Thorough discussion of data-driven and first principles models for energy-relevant systems and processes, approached through various in-depth case studies Applied AI Techniques in the Process Industry identifies and categorizes the various hybrid models available that integrate data-driven models for energy-relevant systems and processes with different forms of process knowledge and domain expertise. State-of-the-art techniques such as reduced-order modeling, sparse identification, and physics-informed neural networks are comprehensively summarized, along with their benefits, such as improved interpretability and predictive power. Numerous in-depth case studies regarding the covered models and methods for data-driven modeling, process optimization, and machine learning are presented, from screening high-performance ionic liquids and AI-assisted drug design to designing heat exchangers with physics-informed deep learning. Edited by two highly qualified academics and contributed to by a number of leading experts in the field, Applied AI Techniques in the Process Industry includes information on: Integration of observed data and reaction mechanisms in deep learning for designing sustainable glycolic acid Machine learning-aided rational screening of task-specific ionic liquids and AI for property modeling and solvent tailoring Integration of incomplete prior knowledge into data-driven inferential sensor models under the variational Bayesian framework AI-aided high-throughput screening, optimistic design of MOF materials for adsorptive gas separation, and reduced-order modeling and optimization of cooling tower systems Surrogate modeling for accelerating optimization of complex systems in chemical engineering Applied AI Techniques in the Process Industry is an essential reference on the subject for process, chemical, and pharmaceutical engineers seeking to improve physical interpretability in data-driven models to enable usage that scales with a system and reduce inaccuracies and mismatch issues.

Table of Contents:
Preface xi 1 AI for Property Modeling, Solvent Tailoring, and Process Design 1 Yuqiu Chen 1.1 AI-Assisted Property Modeling 1 1.2 AI-assisted Solvent Tailoring 10 1.3 AI-Assisted Process Design 14 1.4 Conclusions 17 References 19 2 Hunting for Better Aromatic Chemicals with AI Techniques 23 Qilei Liu, Haitao Mao, Lu Wang, and Lei Zhang 2.1 Introduction 23 2.2 Machine Learning-Based Odor Prediction Models 25 2.2.1 Odor Predictions for Pure Aromatic Chemicals Using Group-Based Machine Learning Method 25 2.2.1.1 Database Preparation 25 2.2.1.2 Molecular Representation 26 2.2.1.3 Model Architecture 27 2.2.1.4 Results and Discussions 27 2.2.2 Odor Prediction for Mixture Aromatic Chemicals Using σ-Profiles-Based Machine Learning Method 29 2.2.2.1 Database Preparation 29 2.2.2.2 Molecular Representation 31 2.2.2.3 Model Architecture 34 2.2.2.4 Results and Discussions 35 2.3 Computer-Aided Aroma Design (CAAD) Framework 36 2.3.1 CAAD for Pure Aromatic Chemicals 36 2.3.1.1 Identify Product Attributes 36 2.3.1.2 Convert Product Attributes to Properties and Their Constraints 38 2.3.1.3 Choose Property Prediction Model for Estimating Properties 38 2.3.1.4 Formulate MILP/MINLP Model 38 2.3.1.5 Solve the Model Using Decomposition-Based Algorithm 38 2.3.1.6 Verification 39 2.3.2 CAAD for Mixture Aromatic Chemicals 39 2.3.2.1 Identify Product Attributes 41 2.3.2.2 Convert Product Attributes into Properties and Corresponding Constraints 42 2.3.2.3 Establish Property Models 42 2.3.2.4 Ingredient Screening 42 2.3.2.5 Verification 43 2.4 Case Studies 44 2.4.1 Pure Aroma Design for Shampoo Additives 44 2.4.2 Pure Aroma Design for the Ingredient in Insect Repellent Spray 45 2.4.3 Mixture Aroma Design for Aroma Substitutes 48 2.4.4 Mixture Aroma Design for Odor Tuning 51 2.5 Conclusions 53 2.a The CAS Number of Molecules and the Selected Groups 54 2.b The Calculation Formula of Odor Score 56 2.c The Parameters and Results of the ANN Model 57 2.d The Designed Results of Molecules for Case Study 2 59 2.e Aroma Compounds for Ingredient Screening 61 Acknowledgments 75 References 75 3 Machine Learning-Aided Rational Screening of Task-Specific Ionic Liquids 79 Ruofan Gu and Zhen Song 3.1 Introduction 79 3.2 Molecule Representation of ILs 80 3.2.1 Groups or Fragment-Based Representation 81 3.2.2 COSMO-Derived Descriptors or Fingerprints 83 3.2.3 Machine-Learned Representations 83 3.3 Machine Learning-Based Structure–Property Models 84 3.3.1 Machine Learning for COSMO-Based Models 84 3.3.2 Machine Learning for UNIFAC Extensions 86 3.3.3 Pure Machine Learning Models 89 3.4 Applications 90 3.4.1 Computer-Aided IL Screening 90 3.4.2 Computer-Aided IL Design 93 3.4.3 Extensions to IL Mixtures 95 3.5 Conclusion and Perspectives 97 References 99 4 Integration of Observed Data and Reaction Mechanisms in Deep Learning for Designing Sustainable Glycolic Acid 105 Xin Zhou 4.1 Introduction 105 4.2 Methodology 107 4.2.1 Database Generation 109 4.2.2 Deep Learning 111 4.2.2.1 Deep Neural Networks 111 4.2.2.2 Deep Belief Networks 112 4.2.2.3 Fully Connected Residual Networks 114 4.2.2.4 Random Forest 115 4.2.3 Optimization and Prediction 116 4.2.4 Life Cycle Multidimensional Evaluation 116 4.3 Results and Discussion 117 4.3.1 Data Analysis and Statistics Before Modeling 117 4.3.1.1 Analysis of Experimental Data 117 4.3.1.2 Data Dependence Analysis 117 4.3.2 Model Comparison and Feature Important Analysis 118 4.3.2.1 Model Comparison 118 4.3.2.2 Feature Important Analysis 123 4.3.3 Performance and Feature Analysis of the Optimized FC-ResNet–GA Model 125 4.3.4 Process Multi-objective Optimization and Experimental Verification 126 4.3.5 LCSA Based on the Optimized Parameters 128 4.3.5.1 Original Life Cycle Framework 128 4.3.5.2 Life Cycle Inventory Analysis 128 4.3.5.3 Life Cycle Sustainable Interpretation and Assessment 129 4.4 Conclusion 131 4.a Pareto Optimization Set 132 4.b Experimental Data 133 4.c Construction Method of Process Simulation Database Using Reaction Mechanism 134 4.c.1 Elimination of the Diffusion Limitations 136 4.c.2 Reaction Kinetics 138 References 139 5 Innovation of Gas Separation Processes: Integrating Computational MOF Design and Adsorption Process Optimization 145 Xiang Zhang and Teng Zhou 5.1 Introduction 145 5.2 Step One: Descriptor Optimization 147 5.2.1 Material–Property Relationship of MOFs 148 5.2.1.1 MOF Representation 148 5.2.1.2 Data-Driven Model for Single-Component Adsorption Isotherm 149 5.2.1.3 Multicomponent Dual-Site Langmuir Isotherm Model 150 5.2.2 Integrated Optimization of MOF Descriptors and PSA Operating Conditions 151 5.2.2.1 Descriptor Design Space 151 5.2.2.2 P/VSA Process Model 152 5.2.2.3 Integrated Design Formulation 153 5.2.3 Results 154 5.2.3.1 Benchmark Process Using Cu-BTC 154 5.2.3.2 Optimal MOF and Process from Integrated Design 154 5.3 Step Two: MOF Matching 157 5.3.1 Material–Property Relationship of MOFs 157 5.3.1.1 Property–Performance Relationship for PE/PA Separation 157 5.3.1.2 Validation with 471 CoRE MOFs 159 5.3.2 From Computational MOF Design to Model-Based MOF Screening 161 5.3.2.1 Identification of MOF Building Blocks 161 5.3.2.2 In Silico Synthesis of Hypothetical MOFs 162 5.3.2.3 MOF Screening via Validity and Feasibility Constraints 164 5.3.2.4 MOF Screening via GCMC Simulations 165 5.3.2.5 MOF Screening via PSA Process Optimization 165 5.3.2.6 Optimal Results of SMOF- 1 165 5.4 Conclusion 167 References 168 6 Reverse Design of Heat Exchange Systems Using Physics-Informed Machine Learning 173 Chang He and Yunquan Chen 6.1 Introduction 173 6.2 PINN-Based Inverse Design Method 176 6.2.1 Overview of Inverse Design 176 6.2.1.1 Standard Physics-Informed Neural Networks 177 6.2.1.2 Design Optimization and Decision-making Methods 180 6.3 Example 1: Finned Heat Sink Model 181 6.3.1 System Description and Objectives 181 6.3.2 Improved PINN Structure 185 6.3.3 Results 185 6.4 Illustrative Example 2: Tubular Air Cooler Model 191 6.4.1 System Description and Objectives 191 6.4.2 Improved PINN Structure 195 6.4.3 Transfer Learning 196 6.4.4 Results 199 6.5 Conclusion 204 References 205 7 Integrating Incomplete Prior Knowledge into Data-Driven Inferential Sensor Models Under Variational Bayesian Framework 211 Zhichao Chen, Hao Wang, Yiran Ma, Cheng Qiu, Le Yao, Xinmin Zhang, and Zhihuan Song 7.1 Introduction 211 7.2 Literature Review 213 7.2.1 Transport Process Scale 213 7.2.2 Unit Operation Scale 214 7.2.3 Overall Summary and Technical Gap 214 7.3 Proposed Approach 214 7.3.1 Loss Function Derivation 215 7.3.2 Knowledge Representation 216 7.3.2.1 Knowledge Description 216 7.3.2.2 Knowledge Section via Self-Attention Mechanism 216 7.3.2.3 Similarity of GCN and SAM 217 7.3.2.4 Sampling from Posterior 219 7.3.3 Model Expressions 221 7.4 Experimental Results 223 7.4.1 Evaluation Metrics 224 7.4.2 Process Description 224 7.4.3 Prior Knowledge Analysis 225 7.4.4 Baseline Models 227 7.4.5 Model Performance Comparisons 228 7.4.6 Comparison with L1 and L2 Regularization Terms 229 7.4.7 Sensitivity Analysis 230 7.5 Conclusions 230 7.A Experimental Settings 232 References 233 8 Data-Driven and Physics-Based Reduced-Order Modeling and Optimization of Cooling Tower Systems 239 Chang He and Zhiqiang Wu 8.1 Introduction 239 8.2 Full-Scale Physical Model of Cooling Towers 241 8.3 Bi-level Reduced-Order Models 244 8.3.1 Design of Optimal Experiments 245 8.3.2 Multi-sample CFD Simulations 247 8.3.3 Model Reduction 248 8.4 Thermodynamic Performance Indicators 251 8.5 Optimization Model 253 8.6 Illustrative Example 254 8.7 Conclusion 261 References 261 9 AI-Aided High-Throughput Screening and Optimistic Design of MOF Materials for Adsorptive Gas Separation 265 li Zhou, Min Cheng, Shihui Wang, and Xu Ji 9.1 Introduction 265 9.2 Methodology 266 9.2.1 Molecular Level Simulation and Screening Driven by Rigorous Molecular Simulation and Machine Learning 266 9.2.1.1 Molecular Characterization 266 9.2.1.2 Structural/Chemical Analysis-Based Prescreening 267 9.2.1.3 Diversity Analysis and Dataset Splitting 268 9.2.1.4 Molecular Simulation and Performance Evaluation Metrics 268 9.2.1.5 AI-Aided Quantitative Structure–Property Relationship Development and Rapid Screening 269 9.2.1.6 Process Level Simulation 270 9.2.1.7 Reverse Molecular Design 270 9.3 Case Studies 273 9.3.1 High-Throughput Screening of Metal–Organic Frameworks for Hydrogen Purification 273 9.3.1.1 Prescreening 273 9.3.1.2 Rapid Screening 274 9.3.1.3 Rigorous Validation 274 9.3.1.4 Structure–Property Relationship Analysis 276 9.3.1.5 Investigation on Practical Factors 277 9.4 Conclusions 283 References 283 10 Surrogate Modeling for Accelerating Optimization of Complex Systems in Chemical Engineering 287 Jianzhao Zhou and Jingzheng Ren 10.1 Introduction 287 10.2 Surrogate Modeling Techniques 289 10.2.1 Polynomial Regression (PR) 290 10.2.2 Polynomial Chaos Expansion 291 10.2.3 Kriging 291 10.2.4 Radial Basis Functions (RBF) 292 10.2.5 High-Dimensional Model Representation (HDMR) 293 10.2.6 Decision Tree (DT) 294 10.2.7 Support Vector Machine (SVM) 295 10.2.8 Artificial Neural Network (ANN) 296 10.3 Application of Surrogate Model in Optimization of Chemical Processes 297 10.3.1 Reaction Engineering 297 10.3.2 Separation Engineering 299 10.3.3 Heat Exchange and Integration 300 10.3.4 Process Design and Synthesis 301 10.4 Conclusion 303 Acknowledgment 303 References 303 Index 313