reduce the gap further, it was

suggested that the level of the highest-temperature refrigeration

utility be increased from 10C to 3C and then the loads be

readjusted as before; the result is shown in Figure 11.12. This modified

configuration yielded a 10 percent reduction in the shaftwork

requirement. Other temperature changes were suggested to reduce

further the shaftwork requirements, but these changes were found to

be less cost-effective.

Varbanov and colleagues (2005) demonstrated the synthesis of a

utility system (CHP network) of an industrial Total Site by applying

a combination of targeting and Mathematical Programming

techniques. Figure 11.13 shows the heating and cooling demands of

the chemical site studied. There are two operating scenariosfor

winter and summer, with different prices for power and fuel. The

basic data for the problem are listed in Tables 11.3 and 11.4.

Generic estimates of the corresponding coefficients were specified

for the boiler performance (field-erected boilers are slightly more

efficient than packaged units), and the capital cost estimates were

obtained online (Boiler Cost 2003). These estimates are a function of

boiler capacity and steam pressure. The performance and cost of gas

0

0.05 EGCC

Utility curve

0.10

0.15

0.20

0.25

c [1]

FIGURE 11.12 Area between the EGCC and the utility curve is further

decreased by changing the temperature of the first refrigeration level from

10C to 3C (after Fritzson and Berntsson, 2006).

272 Chapter Eleven

Winter Summer

T [C] T [C]

300 HP level candidates 300 HP level candidates

209.15 209.15

200 200

198.89 198.89

LP level candidates

40 20 0 20 40 60 80 60 40 20 0 20 40 60 80

H [MW] H [MW]

FIGURE 11.13 Total Site Profiles and candidate steam pressure levels for three

steam mains of an industrial chemical site (the Heat Source Profile for summer

provides more heat than that for winter).

Winter Summer

Fraction of year 0.55 0.45

Price of cooling water [$/t] 0.0185 0.0212

Site power demand [MW] 25 31

Maximum export allowed [MW] 10 10

Price of power (import and export) [$/MWh] 20 30

Interest rate [%] 8

Life of plant [y] 10

Capital installation factor 4.8

TCW [C] 10

TBFW [C] 120

P VHP [bar(a)] 90

TVHP,SH [C] 200

TVP [C] 50

Natural gas price [$/MWh] 9.369

Distillate oil price [$/MWh] 10.734

Fuel gas price [$/MWh] 4.984

Fuel oil price [$/MWh] 6.226

Site

Industrial Applications and Case Studies 273

turbines (ranging from 11.7 to 85.4 MW) were estimated using data

from Gas Turbine World (2001).

The preliminary Total Site targets estimated minimum utility

demand of about 80 t/h for the winter scenario, leading to the choice

of a field-erected boiler for stand-alone steam generation. A gas

turbine with a Heat Recovery Steam Generator (HRSG) is also selected.

For this case, three steam mains (headers) were considered: one for

very-high-pressure (VHP) steam generated by the boiler and by the

HRSG, and two more intermediate steam mains. The VHP header

properties were specified by the problem definition and are given in

Table 11.4. Hence, locations for the two other headers still have to be

determined. Figure 11.13 illustrates the partitioning of the candidate

steam levels, whose ranges are highlighted. For the upper steam

main, this graph was used to identify three candidate levels with

respective saturation temperatures of 270.68C, 209.15C, and

198.89C; the only candidate for the lower steam main there has a

saturation temperature of 130.51C. The problem superstructure is

illustrated in Figure 11.14.

Optimizing and reducing the superstructure yields the flowsheet

shown in Figure 11.15. It features two fired steam boilers and one

steam turbine sized to the steam flow for the winter period. This

flowsheet features relatively low on-site power generation and a

significant amount of power import. The main reasons behind this

approach are that power is cheap and capital is expensive; these

factors preclude a better utilization of the steam systems potential

cogeneration. This design also involves a significant amount of CO2

emissions.

Gas turbines

gt01 gt01

Process heating

Process cooling demands

demands HRSG HRSG b01 b02 Winter Summer

Winter Summer

VHP: 90 bara, 503.35 C 3.550 3.175

12.000 16.885 MW MW

MW MW Canditate

hdr01: st01 hdr01: st02 9.038 5.956

pressure levels

7.700 13.383 MW MW

hdr01 tb01

MW MW

tb02 1.635 1.501

tb03 MW MW

1.300 2.877

MW MW

hdr02: st01 hdr02: st02 50.591 43.777

15.500 18.136 MW MW

MW MW hdr02 : tb01

14.660 10.632

8.210 10.632 MW MW

CT: st01 CT: st02

MW MW

COND: 0.1235 bara

Condensate return

To cooling water

DA Make-up water

274 Chapter Eleven

emissions were evaluated. The sensitivity analysis addressed four

cases, as summarized in Table 11.5. Case 1 is the base case and thus

represents the conditions discussed so far. The other cases gradually

increase the price of power and fuel as well as the penalties on

emissions. The optimal utility system flowsheet for case 3 is shown

in Figure 11.16. Case 4 also replaces fuel oil with a biofuel, which is

assumed to have zero CO2 emissions.

Analysis of these emission reduction options leads to the

following conclusions: (1) Increasing the system efficiency is the

cheapest option for CO2 abatement, but it has a relatively limited

scope. (2) The next economic option for this particular problem is to

close the carbon cycle by using biofuels; in general, however, CO2

capture and sequestration could also be considered.

Capacity: 51.334 t/h Capacity: 70.000 t/h

Winter: 51.334 t/h b02 Winter: 41.990 t/h Winter Summer

Winter Summer b01

Summer: 45.049 t/h Summer: 0 t/h

6.289 t/h 5.625 t/h

12.000 MW 16.885 MW

VHP: 90 bara, 503.35 C 3.550 MW 3.175 MW

7.700 MW 13.383 MW Capacity: 52.409 t/h hdr : st POWER 18.907 t/h 13.210 t/h

01 01

Winter: 52.409 t/h Winter: 4.767 MW

Summer: 1.941 MW 9.038 MW 5.956 MW

Summer: 26.213 t/h

Winter: 15.717 t/h

1.300 MW 2.877 MW

Summer: 0.001 t/h

TSAT=198.89 C; P=15.19 bara 1.635 MW 1.501 MW

33.139 t/h 52.304 t/h

(tb03)

Winter: 16.645 t/h 84.620 t/h 77.960 t/h

25.240 t/h 29.533 t/h

Summer: 0.558 t/h

50.591 MW 43.777 MW

15.500 MW 18.136 MW TSAT=130.51 C (tb01); P=2.74 bara

Winter: 18.769 t/h 23.116 t/h 17.597 t/h

Summer: 12.493 t/h

8.210 MW 10.632 MW 14.660 MW 10.632 MW

Winter: 114.025 t/h

COND: 0.1235 bara Summer: 101.182 t/h

To cooling water

Condensate return

DA

Winter: 136.671 t/h Winter: 22.646 t/h from processes

Summer: 116.369 t/h Summer: 15.187 t/h

To the stream Make-up

generators Winter: 154.502 t/h water Power import [MW]: 21.1 (Winter); 30.0 (Summer)

Summer: 128.237 t/h Total annualized cost: 13.124 106 $/y

CO2 [$/t] SOx [$/t] NOx [$/t] [$/MWh]

1 0 0 0 20.00 30.00

2 0 0 0 30.00 45.00

3 40 500 1000 40.00 60.00

4 40 500 1000 40.00 60.00

Parameters

Industrial Applications and Case Studies 275

Power

gt01

Capacity: 33.917 MW Capacity: 28.570 t/h

Winter: 30.085 MW Winter: 95.230 t/h Winter: 0.000 t/h

Summer: 33.916 MW Summer: 45.022 t/h Summer: 0.000 t/h

Winter Summer b02 Winter Summer

HRSG

6.289 t/h 5.625 t/h

12.000 MW 16.885 MW

VHP: 90 bara, 503.35C 3.550 MW 3.175 MW

Capacity: 52.017 t/h hdr01 : st01

Winter: 52.017 t/h Winter: 7.722 MW

9.038 MW 5.956 MW

Summer: 26.023 t/h Summer: 1.942 MW

Winter: 18.016 t/h

1.300 MW 2.877 MW

Summer: 0.164 t/h

TSAT = 198.89C; P=15.19 bara (tb03)

1.635 MW 1.501 MW

33.139 t/h 52.304 t/h

Power

Capacity: 20.000 t/h hdr02 : st01 84.271 t/h 77.935 t/h

25.240 t/h 29.533 t/h Winter: 0.733 MW

Winter: 13.997 t/h

Summer: 0.000 MW

Summer: 0.000 t/h 50.591 MW 43.777 MW

15.500 MW 18.136 MW Winter: 4.904 t/h

Summer: 0.555 t/h

TSAT = 130.51C; P=2.74 bara (tb01) 24.272 t/h 17.597 t/h

8.210 MW 10.632 MW Winter: 19.869 t/h 14.660 MW 10.632 MW

Summer: 12.491 t/h

COND: 0.1235 bara

To cooling water

Winter: 114.832 t/h Condensate return

DA

Winter: 137.590 t/h Summer: 101.157 t/h from processes

Winter: 22.758 t/h

Summer: 116.343 t/h Summer: 15.119 t/h

To the stream Make-up

generators Winter: 156.466 t/h water Power export [MW]: 10.0 (Winter); 4.4 (Summer)

Summer: 128.209 t/h Total annualized cost: 23.060 106 $/y

Building Complexes

Herrera, Islas, and Arriola (2003) studied a hospital complex that

included an institute, a general hospital, a regional laundry center, a

sports center, and some other public buildings. The use of diesel fuel

represented 75 percent of its total energy consumption and 68 percent

of its total energy cost, which was $396,131 in 1999.

In the hospital complex, the heat demand is met by producing

steam in boilers fueled by high-price diesel fuel. There is no heat

recovery between the existing heat sources and heat sinks. The hot

streams were identified as the soiled soapy water from the laundry

and the flow of condensed steam not recovered in the condensation

network. The stream data are presented in Table 11.6.

For this hospital complex, the amount of external heating required

(i.e., the hot utility target) is 388.64 kW, which can be seen on the hot

and cold Composite Curves in Figure 11.17. This plot was employed

to determine what temperature levels of the utilities would satisfy

this requirement. The heating utility target of 388.64 kW translates

to an annual energy requirement of 12.26 TJ/y.

The amount of heating provided is actually 625.28 kW, which

represents the heat services that are currently transferred to the

complex. This figure represents a potential for energy savings of

38 percent, which is equivalent to an annual reduction in diesel fuel

of 246,000 liters (worth about $100,000). To reduce heating energy

demands to the targeted value, the Heat Integration Analysis

276 Chapter Eleven

[C] [C] [kW] [kW/C]

1 Hot Soapy water 85 40 23.70 0.53

2 Hot Condensed 80 40 96.32 2.41

steam

3 Cold Laundry 25 55 17.60 0.59

sanitary water

4 Cold Laundry 55 85 77.27 2.58

5 Cold Boiler feed 30 60 7.13 0.24

water

6 Cold Sanitary water 25 60 77.12 2.20

7 Cold Sterilization 30 121 12.50 0.14

8 Cold Swimming pool 25 28 151.67 50.56

water

9 Cold Cooking 30 100 59.63 0.85

10 Cold Heating 18 25 100.82 14.40

11 Cold Bedpan 21 121 4.94 0.05

washers

TABLE 11.6 Process Stream Data of Hospital Complex (Herrera, Islas, and

Arriola, 2003)

T [C]

120 Hot CC

Cold CC

100

80

60

40

PINCH

20

H [MW]

FIGURE 11.17 Process Composite Curves for hospital complex (Tmin = 20C).

network. Two are needed in the laundry to cover part of the heat

demand, a third is needed in the machinery rooms that help to heat

boiler feed water, and a fourth is needed in the condensation tank

area that heats the sanitary water. The analysis could be refined

Industrial Applications and Case Studies 277

further by considering several other issues, such as fouling, pressure

drop, and variation in the heat demand.

In the global market, optimal management control of a company is

necessary for survival. Such control involves a series of effective

strategic decisions that are concerned with various aspects of the

supply chainfor example, decisions regarding the products

themselves as well as the location of production facilities and

distribution centers. When the subject is the development or

operation of highly complex business processes (e.g., supply

chains, value chains), computer-aided decisions are preferred.

A conventional approach for systematically evaluating decision

alternatives is Mathematical Programming, although a daily

management problem is usually not resolved via MPR methods.

Even when the MPR can be constructed for the problem, it is still

difficult to verify whether the mathematical formulation of relevant

decision alternatives is sufficiently accurate and complete to

identify the optimal solution. When the mathematical model is not

generated systematically, the models chances of embodying the

optimal solution are very low.

This case study involves determining the optimal design of a

supply chain. Here, the purpose of the supply chain is to meet a given

volume demand for commodity C at location L1. Three options are

considered: produce commodity C at location L1; produce commodity

C at location L2 and then transport it to location L1; or some combination

of these. Production requires the availability of part A and part B at

the same location. Part A is available at location L3 and, to a limited

extent, at location L4; it can be transported to location L1 or to

location L2. Part B is available at location L2 and can be transported to

location L1. The list of potential activities is given in Table 11.7.

PL1 Production L1 A and B at L1 C at L1

PL2 Production L2 A and B at L2 C at L2

TAL3L1 Transportation From L3 to L1 A at L3 A at L1

TAL3L2 Transportation From L3 to L2 A at L3 A at L2

TAL4L1 Transportation From L4 to L1 A at L4 A at L1

TAL4L2 Transportation From L4 to L2 A at L4 A at L2

TBL2L1 Transportation From L2 to L1 B at L2 B at L1

TCL2L1 Transportation From L2 to L1 C at L2 C at L1

278 Chapter Eleven

(activating entities) and effects (resulting entities), the Maximal

Structure Generation (MSG) algorithm then produces the maximal

structure (Friedler et al., 1993). Next, when applied to the maximal

structure so derived, the Solution Structures Generation (SSG)

algorithm (Friedler et al., 1995) enumerates 15 combinatorially

feasible business process structures for the problem.

To determine the optimal business practice, the following

quantitative information is provided in addition to the case studys

15 structural alternatives. The required volume of the demand is

20,000 pieces annually. Producing one piece of commodity C requires

the availability of one piece of part A and one of part B. At most 5000

pieces of part A are available at location L4 for 230 each. An

unlimited number of part A can be purchased at location L3 for 250

each, and an unlimited number of part B can be purchased at

location L2 for 310 each.

The cost of an activity depends on its volume. To estimate the

increase in the cost of an activity as a function of its volume, costs

are given for handling 1000 and 2000 pieces per year. If a linear cost

function with a fixed charge is adopted to estimate the cost of the

activities with different volumes, then this function is determined

by the fixed charge and the proportionality constant. Table 11.8

summarizes the cost of processing 1000 versus 2000 pieces per

annum as well as the parameters of the fixed-charge linear cost

function for each activity.

The optimal business process, as determined by the Accelerated

Branch-and-Bound (ABB) algorithm (Friedler et al., 1996), corresponds

to the activities given in Table 11.9 whose annual cost totals

ID 1000 pcs/y 2000 pcs/y Fixed charge Proportionality

[/y] [/y] [/y] constant

[/pcs/y]

PL1 8,000 10,000 6,000 2

PL2 9,000 10,000 8,000 1

TAL3L1 4,000 8,000 0 4

TAL3L2 10,000 20,000 0 10

TAL4L1 12,000 24,000 0 12

TAL4L2 2,000 4,000 0 2

TBL2L1 10,000 20,000 0 10

TCL2L1 14,000 28,000 0 14

Industrial Applications and Case Studies 279

Volume [pcs/y]

PL1 15,000 20,000

PL2 5,000 20,000

TAL3L1 15,000 15,000

TAL3L2 15,000

TAL4L1 5,000

TAL4L2 5,000 5,000

TBL2L1 15,000 20,000

TCL2L1 5,000 20,000

Total cost [/y] 11,439,000 11,466,000 11,568,000

Processes

of 11,466,000, and the third-best business process has a total annual

cost of 11,568,000.

Paint production usually consists of three major operations: grinding

and dispersion, mixing and coloring, and discharging and packaging.

Paints and coatings are typically produced in batches. They are made

in stationary and portable equipment units such as high-speed

dispersion mixers, rotary batch mixers, blenders, sand mills, and

tanks. The raw materials are solvents, resins, pigments, and additives

that include inorganic and organic chemicals. Paint manufacturing

does not usually involve chemical reactions between the raw

materials, so the finished product consists of a mixture of the different

raw materials. Several dozens of products are produced at the

manufacturing site, so the corresponding scheduling problem is

bound to be highly complex.

The S-graph framework of batch scheduling (see Chapter 7) has

been extended to solve complex paint production problems (Adonyi

et al., 2008). Changeover time is defined for any equipment unit that

requires cleaning. Traditionally, minimizing makespan (total time to

completion) is the criterion used when assigning equipment units to

tasks and scheduling the tasks. Such schedules maximize the

production systems efficiency, but they may lead to unnecessarily

high levels of waste generation. Thus, determining which task

schedule minimizes cleaning cost will require that the problems

objective function be modified. Now, rather than minimizing

makespan as in the original problem, the reformulation seeks to

280 Chapter Eleven

minimize the cleaning cost. This change in criterion has only a minor

effect on the solution procedure, so an effective solver for the original

problem is also useful for the reformulated problem.

Twenty-three equipment units, E1 through E23, are available to

generate six products, A through F. The changeover time is 70 minutes

for equipment units E6, E7, E8, and E9 but 100 minutes for equipment

units E1 through E5 and E10 through E20. All other changeover times

are presumed to be zero. The number of batches to be produced is

given in Table 11.10.

Cleaning the equipment units is a costly operation that involves

many pollutants. The minimal makespan schedule contains 11

cleaning operations, which are denoted by the dotted changeover

arcs on its S-graph; see Figure 11.18.

The cleaning cost of the solution with minimal makespan is

$14,000. In contrast, the solution based on minimizing the cost

involves four (rather than 11) cleaning operations and only $3,500 in

cleaning cost; its makespan is 6,910 minutes. If the cleaning cost is

limited to reach $5,500, then the corresponding makespan is reduced

to 6,700 minutes.

Product A B C D E F

Number of batches 3 5 1 3 9 3

E1 E6 E11 E22 97 E4 E6 E10 E21 109

A 0 E

0 0 0 0

0

60 310 120 540 40 120 90 720 110

E1 E6 E11 E22 98 E4 E6 E12 E21

A E

0 0 0 70 0 0 0 0

0

60 310 120 540 40 120 60 720 111

E1 E6 E11 E22 99 E4 E6 E10 E21

70 0 A E

0 0 0 0

60 240 120 540 40 300 90 720 112

E1 E7 E15 E22 100 E4 E6 E12 E21

B E

0 0 0 0 0 0

0

70 40 300 60

60 E7 240 E19 120 E22 540 101 B E6 E21 720 113

E1 E4 E10

E

0 0 100 0 0 0 0

0

60 240 120 40 300 90 720

E1 E7 E11 E22 540 102 E4 E6 E12 E21 114

B E

0 0 0 0 0

0 0

60 240 60 540 40 300 60 720

E1 E7 E13 E22 103 E4 E6 E10 E21 115

B E

0 0 0 0 0 0

100 0

60 240 120 540 40 300 90 720

E1 E7 E15 E22 104 E4 E6 E12 E21 116

B E

70 0 0 0 0

100 40 300 90 720

E2 60 E8 120 E16 50 E22 540 105 E4 E6 E16 E21 117

C 100 E

0

60 240 90 540 40 240 120 720

E3 E7 E18 E22 106 E5 E7 E19 E23 118

D F

0 0 0 0 0 0

60 240 90 540 40 240 60 720

E3 E7 E14 E22 107 E5 E7 E20 E23 119

D 0 F

0 0 100 0 0

0

60 240 60 720 40 240 120 720

E3 E7 E20 E23 108 E5 E7 E15 E23 120

D F

70 100

FIGURE 11.18 Schedule graph of the solution that minimizes makespan (after

Adonyi et al., 2008).

CHAPTER 12

Typical Pitfalls and

How to Avoid Them

P

rocess Integration (PI) has proven to be a powerful optimization

tool for designing processes that are energy-efficient,

environmentally friendly, and sustainable. The methodology

provides clear insight into the design process, but this direct

simplicity is sometimes misunderstood by potential users. Once

proper data are made available, the procedure generates an excellent

lead for the design process. But just as with all optimization tools,

potential pitfalls when using PI include improper formulation of the

problem and incorrect data extraction.

Even when the most efficient and well-developed methodology

is used to solve an optimization problem with high precision, the

results may be suspicious unless we have been solving the right

problem. In other words, has the problem been formulated in a way

that closely reflects the real process under consideration and

(especially) has the correct data been extracted? Negative answers to

these questions explain such published statements as: Pinch

technology [or process integration] did not work for this problem.

When these problems are revisited it usually becomes obvious that

the PI methodology is not at fault; rather, an inexperienced or

overconfident user is typically the cause.

Therefore, this chapter is devoted to various mistakes that a

designer might unwittingly make. The most basic issue concerns

how one starts a PI-based project and how it is run. Kemp (2007)

summarized the key steps listed below, which have been further

developed based on the authors experience. These steps are

specifically related to Heat Integration; however, they apply with

only small adjustments to mass, water, and other integration as

well.

way is to closely liaise with the process designer and/or plant

manager, especially if the plant is already operating.

281

282 Chapter Twelve

the designed process flowsheet data and calculations and/

or on measurements taken from the operating plant (if the

study is for a retrofit).

3. Select the streams. This is a critical step and, as will be shown

in this chapter, not as straightforward as it may seem.

4. Remove all the existing units related to the PI analysis. For

Heat Integration, remove all heat-transferring units, for

mass water integration remove all water interconnections

(the pipes). This step is also criticalwithout it, the

optimized design would not differ from the initial design.

Section 4.2.4 provides an example illustrating this activity.

5. Extract the stream data for the PI analysis. Different data are

relevant for each PI analysis type. For HI heat loads and

temperatures are extracted.

6. Make a qualified initial guess for the Tmin value; this value

can be adjusted later at various stages of the design

optimization.

7. Perform the Pinch analysis: obtain the Pinch temperatures

and the utility targets.

8. Design the initial (heat exchanger) network using the

criterion of maximizing energy recovery.

9. Check for a cross-Pinch transfer and for inappropriate

placement of utilities.

10. Check for proper placement of reactors, separation columns,

heat engines, and heat pumps.

11. Investigate the potential for further modifying the process

in order to minimize energy consumption and reduce

capital costs. Investigate the potential benefits of applying

the plus-minus principle (see Chapter 4) and the KHSH and

KCSC principles (see Figure 4.44).

12. Investigate the potential for integration with other processes

that is, Total Site Analysis.

13. Consider the implications of pressure drop (trade-offs between

heat savings and extra energy for pumping) and the physical

layout (capital cost of heat exchangers and/or piping).

14. Make the preselection of heat exchange equipment and

perform the preliminary costing. Provision should be made

for variations in the future price of energy.

15. Make the first optimization run of the predesign plant or

site, and make adjustments to Tmin.

Typical Pit falls and How to Avoid Them 283

16. Based on the optimization, extract adjusted data and return

to step 7. Perform an additional loop (or loops) while

screening and scoping for potential simplifications.

17. Consider real plant constraints; these include safety,

technology limitations, controllability, operability, flexibility,

availability, and maintainability.

18. Pay attention to start-up and shutdown of the process; some

early designs for highly integrated plants had problems in

this area.

19. Run a second optimization for the final tuning accounting

for the information added during steps 16 to 18. If necessary,

return to any appropriate previous step for adjustment.

20. The design is now ready for detailing. However, optimization

is a never-ending procedure, and designs may need to be

modified in response to changes in operating conditions

(e.g., plant capacity) or the economic environment (e.g., tax

policy; prices for energy, materials, and production).

As emphasized previously, data extraction is a crucial step. Bodo

Linnhoff presented one of his last plenary lectures (Linnhoff and

Akinradewo, 1998) on the automated interface between simulation

and integration. This has been a substantial step toward data

extraction for PI software tools. The plenary has been fairly

comprehensive and suggested the way forward for this important

task. The problem received increased interest following this lecture,

and several software packages now offer support for solving it.

Nonetheless, more work in this area is needed to satisfy the

requirements of routine industrial applications.

In their SuperTarget and Pinch Express software packages,

Linnhoff March (1998) included procedures for automatic extraction

of data for Heat Integration. Even so, thermal data, which involve the

stream heating and cooling information and utilities information,

are the most critical data required for Pinch Analysis. There are

several possibilities for extracting the thermal data from a given heat

and material balance. This must be done carefully, as poor data

extraction can easily lead to missed opportunities for improved

process design. In extreme cases, poor data extraction can falsely

present the existing process flow-sheet as optimal in terms of energy

efficiency. If the data extraction accepts all the features of the existing

flow-sheet then there will be no scope for improvement. If it does not

accept any features of the existing flow-sheet then Pinch Analysis

284 Chapter Twelve

accepts only the critical sections of the plant which cannot be

changed. Data extraction skill develops with increased experience in

the application of Pinch Technology (Linnhoff March, 1998).

Since the release of these packages, the methodology has

developed further and more attempts have been made to extract

data automatically. However, experience and following the proper

rules remain valuable assets. Basic questions to ask include the

following:

2. How precise must the data be at each step?

3. How can considerable changes in specific heat capacities be

handled?

4. What rules and guidelines must be followed to extract data

properly?

5. How can the heat loads, heat capacities, and temperatures of

an extracted stream be calculated?

6. How soft are the data in a plant or process flowsheet?

7. How can capital costs and operating costs be estimated?

To those unfamiliar with PI, identifying a stream seems fairly

straightforward. In fact, many considerations are involved, and

accounting for them properly is key to setting up the problem. First,

we need not consider any stream that neither gains nor provides heat;

no data needs to be extracted from a stream with identical supply

and target temperatures and enthalpies. (Of course, in the absence of

perfect insulation, every stream loses or gains some heat; in many

cases, however, these small amounts of losses and gains can be

neglected.) If we do not extract data from such streams, the problem

is considerably simplified.

There are also streams that, for one reason or another, should not

be included in the PI problemfor example, streams that are remote

and streams that should not be altered for safety, product purity, and

operational reasons and for other (mostly practical) considerations.

Finally, Heat Integration deals with heat flows, which can be carried

not only by a pipe line but also by radiation or conduction.

Consider the example depicted in Figure 12.1. This example was

introduced, along with some data extraction rules, by Linnhoff and

colleagues (1982) and was later modified for use in many follow-up

books (e.g., Smith, 1995; Smith, 2005; Kemp, 2007) and in many courses

based on UMIST (later the University of Manchester) teaching

materials. The figure shows part of a flowsheet in which the feed

stream is heated to 45C by recuperated heat in a heat exchanger and

then enters a processing unit. After leaving this unit, the stream is

Typical Pit falls and How to Avoid Them 285

150

160

H3

190

120 40

Reactor 80 10

45 45

H2 H1

Feed

140 Unit 140

heated further by two heat exchangers and then enters a reactor. The

reactor requires the feed stream to be at 165C. The question is: How

many streams should be extracted?

1. Three streams: 1045C, 4580C, and 80165C

2. Two streams: 1045C and 45165C

3. One stream: 10165C

there will again be three heat exchangers with the same heat transfer

duties as before. This case accounts for claims of no process

improvement by some PI critics. Option 2 presents more degrees of

freedom: the first heat exchanger would be the same, but the other

could be modified. Extracting two streams would be appropriate for

cases when the processing unit requires a feed temperature of close

to 80C.

Option 3 provides the most degrees of freedom and scope for

improvement, but with this design the processing unit feed could be

at any temperature between the 10C supply and the reactor target of

165C. If the processing unit is, say, a filter (as assumed by Smith,

2005), then there would probably be some restriction on the filter

supply temperature to ensure proper operation of the filter. If the

processing unit is storage (as assumed by Linnhoff et al., 1982), then

the supply temperature might be restricted to a different range

(depending, e.g., on whether liquid or gas is stored). This simple

example demonstrates that choosing the right stream from which to

extract data cannot be a fully automatic process. Making this choice

requires assessments related to specific processing units and

performance requirements for the plant.

There are frequently questions about required data precision, and a

common excuse for not applying PI analysis is that the data on a

plant is not sufficiently precise. However, the process of applying PI

286 Chapter Twelve

during the course of several design loops. At the beginning there are

no specifications for the heat-transferring units, and neither are the

feed temperatures (and several other important factors) fully fixed.

At this stage, then, extremely precise data is not needed. It is important

to recognize that PI and the initial optimization are more about

screening and scoping than about detailed design. In seeking

potential energy savings, it is the general direction in which

optimization should proceed that is the initial concern. If the result

of this first step is that a 15 percent savings in energy is possible, then

this figure is sufficient to confirm the design approach and it doesnt

much matter if the precise figure is actually 13 or 17 percent.

It is in the regions close to the Pinch that the data should be as

precise as possible (Linnhoff et al., 1982). Also, it is best to remain

inside the Composite Curves in the plot of temperature versus

enthalpy. At the start of data extraction, we might have only a vague

idea of where (and at what temperature) the Process Pinch will occur;

also, the Composite Curves are based solely on data extraction.

Therefore, data extraction must necessarily start from rough

assessments and then be corrected step by step.

Capacities Be Handled?

Further analysis of the flowsheet in Figure 12.1 reveals that phase

changes are very likely to occur when the temperature increases

from 10C to 165C. In general, Cp varies with temperature, but latent

heat is also a determining factor. Clearly, using a constant value for

Cp would not be sufficiently realistic.

A segmentation technique has been developed to deal with this

problem. This technique is used, for example, in the software tools

SPRINT (2009) and STAR (2009). The software tools treat the segments

as individual streams that are combined to form input and output

streams. Data extraction is affected by how many segments are used

and their boundary temperatures (see Figure 12.2). Increasing the

number of segments naturally increases the complexity, so this

T 1 T 2 T 3 T 4

? ? ? ?

H H H H

Typical Pit falls and How to Avoid Them 287

number should be minimized for industrial problems involving

many streams.

Followed to Extract Data Properly?

As mentioned in Section 12.1.1, data extraction rules were first

introduced by Linnhoff et al. (1982) and frequently used with some

modifications thereafter (Smith, 1995; CPI, 2004 and 2005; Smith,

2005; Kemp, 2007). Most of this work is related to Heat Integration,

but the principles apply as well to mass (water) integration. These

rules are reviewed briefly in this section.

When two or more streams of different temperatures are mixed,

this nonisothermal mixing constitutes a heat exchange with

degradation of the higher temperature. In some cases, such mixing

can also cause cross-Pinch transfer problemsas in Figure 12.3(a),

stream mixing extracted as

(a) three streams and (b) two

streams.

110

70 40 75 35

110 160

40 70

35 75

(b) 160

110

70 40 75 35

35 160

35 70

288 Chapter Twelve

where three streams are extracted. The correct data extraction for this

case involves two streams, as shown in Figure 12.3(b).

General guidelines for data extraction may be summarized as

follows:

they should not be neglected when streams (mass and heat

flow in pipes) are long or subject to varied temperatures. In

such cases, the solution is to introduce hypothetical coolers (or

heaters) that represent the heat loss.

2. Extracting utilities. The utilities should never be extracted from

the existing plant or flowsheet, for then the solution would

likely arrive at the same utility values and perhaps neglect

some options that would be more efficient. This rule of thumb

applies especially to cases where utilities can be generated

on-site and thereby (at least partially) replace costly existing

utilities. In this connection it should be remembered that

steam is not always a utilitysometimes it is also a process

stream (e.g., stripping steam in separation columns). Process

streams should remain in place and not be removed.

3. Generating utilities. The Heat Integration analysis may indicate

some valuable options for using otherwise wasted heat or

cold to generate utilities. The Grand Composite Curves can

be used to locate such options. However, when extracting

data it must be recognized that steam requires the boiler

feedwater (BFW) to be heated, water to be evaporated, and

the steam to be superheated; see Figure 12.4. Many mistakes

have been caused by designers who simply matched the

Specified Steam

Conditions

BFW Conditions

preheating, evaporation, and superheating.

Typical Pit falls and How to Avoid Them 289

steam generation line without making provisions for

preheating and superheating.

4. Extracting at the effective temperature. In some cases a stream

cannot be extracted directly because it has to still be used by

a related process. For example, a hot stream should be

extracted at temperatures at which the heat becomes

available. An example is given in Smith (2005, p. 433) for a

reactor using a quench liquid.

5. Forced and prohibited matches. In almost any process there exist

matches that are necessary for technological reasonsthe

forced matchesas well as those, such as hot and cold stream

matches in a heat exchanger, that should be prohibited (e.g., to

prevent contamination of one of the streams). In manual design

these constraints have to be observed by the designer, but

software tools usually offer this option. If not then the

constraints can be secured by an appropriate penalty or bonus

(as applies) in the objective function used for the optimization.

6. Keeping streams separate only when necessary. If streams can be

merged then it may be possible to eliminate some heat-

exchanging units. For example, streams that leave the plant

to be treated as wastewater often have some heat content that

can be utilized.

Temperatures of an Extracted Stream

Be Calculated?

Once a stream has been extracted, the next problem is calculating the

heat-related data. There are standard engineering procedures

available for the running plants as the measurements with the

following data reconciliation (Kleme, Lucha, and Vaek, 1979;

Minet et al., 2001; VALI III User Guide, 2003). The other option is to

develop a flowsheeting simulation model (Kleme, 1977). An

overview of flowsheeting and balancing simulators was given in

Chapter 9. If a plant is being designed, some data could be also

extracted from the process flow diagram (PFD). But all those options

consume time and resources, so in the early design stages (when the

process structure is still under development and likely to be changed

as a result of PI analysis), it is reasonable and easier to use a simplified

approach based on the extracted data. Such an approach is

demonstrated in Figure 12.5 for a part of the flowsheet from

Figure 12.1. The CPs of the stream segments are assumed constant

and calculated from the temperatures and the duties given in the

flowsheet. Experience has indicated that the resulting rough

preliminary data are sufficient and can later be made more precise

by one or more of the procedures listed previously.

290 Chapter Twelve

T

160

80

45

150 10

160

H

H3 H1 H2 H3

190

120 40

Reactor

80 45 45 10

H2 H1

Feed

140 Unit 140

FIGURE 12.5 Obtaining rough data from flowsheet heat loads and

temperatures.

Process Flowsheet?

Distinguishing soft data from hard data is one of the most important

aspects of data extraction. Inexperienced persons are usually trying

to stick the temperatures shown in the PFD, extract those temperatures,

and then perform the PI analysis. However, this approach usually

ends up overlooking many opportunities. A better approach is to

question every temperature, discuss each one with the process

engineer (or plant designer or plant manager), and thereby establish

which temperatures are critical (the hard data) while the rest (the

soft data) can be in some way compromised. In practice, most data

are at least a little soft, and designers can use this fact to their

advantage. Typically, streams that are leaving the plant (see

Figure 12.6) are characterized by soft data and thus are suitable for

optimization via the plus-minus principle (Figure 12.7). Data softness

is closely related to changing conditions and to a designs flexibility,

operability, and resilience.

Costs Be Estimated?

The need to find cost data arises when the appropriate Tmin (which

should be close to the optimum) is being selected. The optimum Tmin

depends strongly on economic parameters, and its value is important

for both grassroots design and retrofit. Estimating capital costs is

usually a time-consuming procedure. However, it is possible to use

Typical Pit falls and How to Avoid Them 291

70 20

Cooler Cooler

T T Increase (+)

Hot Stream

Decrease ()

Hot Stream

Utility Target Utility Target

H H

T T Decrease ()

Increase (+) Cold Stream

Cold Stream

Utility Target Utility Target

H H

targets by using properly extracted soft data.

when little is known about the types of heat transfer units to be used,

the ultimate design, the materials required, or the temperature,

pressure, and composition of streams. The estimates so derived will

usually suffice until the final detailed design cost date, at which time

information is obtained from selected manufacturers. Note, however,

that equipment cost may vary regionally and may also be related to

market conditions (e.g., a large customer can secure discounts, and

prices fall during a recession).

It is much more difficult to establish operating cost, which is

affected by labor, taxation, and so forth but is mainly a function of

energy cost. Here the most obvious problemand greatest potential

pitfallis using the current price of energy. (It is possible to find

292 Chapter Twelve

many publications and projects where this rule was not followed.) It

is better to use the anticipated average energy price for the life span

of the plant or, in the case of a retrofit, for the payback period. The

problem then becomes one of estimating this future energy price for

periods that may be as long as five or ten years. It has been shown

(see, e.g., Kleme and Bulatov, 2001; Donnelly, Kleme, and Perry,

2005) that even the forecasts of highly qualified experts are frequently

inaccurate. One of the obvious potential approaches is to use the

scenarios and target the most flexible design that would provide a

balanced optimum for various situations.

Demand and Supply

The integration of renewable energy sources was discussed in

Chapter 6. Renewable resources are usually available on a smaller

scale and are often distributed over a certain region. Their availability

(with the exception of biomass) varies significantly with time and

location. This variability is due to changing weather and geographic

conditions. The energy demands (heating, cooling, and power) of

sites also vary significantly with time of the day and period of the

year. These variations in the supply and demand of renewables can

be predicted in part, and some of the variation is fairly regularfor

instance, day versus night in predominantly cloudless areas for solar

energy. The availability of wind-generated energy can be less

predictable. One approach to dealing with these problems is the

advanced PI technique that employs time as an additional problem

dimension. A basic methodology along these lines (involving Time

Slices and Time Average Composite Curves) was developed for the

Heat Integration of batch processes (Kleme et al., 1994; Kemp and

Deakin, 1998). This methodology was recently revisited by Foo,

Chew, and Lee (2008).

This methodology has also been extended to the Heat Integration

of renewables. Important steps in this direction were reported by

Perry, Kleme, and Bulatov (2008) and Varbanov and Kleme (2010).

Dealing with variation and fluctuation brought another complexity

into data extraction. Data should be collected for all time slices that

increase the complexity. Especially, for each case, it is necessary to

choose the time horizon for the analysis and number of time slices.

This is a fast-developing field, so it is advisable to monitor recently

published research papers and conference presentations.

It has been assumed that all analyzed and optimized processes

operate in a steady state. Many industrial processes do operate in this

Typical Pit falls and How to Avoid Them 293

fashion, and a common control task is to maintain such processes in a

steady state. However, there are also many situations in which the

working regime must be changed. This occurs not only at start-up

and shutdown but also in response to changes in operating conditions

(e.g., outside temperatures) and production conditions (e.g., capacity,

volume). Problems involving such variation are generally addressed

in one of two ways: (1) accommodating various scenarios via a design

objective of minimizing sensitivity to the possible changes or

(2) designing to optimize the total systems overall dynamic

performance. The second option is much more complicated and

requires more time and resources; hence it is used only when

substantial deviations from steady-state operations are anticipated.

After data extraction, the most important aspect of PI analysis and

optimization is the correct interpretation of results. These results are

typically presented in the form of a printout generated by some

software tool; in most cases, the software output is in the form of

a grid diagram (e.g., STAR, 2009) or PFD supported by printout

tablesas with ASPEN (AspenTech, 2009a, 2009b, 2009c) and

gPROMS (PSE, 2009). In order to minimize typos and misinterpretation,

many software tools feature an interface that facilitates the transfer

of data files. This technology is now fairly well developed.

The most challenging aspect of data interpretation is assessing

the results in terms of possible further development or correction of

the important process features. The interpretation will depend on the

extent to which the factors mentioned previously (i.e., data uncertainty,

data softness, flexibility, operability, controllability, safety, availability,

and maintenance) have already been incorporated into the data

extraction. If the data extraction did not reflect these considerations,

then they must be addressed during the data interpretation. At this

stage, the close collaboration of a team that includes the main involved

professionals is highly recommended.

Designers are strongly advised not to stick with just one solution

but rather to explore different potential scenarios associated with

various operating conditions and then test their designs sensitivity

to the possible variations. For screening and scoping it is helpful to

use all types of Composite Curves as well as information (if provided

by software) on which streams are contributing to specific parts of

the curves.

Even when a sustainable and near-optimum design is developed,

it must still be put into practice. This involves selling the

294 Chapter Twelve

investors, and contractors. This was a big problem when PI was just

beginning, and great strides in this area were made by UMIST, Bodo

Linnhoff, and his company Linnhoff March in the 1980s and 1990s.

PI has since proven itself and gained in popularity, so decision

makers have become more receptive.

Much of the situations improvement is due to multinational

companies that have incorporated PI into their design and operational

practice. Because the methodology has become widespread, it is not

possible to list all of these companies. However, among the pioneers

were members of the UMIST and, after the merger, the University of

Manchester Process Integration Research Consortium: Air Products,

Aspen, BASF, Bayer, BOC, BP, Canmet, Degussa, EDF, Engineers

India, Exxon Mobil, Hydro, IFP, JGC, KBC, Mitsubishi Chemical

Corporation, MOL, MW Kellogg, Petrobras, Petroleum Research

Centre, Petrom, Petronas, Saudi Aramco, Shell, Sinopec, Technip,

Total, UOP, and Vito. These firms were joined in the consortium by

several universities, including University POLITEHNICA Bucharest

and Petronas Technological University.

There have also been strong supporters of PI in the United

States, both at universities and in the industry; some of them are

listed in Chapter 13. A major goal remains a close collaboration and

smooth joint effort among PI specialists and the designers,

managers, and owners (or contractors) of processing plants. A

projects best chance of success is when all these stakeholders share

the goal of developing an optimized and sustainable process.

CHAPTER 13

Information Sources

and Further Reading

T

here are various sources of information on optimization and

integration in the process industry. The aim of this chapter is

to summarize many of these sources, although their number is

growing rapidly. No such list could be fully comprehensive, but

every attempt has been made to include the most important sources

of information.

The chapter is divided into five sections as follows: (1) general

sources of information, (2) Heat Integration, (3) Mass Integration,

(4) combined analysis, and (5) optimization for sustainable industry.

Within each section, listings are further divided into four groups:

conferences, journals, service providers, and projects. A listing is

repeated when the information is relevant to more than one category;

this makes searching more efficient for users.

13.1.1 Conferences

Conferences that address Process Integration (PI) approaches to

minimizing the use of energy, water, and other resources can be

sorted into three groups. The first group consists of conferences that

are directly related to energy and resource minimization.

Saving and Pollution Reduction (PRES), organized annually

since 1998, <www.conferencepres.com>

European Symposium on Computer Aided Process

Engineering (ESCAPE), organized annually since 1992,

<www.cape-wp.eu>

295

296 Chapter Thirteen

Engineering (ICOSSE), organized since 2009, <www.aiche.

org/IFS/Conferences/index.aspx>

Dubrovnik Conference on Sustainable Development of

Energy, Water and Environment Systems (SDEWES),

organized biannually since 2002, <www.sdewes.fsb.hr>

Italian Conference on Chemical and Process Engineering

(ICheaP), organized biannually since 1993, <www.aidic.it/

italiano/pubblicazioni/confse.htm>

Europe Energy Efficiency Conference, <www.esv.or.at/esv/

index.php?id=1484&L=1>

throughout the world and for which PI topics are part of their

scientific programs.

and cosponsored conferences, <www.aiche.org/Conferences>

European Congress of Chemical Engineering (ECCE),

organized biennially

Chemical Engineering, Chemical Equipment Design and

Automation (CHISA), organized biennially since 1962 (odd

years), <www.chisa.cz>

World Congress of Chemical Engineering,<www.wcce8.org/

index.html> (the 9th World Congress of Chemical

Engineering, incorporating the 14th APCChE Congress, will

be held in 2013)

World Bioenergy, <www.elmia.se/worldbioenergy/>

World Sustainable Energy Days, <www.wsed.at>

World Renewable Energy Congress: Innovation in Europe,

<www.wrenuk.co.uk>

specific issues (food processing, pulp and paper, etc.) and at which

energy and resource efficiency are addressed as special cases.

2009>

TAPPI EPE (Engineering, Pulping and Environmental)

Conference, organized annually, <www.tappiepe.org>

CLIMA Congress, official congress of the Federation of

European Heating and Air-Conditioning Societies (REHVA),

<www.clima2010.org>

Information Sources and Further Reading 297

13.1.2 Journals

There are many journals that cover topics related to energy, water,

and resource minimization from the Process Integration standpoint;

just a few of them are listed (alphabetically) here.

Journal of Clean Technologies and Environmental Policy, <www.

springer.com/environment/sustainable+development/journal/

10098>

Journal of Cleaner Production, <www.elsevier.com/wps/find/

journaldescription.cws_home/30440/description#description>

Resources, Conservation & Recycling, <www.elsevier.com/

wps/find/journaldescription.cws_home/503358/description

#description>

The following alphabetical list includes service providers as well as

professional bodies and networks.

Wolverhampton, WV6 8TQ, UK, <www.adas.co.uk/contact/

index.html>

Artie McFerrin Department of Chemical Engineering, Texas

A&M University, College Station, TX 77843, USA, <www.che.

tamu.edu>, contact person: Mahmoud M. El-Halwagi

AspenTech, Aspen Technology, Inc., 200 Wheeler Rd.,

Burlington, MA 01803, USA, <www.aspentech.com>

Dansk Energi Analyse A/S, <www.dea.dk>

Department of Chemical Engineering, Auburn University,

212 Ross Hall, Auburn University, AL 36849-5127, USA, <eng.

auburn.edu/programs/chen/about/index.html>

BIS, Department for Business, Innovation and Skills, UK,

<www.bis.gov.uk>

Centre for Advanced Process Decision-Making, Carnegie

Mellon University, Department of Chemical Engineering,

5000 Forbes Ave., Pittsburgh, PA 15213, USA, <capd.cheme.

cmu.edu>

Center for Engineering and Sustainable Development

Research, De La Salle University, Manila, 2401 Taft Avenue,

1004 Manila, Philippines, <www.dlsu.edu.ph/research/

centers/cesdr/strg.asp>, contact person: Raymond Tan

Centre for Process Integration, School of Chemical

Engineering and Analytical Science, The University of

298 Chapter Thirteen

<www.ceas.manchester.ac.uk/research/researchcentres/

centreforprocessintegration>, contact person: Robin Smith

Centre for Process Integration and Intensification (CPI2),

European Community Project Marie Curie Chair (EXC)

MEXC-CT-2003-042618 INEMAGLOW, Research Institute of

Chemical and Process Engineering, Faculty of Information

Technology, University of Pannonia, Egyetem u.10, Veszprem,

H-8200, Hungary, <inemaglow.dcs.uni-pannon.hu>, contact

person: Ji Kleme

Centre for Process Systems Engineering, Imperial College

London, C507 Roderic Hill Bldg., South Kensington Campus,

London, UK, <www3.imperial.ac.uk/centreforprocess

systemsengineering>, contact person: Efstratios Pistikopoulos

Centre for Technology Transfer in the Process Industries,

University POLITEHNICA Bucharest, 1 Polizu St., Bldg. A,

RO-011061, Bucharest, Romania, <www.chim.upb.ro/CTTPI>,

contact person: Valentin Pleu

Charles Parsons Institute, University of Limerick, Limerick,

Ireland, <www2.ul.ie/web/WWW/Faculties/Science_%26_

Engineering/Research/Research_Institutes/CPI>, contact

person: Toshko Zhelev

Chiyoda Corporation, Energy Frontier Business Development

Office, 2-12-1 Tsurumichuo, Tsurumi-ku, Yokohama 230-8601,

Japan, <www.chiyoda-corp.com/en>, contact person: Kazuo

Matsuda

COWI A/S, Parallelvej 2, DK-2800, Kongens Lyngby, Denmark,

tel. +45 45 97 22 11, <www.cowi.com/>, cowi@cowi.dk

DEFRA, Department for Environment, Food and Rural

Affairs, UK, <www.defra.gov.uk>

Department of Chemical and Environmental Engineering,

University of Nottingham Malaysia, Broga Rd., 43500

Semenyih, Selangor, Malaysia, <www.nottingham.edu.

my/Faculties/Engineering/Research/ENV>, contact person:

Dominic Foo

Department of Chemical Engineering, University of Maribor,

Smetanova ulica 17, Maribor, Slovenia, <atom.uni-mb.si>,

contact person Zdravko Kravanja

Department of Chemical Engineering, Massachusetts

Institute of Technology, 77 Massachusetts Ave., Room 66350,

Cambridge, MA 02139, <web.mit.edu/cheme/index.html>

Department of Chemical Engineering, University of Pretoria,

Lynwood Rd., Pretoria 0002, South Africa, <web.up.ac.za/

Information Sources and Further Reading 299

default.asp?ipkCategoryID=2063&language=0>, contact person:

Thoko Majozi

Department of Computer Science and Systems Technology,

Faculty of Information Technology, University of Pannonia,

Egyetem u.10, Veszprem, H-8200, Hungary, <www.dcs.vein.

hu>, contact person: Ferenc Friedler

Department of Energy and Climate Change, UK, <www.

decc.gov.uk>

Department of Energy and En