Pattern Recognition in Industry (eBook)

Front Cover 1
Pattern Recognition in Industry 4
Copyright Page 5
Contents 14
Preface 8
Acknowledgments 10
About the Author 12
Part I: Philosophy 22
Chapter 1. Introduction 24
1.1. Distinguishing Knowledge and Information from Data 24
1.2. Whence Pattern Recognition Technology 25
1.3. Thermodynamic Concept of Order Leading to Information Theory 26
1.4. Modeling Informed by Observation 27
1.5. Pattern Recognition Technology Triad 28
References 29
Chapter 2. Patterns Within Data 30
2.1. Types of Data 30
2.2. Characterizing Data 30
2.3. Distance Between Data 32
2.4. Organizing Data—Clustering / Auto-Classification 34
2.5. Organizing Data—Data Series Resonance 35
2.6. Organizing Data—Correlative Modeling 35
References 36
Chapter 3. Adapting Biological Principles for Deployment in Computational Science 38
3.1. Learning Organisms—An Introduction to Neural Nets 38
3.2. Supervised Learning 40
3.3. Unsupervised Learning 42
3.4. Models that Self-Organize Data (Unsupervised Learning) as well as Correlate them with Dependent Outcomes (Supervised Learning) 43
3.5. Genetic Algorithms 45
References 46
Chapter 4. Issues in Predictive Empirical Modeling 48
4.1. Pre-Conditioning Data: Pre- and Post-Processing 48
4.2. Detecting Extrapolative Conditions 49
4.3. Embedding Mechanistic Understanding / Experiential Judgment to Enhance Extrapolative Robustness 49
4.4. Insight into Model Behavior 50
Part II: Technology 52
Chapter 5. Supervised Learning—Correlative Neural Nets 54
5.1. Supervised Learning with Back-Propagation Neural Nets 54
5.2. Feedforward—Exercising the BP Net in Predictive Mode—Neuron Transformation Function 54
5.3. BP Training—Connection Weights Adjusted by the “Delta Rule” to Minimize Learning Errors 57
5.4. Back-Propagation Equations for General Transformation Functions 58
5.5. Back-Propagation Equations for Sigmoidal Transformation Functions 60
5.6. Conjugate Gradient Methodology for Rapid and Robust Convergence 61
5.7. Separating Signal from Noise in Training 62
5.8. Pre-Conditioning Data for BP Nets 63
5.9. Supervised Learning with Radial Basis Function Neural Nets 65
5.10. Seeding the Input Data Space with RBF Cluster Centers 65
5.11. Assigning Spheres of Influence to each Cluster 67
5.12. Activating Clusters from a Point in the Data Space 67
5.13. Developing RBF Correlation Models—Assigning Weights to Map Outcome 68
5.14. Pre-Conditioning Data for RBF Nets 68
5.15. Neural Net Correlation Models 68
References 69
Chapter 6. Unsupervised Learning: Auto-Clustering and Self-Organizing Data 70
6.1. Unsupervised Learning—Value to Industry 70
6.2. Auto-Clustering Using Radial Basis Functions 70
6.3. RBF Cluster Radius 71
6.4. Competitive Learning 72
6.5. Data Pre-Conditioning for Competitive Learning 76
References 77
Chapter 7. Customizing for Industrial Strength Applications 78
7.1. Modeling: The Quest for Explaining and Predicting Processes 78
7.2. Combining Empiricism with Mechanistic Understanding 78
7.3. Embedding an Idealized (Partially Correct) Model 79
7.4. Embedding A Priori Understanding in the Form of Constraints 85
7.5. Incorporating Mixed Data Types 87
7.6. Confidence Measure for Characterizing Predictions 88
7.7. Interpreting Trained Neural Net Structures 90
7.8. Graphical Interpretation of Trained Neural Net Structures 93
References 94
Chapter 8. Characterizing and Classifying Textual Material 96
8.1. Capturing a Document’s Essential Features through Fingerprinting 96
8.2. Similar Documents Auto-Classified into Distinct Clusters 97
8.3. Activity Profiles of Authors Provide Competitive Insight 98
8.4. Visualizing a Document’s Contents 99
8.5. Identifying Keywords through Entropic Analysis of Text Documents 100
8.6. Automation Shrinks Time and Resources Required to Keep up with the World 103
References 103
Chapter 9. Pattern Recognition in Time Series Analysis 104
9.1. Leading Indices as Drivers 104
9.2. Concept of Resonance in Quantifying Similarities between Time Series 104
9.3. Identifying Leading Indicators 105
9.4. Forecasting 107
Reference 108
Chapter 10. Genetic Algorithms 110
10.1. Background 110
10.2. Definitions 111
10.3. Setting the Stage 111
10.4. Selection 113
10.5. Mating 114
10.6. Mutation 115
10.7. “Breeding” Fit Solutions 117
10.8. Discovering Profitable Operating Strategies 119
10.9. Product Formulation 119
References 121
Part III: Case Studies 122
Chapter 11. Harnessing the Technology for Profitability 124
11.1. Process Industry Application Modes 124
11.2. Business Applications 128
11.3. Case Studies that Follow 130
Chapter 12. Reactor Modeling Through in Situ Adaptive Learning 132
12.1. Background 132
12.2. Reactor Catalyst Deactivation 132
12.3. Model Configuration 133
12.4. In Situ Modeling Scheme 134
12.5. Validation Procedure 136
12.6. Validation Results 136
12.7. Roles Played by Modeling and Plant Operational Teams 138
12.8. Competitive Advantage Derived through this Approach 138
Reference 139
Chapter 13. Predicting Plant Stack Emissions to Meet Environmental Limits 140
13.1. Background 140
13.2. Reactor Flow and Model Configuration 140
13.3. Model Training and Results 142
13.4. Identifying Optimal Operating Windows for Enhancing Profits 143
Chapter 14. Predicting Fouling/Coking in Fired Heaters 144
14.1. Background 144
14.2. Model Configuration 144
14.3. Model Results 146
14.4. Conclusions 147
Chapter 15. Predicting Operational Credits 148
15.1. Background 148
15.2. Issues 148
15.3. Model Configuration 149
15.4. Model Results 150
15.5. Plant Follow-Up 151
Chapter 16. Pilot Plant Scale-up by Interpreting Tracer Diagnostics 152
16.1. Background 152
16.2. Issue 152
16.3. Genetic Algorithm–Simulation Model Coupling 153
16.4. Results and Conclusion 156
Chapter 17. Predicting Distillation Tower Temperatures: Mining Data for Capturing Distinct Operational Variability 158
17.1. Background 158
17.2. Issue 158
17.3. Model Configuration 158
17.4. Identifying Distinctly Different Operating Conditions 159
17.5. Results 160
Chapter 18. Enabling New Process Design Based on Laboratory Data 164
18.1. Background 164
18.2. Model Configuration—Bi-Level Focus for “Spot-Lighting” Region of Interest 164
18.3. Model Results 166
18.4. Conclusion 168
Chapter 19. Forecasting Price Changes of a Composite Basket of Commodities 170
19.1. Background 170
19.2. Approach and Model Configuration 170
19.3. Model Results 172
19.4. Conclusions 174
Chapter 20. Corporate Demographic Trend Analysis 176
20.1. Background 176
20.2. Issues 176
20.3. Approach and Model Configuration 176
20.4. Model Results and Conclusions 177
Epilogue 180
Appendices 182
Appendix A1. Thermodynamics and Information Theory 184
A1.1. Thermodynamic Concepts Set the Stage for Quantifying Information 184
A1.2. Equilibrium as a State of Disorder—Organization as a Value-Adding Process 185
A1.3. Entropy, Disorder, and Uncertainty 186
A1.4. Opportunities Found in Imbalances 187
A1.5. Appreciation through Quantification 188
A1.6. Quantifying Information Transfer 188
A1.7. Information Content in a System 189
References 191
Appendix A2. Modeling 192
A2.1. What Are Models 192
A2.2. Mechanistic Modeling—General Laws 192
A2.3. Particular Laws and Constitutive Relations 194
A2.4. Combining General Laws and Constitutive Relations 195
A2.5. Modeling Directly from Data 195
Reference 195
Index 196