Cover
Title Page
Copyright
Preface
Biography
1 The Basic Ideas
1. 1.1 Introduction
2. Notes
2 Design of Experiments – DOE
1. 2.1 The 2² Factorial Designs
2. 2.2 Effects for 2² Factorial Designs
3. 2.3 Interactions Between Factors
4. 2.4 Standard Error for the Effects
5. 2.5 2³ Factorial Design
6. 2.6 Effects for the 2³ Factorial Designs
7. 2.7 Standard Errors of Effects for Two‐ and Three‐Level Factorial Designs
3 Neural Network Modeling – Data Mining
1. 3.1 Data Preprocessing
2. 3.2 Building, Training, and Verifying Model
3. 3.3 Model Analyzing
4. 3.4 What‐Ifs Optimization
5. 3.5 DOE Experiment Using Neural Networks Model
6. Note
4 Evolutionary Operation – EVOP
1. 4.1 Small‐Scale and Plant‐Scale Investigation
2. 4.2 Scale‐up
3. 4.3 Static and Evolutionary Operation
4. 4.4 Analysis of Information Board
5. 4.5 Three‐Factor Scheme
6. 4.6 Current Best‐Known Conditions
7. 4.7 Change in Mean for a 2² Factorial Design with Center Point
8. 4.8 Standard Errors for the Effects
9. 4.9 The Effects and Their Standard Errors for a 2² Design with Center Point
10. 4.10 Analysis of Information Board for Three Responses Using Factorial Effects
11. 4.11 2³ Factorial Design Effects, Interpretation, and Information Board
12. 4.12 Dividing the 2³ Factorial Design Into Two Blocks
13. 4.13 2³ Design with Two Center Points Run in Two Blocks
5 Different Techniques of EVOP
1. 5.1 Box EVOP – BEVOP
2. 5.2 Calculation Procedure for Two‐Factor EVOP
3. 5.3 Calculation Procedure for Three‐Factor EVOP
4. 5.4 BEVOP in Plant‐Scale Experiments
5. 5.5 BEVOP Applications
6. 5.6 BEVOP Advantages and Disadvantages
7. 5.7 BEVOP Simulation
8. 5.8 Rotating Square Evolutionary Operation – ROVOP
9. 5.9 Random Evolutionary Operation – REVOP
10. 5.10 Quick‐Start EVOP – QSEVOP
11. 5.11 QSEVOP Simulation
12. 5.12 Simplex Evolutionary Operation – SEVOP
13. 5.13 Some Practical Advice About Using EVOP
6 EVOP Software
Appendix A: The Approximate Method of Estimating the Standard Deviation in EVOP
Appendix B: 2²‐ and 2³‐Factor Box EVOP Calculations with Center Point
1. B.1 2³ Three‐Factor Box EVOP Calculations with Center Point
Appendix C: Short Table of Random Normal Deviates
Appendix D: How Many Cycles Are Necessary to Detect Effects of Reasonable Size
Appendix E: Multiple Responses: The Desirability Approach
References
Index
End User License Agreement

List of Tables

Chapter 2
1. Table 2.1 2² Factorial design.
2. Table 2.2 2² Factorial design in coded values.
3. Table 2.3 Factors settings.
4. Table 2.4 The eight sets of conditions of a 2³ factorial design.
5. Table 2.5 2³ Factorial design in coded values.
6. Table 2.6 Standard Errors (S.E.) for effects.
7. Table 2.7 Response surface design (n = 2; b = 4; N = 20).
8. Table 2.8 D‐Optimal design (n = 3; N = 19).
Chapter 3
1. Table 3.1 Five years, process data and fiber strength.
2. Table 3.2 Sensitivity report.
3. Table 3.3 Set points and What‐Ifs optimization tool.
4. Table 3.4 The central composite rotatable design.
5. Table 3.5 The central composite rotatable design matrix.
6. Table 3.6 The factor's levels.
Chapter 4
1. Table 4.1 The effects and standard errors.
2. Table 4.2 Information board for by‐product yields after three cycles.
3. Table 4.3 By‐product data for three repetitions/cycles of a 2³ design.
4. Table 4.4 Values of f_k,n.
5. Table 4.5 Calculation for S₂ and S₃.
Chapter 5
1. Table 5.1 Individual batch yields.
2. Table 5.2 Information board.
3. Table 5.3 Individual batch yields.
4. Table 5.4 Information board.
5. Table 5.5 2² EVOP information board.
6. Table 5.6 Initial experimental conditions.
7. Table 5.7 Information board.
8. Table 5.8 BEVOP calculation.
9. Table 5.9 Information board.
10. Table 5.10 BEVOP calculation.
11. Table 5.11 Information board.
12. Table 5.12 BEVOP calculation.
13. Table 5.13 Information board.
14. Table 5.14 BEVOP calculation.
15. Table 5.15 Information board.
16. Table 5.16 Initial experimental conditions.
17. Table 5.17 Information board.
18. Table 5.18 BEVOP calculation.
19. Table 5.19 Information board.
20. Table 5.20 BEVOP calculation.
21. Table 5.21 Information board.
22. Table 5.22 BEVOP calculation.
23. Table 5.23 Information board.
24. Table 5.24 BEVOP calculation.
25. Table 5.25 Information board.
26. Table 5.26 2² BEVOP simulation.
27. Table 5.27 2² BEVOP information board.
28. Table 5.28 2² BEVOP simulation.
29. Table 5.29 2² BEVOP information board.
30. Table 5.30 2² BEVOP simulation.
31. Table 5.31 2² BEVOP information board.
32. Table 5.32 2² BEVOP simulation.
33. Table 5.33 2² BEVOP information board.
34. Table 5.34 2² BEVOP simulation.
35. Table 5.35 2² BEVOP information board.
36. Table 5.36 2² BEVOP simulation.
37. Table 5.37 2² BEVOP information board.
38. Table 5.38 2² BEVOP simulation.
39. Table 5.39 2² BEVOP information board.
40. Table 5.40 2² BEVOP simulation.
41. Table 5.41 2² BEVOP information board.
42. Table 5.42 2² BEVOP simulation.
43. Table 5.43 2² BEVOP information board.
44. Table 5.44 2² BEVOP simulation.
45. Table 5.45 2² BEVOP information board.
46. Table 5.46 2³ BEVOP simulation.
47. Table 5.47 2³ BEVOP information board.
48. Table 5.48 2³ BEVOP simulation.
49. Table 5.49 2³ BEVOP information board.
50. Table 5.50 2³ BEVOP simulation.
51. Table 5.51 2³ BEVOP information board.
52. Table 5.52 2³ BEVOP simulation.
53. Table 5.53 2³ BEVOP information board.
54. Table 5.54 2² ROVOP simulation.
55. Table 5.55 2² ROVOP information board.
56. Table 5.56 2² ROVOP simulation.
57. Table 5.57 2² ROVOP information board.
58. Table 5.58 2² ROVOP simulation.
59. Table 5.59 2² ROVOP information board.
60. Table 5.60 2² ROVOP simulation.
61. Table 5.61 2² ROVOP information board.
62. Table 5.62 2² ROVOP simulation.
63. Table 5.63 2² ROVOP information board.
64. Table 5.64 2² ROVOP simulation.
65. Table 5.65 2² ROVOP information board.
66. Table 5.66 2² ROVOP simulation.
67. Table 5.67 2² ROVOP information board.
68. Table 5.68 2² ROVOP simulation.
69. Table 5.69 2² ROVOP information board.
70. Table 5.70 2² ROVOP simulation.
71. Table 5.71 2² ROVOP information board.
72. Table 5.72 2² BEVOP simulation.
73. Table 5.73 2² BEVOP information board.
74. Table 5.74 2² BEVOP simulation.
75. Table 5.75 2² BEVOP information board.
76. Table 5.76 2² BEVOP simulation.
77. Table 5.77 2² BEVOP information board.
78. Table 5.78 2² BEVOP simulation.
79. Table 5.79 2² BEVOP information board.
80. Table 5.80 2³ ROVOP simulation.
81. Table 5.81 2³ ROVOP information board.
82. Table 5.82 2³ ROVOP simulation.
83. Table 5.83 2³ ROVOP information board.
84. Table 5.84 2³ ROVOP simulation.
85. Table 5.85 2³ ROVOP information board.
86. Table 5.86 2³ ROVOP simulation.
87. Table 5.87 2³ ROVOP information board.
88. Table 5.88 2³ BEVOP simulation.
89. Table 5.89 2³ BEVOP information board.
90. Table 5.90 2³ BEVOP simulation.
91. Table 5.91 2³ BEVOP information board.
92. Table 5.92 2³ BEVOP simulation.
93. Table 5.93 2³ BEVOP information board.
94. Table 5.94 2³ BEVOP simulation.
95. Table 5.95 2³ BEVOP information board.
96. Table 5.96 REVOP work sheet.
97. Table 5.97 REVOP simulation.
98. Table 5.98 REVOP simulation.
99. Table 5.99 REVOP simulation.
100. Table 5.100 REVOP simulation.
101. Table 5.101 REVOP simulation.
102. Table 5.102 QSEVOP simulation.
103. Table 5.103 QSEVOP simulation.
104. Table 5.104 QSEVOP simulation.
105. Table 5.105 QSEVOP simulation.
106. Table 5.106 QSEVOP simulation.
107. Table 5.107 Coordinates of simplex vertices.
108. Table 5.108 Simplex matrix.
109. Table 5.109 Factors and variation intervals.
110. Table 5.110 Initial simplex.
111. Table 5.111 Initial simplex.
112. Table 5.112 Simplex optimization.
113. Table 5.113 Initial simplex.
114. Table 5.114 Simplex optimization.
115. Table 5.115 SEVOP simulation, Phase I.
116. Table 5.116 SEVOP simulation, Phase XI.
117. Table 5.117 SEVOP simulation, Phase XII.
118. Table 5.118 Summary of simplex movement.
119. Table 5.119 SEVOP simulation, Phase I.
120. Table 5.120 SEVOP simulation, Phase I.
121. Table 5.121 Summary of SEVOP movement.

List of Illustrations

Chapter 2
1. Figure 2.1 Fluidity averages after four cycles.
2. Figure 2.2 2³ factorial design with average by‐product yields after three cy...
3. Figure 2.3 Ozone column with bleaching.
4. Figure 2.4 Ozone product GEB response surface.
5. Figure 2.5 Contour diagram for ozone product GEB response surface.
6. Figure 2.6 Bleach product GEB response surface.
7. Figure 2.7 Contour diagram for bleach product GEB response surface.
Chapter 3
1. Figure 3.1 The Relative Error R² and the strip chart with original and predi...
2. Figure 3.2 The Predicted vs. Actual‐Scatter plot.
3. Figure 3.3 Sensitivity vs. Rank analysis plots.
4. Figure 3.4 Output vs. Percent Plots of the input–output relationships.
5. Figure 3.5 Sensitivity vs. Percent analysis plots.
6. Figure 3.6 Response surfaces.
Chapter 4
1. Figure 4.1 Effects and the “noise level”.
2. Figure 4.2 Information board after four cycles.
3. Figure 4.3 Three‐factor EVOP scheme.
4. Figure 4.4 Interpretation of the change in mean effect.
5. Figure 4.5 Information board at the end of four cycles.
6. Figure 4.6 Approximate contours for (a) cost, (b) impurity, and (c) fluidity...
7. Figure 4.7 The blocks of the 2³ design.
Chapter 5
1. Figure 5.1 2² EVOP with center point.
2. Figure 5.2 Work sheet Cycle No. 1.
3. Figure 5.3 Work sheet Cycle No. 2.
4. Figure 5.4 Work sheet Cycle No. 3.
5. Figure 5.5 Work sheet Cycle No. 4.
6. Figure 5.6 2³ Design with center points.
7. Figure 5.7 Work sheet Cycle No. 1.
8. Figure 5.8 Work sheet Cycle No. 2.
9. Figure 5.9 Work sheet Cycle No. 3.
10. Figure 5.10 2² Design for EVOP.
11. Figure 5.11 Work sheet cycle No. 1.
12. Figure 5.12 Work sheet cycle No. 2.
13. Figure 5.13 Work sheet cycle No. 1.
14. Figure 5.14 Work sheet cycle No. 2.
15. Figure 5.15 Work sheet cycle No. 1.
16. Figure 5.16 Work sheet cycle No. 2.
17. Figure 5.17 Work sheet cycle No. 1.
18. Figure 5.18 Work sheet cycle No. 2.
19. Figure 5.19 Work sheet cycle No. 1.
20. Figure 5.20 Work sheet cycle No. 2.
21. Figure 5.21 Work sheet cycle No. 1.
22. Figure 5.22 Work sheet cycle No. 2.
23. Figure 5.23 Work sheet cycle No. 1.
24. Figure 5.24 Work sheet cycle No. 2.
25. Figure 5.25 Work sheet cycle No. 1.
26. Figure 5.26 Work sheet cycle No. 2.
27. Figure 5.27 Work sheet cycle No. 1.
28. Figure 5.28 Work sheet cycle No. 2.
29. Figure 5.29 Work sheet cycle No. 1.
30. Figure 5.30 Work sheet cycle No. 2.
31. Figure 5.31 Work sheet cycle No. 1.
32. Figure 5.32 Work sheet cycle No. 2.
33. Figure 5.33 Work sheet cycle No. 1.
34. Figure 5.34 Work sheet cycle No. 3.
35. Figure 5.35 Work sheet cycle No. 10.
36. Figure 5.36 Work sheet cycle No. 1.
37. Figure 5.37 Work sheet cycle No. 3.
38. Figure 5.38 Work sheet cycle No. 10.
39. Figure 5.39 Work sheet cycle No. 1.
40. Figure 5.40 Work sheet cycle No. 3.
41. Figure 5.41 Work sheet cycle No. 10.
42. Figure 5.42 Work sheet cycle No. 1.
43. Figure 5.43 Work sheet cycle No. 3.
44. Figure 5.44 Work sheet cycle No. 10.
45. Figure 5.45 Work sheet cycle No. 1.
46. Figure 5.46 Work sheet cycle No. 3.
47. Figure 5.47 Work sheet cycle No. 10.
48. Figure 5.48 Response surface.
49. Figure 5.49 Contour diagram.
50. Figure 5.50 Work sheet cycle No. 1.
51. Figure 5.51 Work sheet cycle No. 3.
52. Figure 5.52 Work sheet cycle No. 10.
53. Figure 5.53 Work sheet cycle No. 1.
54. Figure 5.54 Work sheet cycle No. 3.
55. Figure 5.55 Work sheet cycle No. 10.
56. Figure 5.56 Work sheet cycle No. 1.
57. Figure 5.57 Work sheet cycle No. 3.
58. Figure 5.58 Work sheet cycle No. 10.
59. Figure 5.59 Response surface.
60. Figure 5.60 Contour diagram.
61. Figure 5.61 Work sheet cycle No. 1.
62. Figure 5.62 Work sheet cycle No. 3.
63. Figure 5.63 Work sheet cycle No. 10.
64. Figure 5.64 Work sheet cycle No. 1.
65. Figure 5.65 Work sheet cycle No. 3.
66. Figure 5.66 Work sheet cycle No. 10.
67. Figure 5.67 Response surface.
68. Figure 5.68 Contour diagram.
69. Figure 5.69 Work sheet cycle No. 1.
70. Figure 5.70 Work sheet cycle No. 3.
71. Figure 5.71 Work sheet cycle No. 10.
72. Figure 5.72 Work sheet cycle No. 1.
73. Figure 5.73 Work sheet cycle No. 3.
74. Figure 5.74 Work sheet cycle No. 10.
75. Figure 5.75 Work sheet cycle No. 1.
76. Figure 5.76 Work sheet cycle No. 3.
77. Figure 5.77 Work sheet cycle No. 10.
78. Figure 5.78 Work sheet cycle No. 1.
79. Figure 5.79 Work sheet cycle No. 3.
80. Figure 5.80 Work sheet cycle No. 10.
81. Figure 5.81 Response surface.
82. Figure 5.82 Contour diagram.
83. Figure 5.83 Five successive 2² ROVOP patterns.
84. Figure 5.84 Moving center of the ROVOP without reduction of interval.
85. Figure 5.85 Moving center of the ROVOP with reduction of interval.
86. Figure 5.86 Work sheet replica No. 10.
87. Figure 5.87 Work sheet replica No. 10.
88. Figure 5.88 Work sheet replica No. 10.
89. Figure 5.89 Work sheet replica No. 10.
90. Figure 5.90 Work sheet replica No. 10.
91. Figure 5.91 Work sheet replica No. 10.
92. Figure 5.92 Work sheet replica No. 10.
93. Figure 5.93 Work sheet replica No. 10.
94. Figure 5.94 Work sheet replica No. 10.
95. Figure 5.95 Work sheet cycle No. 10.
96. Figure 5.96 Work sheet cycle No. 10.
97. Figure 5.97 Work sheet cycle No. 10.
98. Figure 5.98 Work sheet cycle No. 10.
99. Figure 5.99 Graphical interpretation of ROVOP moving toward optimum.
100. Figure 5.100 Two successive 2³ ROVOP patterns.
101. Figure 5.101 Simplex movement to optimum.
102. Figure 5.102 Simplex in coordinate system.
103. Figure 5.103 Tetrahedron in coordinate system.
104. Figure 5.104 Coordinates of simplex vertices.
105. Figure 5.105 Simplex movement on response surface.
106. Figure 5.106 Initial simplex.
107. Figure 5.107 Simplex movement.
108. Figure 5.108 Simplex movement on response surface.
109. Figure 5.109 The simplex movement on the response surface.
Appendix E
1. Figure E.1 Desirability functions and optimal solution for example problem....
2. Figure E.2 Overall desirability function for example problem.

Preface

“Let your systems learn the wisdom of age and experience”

It is well known that chemical engineers are involved in each step of the process development (lab‐scale and pilot‐scale) and later in the scale‐up and full‐scale plant process improvement. So far, there are three powerful methodologies/tools for chemical engineers with wide application in industry:

Design of Experiments – DOE
Evolutionary Operation – EVOP
Data Mining using Neural Networks – DM

The objective of this new book is to compare all three methods in full‐scale plant optimization applications and to focus on a simple but powerful technique which provides the experimental tools for full‐scale plant optimization. It is called Evolutionary Operation or EVOP.

Chapter I provides the basic principles necessary for a simple understanding of the difference between these three methodologies.

Chapter II covers only two‐ and three‐factor full factorial designs from the Design of Experiments because of the subject of this book. DOE experiments are very efficient in the lab or pilot plant but NOT in the full‐scale process. There are two main reasons for that:

An owner of the large‐scale process would hardly accept DOE experiment. To detect significant factor's effect in industrial process, a large change in control factors is required, and that on the other hand will produce off‐spec product.
Also replications required by DOE to reduce noise will produce even more off‐spec products.

Chapter III describes theory and application of Data Mining using the Neural Networks modeling tool. In the last twenty or more years, with more data historians installed in process industry, Data Mining, using neural network software, is widely used for industrial process optimization. The main problem with this approach without experiment is static operation of the plant, which means running the plant with almost no change in process factors. Standard Operating Procedures keep factors constant for a very long time. A greater change in process factors is not allowed. If it happens, it will produce off‐spec product. This means, all collected data are in very narrow experimental space range, and usually local or global optimum of the process is out of that range.

An analysis of historical data typically has only 10% to 20% chance of success, despite the fact that hundreds or thousands of data points may be available.

Chapter IV, DOE and Data Mining disadvantages, strongly suggests reusing Evolutionary Operation – EVOP experimental technique. Box G.E.P proposed a special Operation as a technique for improving industrial productivity. Its basic philosophy is that it is nearly always inefficient to run an industrial process to produce a product alone; a process should be run so as to generate a product plus information on how to improve that product. The procedure consists of carefully planned cycle of minor variants on the standard work process. The routine procedure consists of running each of the variants continually repeating the cycle. Usually the effects of these deliberate changes in the factors‐variables are masked by the large errors inherent in large‐scale production units. However, since production will continue anyway, a cycle of variants, which do not affect production significantly, can be run almost indefinitely. Because of constant repetition the effect of small changes can be detected. The wisdom of the age and long experience suggests that EVOP tools help chemical engineers dealing with two different objectives:

Scale‐up from lab and pilot‐plant to full‐scale operation and “fine tuning” of large‐scale production.
Once the optimum is achieved at certain point of time, it is going to be a moving target and it is very hard or impossible to use tools like DOE and DM due to variability of raw materials quality, energy supply fluctuation, tooling wear, gradual equipment misalignment, shifts in ambient temperature and humidity, corrosion, decreasing heat transfer, aging equipment and instrumentation, aging catalyst and scale build‐up, etc.

Chapter V describes theory and applications of five different EVOP methods including Box EVOP, with numerous examples and simulations to compare efficiency:

Box Evolutionary Operation – BEVOP
Rotating Square Evolutionary Operation – ROVOP
Random Evolutionary Operation – REVOP
Simplex Evolutionary Operation – SEVOP
Quick Start EVOP – QSEVOP.

These simulators are the perfect tools for training operators in different EVOP techniques. Also they could be used to run full‐scale plant optimization. In meantime, a special software “EVOP Engine” is under construction and will be released after the book is published.

Research and development scientists and engineers may find the book helpful in optimizing various processes to be more efficient, less energy intensive, less time consuming, more reliable, and less wasteful of raw materials.

This book has come into being a product of many years of research activities at Military Technical Institute, Belgrade, and long experience in process and product development with large‐scale production, including Lenzing Fibers Corporation, Austria; MacDermid, TN, US; and BASF Catalysts LLC., GA, US. The author is especially pleased to offer his gratitude to Helena Smuckler and Anica Lazic for editing of the manuscript. I express my special gratitude to Milica Pojiċ. for helping in a search for the literature.

Petrovac, Montenegro
June 2021

Živorad R. Laziċ

Full Scale Plant Optimization in Chemical Engineering

A Practical Guide

Preface

Biography