Machine Learning and Hybrid Modelling for Reaction Engineering Theory and Applications
Machine Learning and Hybrid Modelling for Reaction Engineering summarises latest research and fills a gap in methodology development of hybrid models for reaction engineering applications.
Saved in:
| Main Authors: | , |
|---|---|
| Format: | eBook |
| Language: | English |
| Published: |
La Vergne
Royal Society of Chemistry, The
2023
|
| Edition: | 1 |
| Subjects: | |
| ISBN: | 1839165634, 9781839165634 |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- 12.3.1 Photobioreactor Light Transmission Simulations -- 12.3.2 Simulation Cost Saving Strategy -- 12.4 Surrogate Modelling and Bayesian Optimisation of an Aeration Driven Photobioreactor -- 12.4.1 Bayesian Optimisation Results -- 12.5 Conclusion -- References -- Chapter 13 Statistical Design of Experiments for Reaction Modelling and Optimisation -- 3.1 The Objective of the Sophorolipid Case Study -- 13.2 Why Empirical Learning? -- 13.3 The Final Model from the Sophorolipid Case Study -- 13.4 The Problem with One-factor-at-a-time Testing -- 13.5 The Complete Experimental Design -- 13.6 The First Stage: A Screening Design -- 13.7 The Second Stage: Axial Points to Estimate Curvilinear Effects -- 13.8 The Third Stage: Augmentation to Explore a New Factor Space -- 13.9 Chapter Summary -- References -- Chapter 14 Autonomous Synthesis and Self-optimizing Reactors -- 14.1 Introduction -- 14.2 Problem Definition -- 14.3 Elements of a Self-optimizing Reactor -- 14.3.1 Experiment Execution -- 14.3.2 Process Analytical Technology (PAT) -- 14.3.3 Algorithms -- 14.4 Applications -- 14.4.1 Nanoparticles -- 14.4.2 Polymer Synthesis -- 14.4.3 Organic Molecules -- 14.4.4 Outlook -- 14.5 Practical Case Studies -- 14.5.1 Case 1 -- 14.5.2 Case 2 -- 14.5.3 Case 3 -- References -- Chapter 15 Industrial Data Science for Batch Reactor Monitoring and Fault Detection -- 15.1 Introduction -- 15.1.1 Whom Is This Book Chapter for? -- 15.2 Industrial Batch Data and Analytics -- 15.2.1 Batch Data -- 15.2.2 Batch Data Alignment -- 15.2.3 Batch Data Analytics -- 15.3 Industrial Machine Learning Applications -- 15.3.1 Correlation Analysis for Batch Processes (Supervised Learning) -- 15.3.2 Anomaly Detection for Batch Processes (Unsupervised Learning) -- 15.4 Challenges and Opportunities -- 15.4.1 OT/IT Integration -- 15.4.2 Process Control vs. Advanced Analytics
- 10.3.1 Hybrid Model Structure -- 10.3.2 Kinetic Structure -- 10.3.3 Parameter Estimation -- 10.3.4 Hybrid Source Model Construction -- 10.3.5 Uncertainty Estimation -- 10.3.6 Hybrid Source Model Results -- 10.4 Hybrid Transfer Model -- 10.4.1 Hybrid Transfer Learning -- 10.4.2 Hybrid Transfer Model Construction -- 10.4.3 Benchmark Model Construction -- 10.4.4 Hybrid Transfer Model and Benchmark Model Results -- 10.5 Conclusion -- References -- Chapter 11 Constructing Time-varying and History-dependent Kinetic Models Via Reinforcement Learning -- 11.1 Recap for Kinetic and Hybrid Modelling -- 11.2 Modelling Time-varying and History-dependent Reactions -- 11.3 Introduction to Complex Reaction Case Studies -- 11.3.1 Scenario 1: Combinatorial Kinetic Systems -- 11.3.2 Scenario 2: History-dependent Kinetic Systems -- 11.3.3 Scenario 3: Time-varying Kinetic Systems -- 11.4 Reinforcement Learning for Hybrid Model Construction -- 11.4.1 Overview of the RL-based Hybrid Modelling Framework -- 11.4.2 Introduction to Reinforcement Learning -- 11.4.3 Integrating Reinforcement Learning with Hybrid Modelling -- 11.4.4 Framework Implementation -- 11.5 Results and Discussion -- 11.5.1 Results of Scenario 1 -- 11.5.2 Results of Scenario 2 -- 11.5.3 Results of Scenario 3 -- 11.5.4 Advantages and Limitations of the Framework -- 11.6 Conclusion -- References -- Part III: Data Intelligence and Industrial Applications -- Chapter 12 Surrogate and Multiscale Modelling for (Bio)reactor Scale-up and Visualisation -- 12.1 Multiscale Modelling of (Bio)reactors -- 12.2 Step I: Bioreactor Hydrodynamics and Turbulence Simulations -- 12.2.1 Numerical Solvers for BioreactorHydrodynamics and Turbulence Simulations -- 12.2.2 CFD Results of Bioreactor Hydrodynamics and Turbulence Simulations -- 12.3 Step 2: Coupling Bioreaction Kinetics to Hydrodynamic Simulations
- 3.3.4 Prediction Results Discussion -- 3.4 Limitations and Challenges for Complex System Simulation -- 3.4.1 Dealing with State Measurement Uncertainty -- 3.4.2 Requirements on Data Sampling Intervals -- 3.4.3 Challenge of Identifying the Best Hybrid Model Structure -- 3.4.4 Challenge of Inductive Bias and Hybrid Model Greyness -- 3.4.5 Simultaneous Parameter and Model Structure Identification -- 3.5 Chapter Summary -- References -- Chapter 4 Model Structure Identification -- 4.1 Introduction -- 4.2 Model Discrimination -- 4.2.1 Information Criteria -- 4.2.2 Case Study for Kinetic Model Selection -- 4.2.3 Model-based Design of Experiments -- 4.2.4 Conclusion -- References -- Chapter 5 Model Uncertainty Analysis -- 5.1 Introduction to Uncertainty Analysis -- 5.2 Case Study - Kinetic Model for Biomass Cultivation -- 5.3 Frequentist Paradigm -- 5.3.1 Frequentist-based Parameter Estimation -- 5.3.2 Individual Confidence Interval -- 5.3.3 Derivation of the Confidence Ellipsoid -- 5.3.4 Confidence Ellipsoid Estimation -- 5.4 Bayesian Paradigm -- 5.4.1 Maximum Likelihood Estimation -- 5.4.2 Maximum A Posteriori (MAP) -- 5.4.3 Bayesian Linear Regression -- 5.4.4 Bayesian Prediction for a Linear Model -- 5.4.5 Bayesian Prediction for a Nonlinear Model -- 5.5 Conclusion -- References -- Part II: Applications in Reaction Engineering -- Chapter 6 Interpretable Machine Learning for Kinetic Rate Model Discovery -- 6.1 Introduction -- 6.2 Problem Definition and Solution Strategy -- 6.3 Symbolic Regression -- 6.4 The Two-step Approach -- 6.5 Results -- 6.6 Conclusion -- References -- Chapter 7 Graph Neural Networks for the Prediction of Molecular Structure-Property Relationships -- 7.1 Introduction -- 7.2 Background -- 7.2.1 Quantitative Structure-Property Relationships -- 7.2.2 Neural Molecular Fingerprints
- Appendix A: Batch Data Alignment
- Cover -- Preface -- Contents -- Part I: Model Construction Theory -- Chapter 1 Physical Model Construction -- 1.1 Kinetic Model Construction -- 1.1.1 Empirical Kinetic Models -- 1.1.2 Chemical Kinetic Models -- 1.1.3 Biochemical Kinetic Models -- 1.2 Fundamentals of Deterministic Optimisation -- 1.2.1 Formulation and Optimality of Constrained Optimisation Problems -- 1.2.2 Solving Nonlinear Optimisation Problems -- 1.3 Dynamic Optimisation and Parameter Estimation -- 1.3.1 Formulation of Parameter Estimation Problems -- 1.3.2 Discretisation of Dynamic Process Constraints -- 1.4 Conclusion -- References -- Chapter 2 Data-driven Model Construction -- 2.1 A New Era of Chemical Reaction Kinetics -- 2.1.1 Data-driven Kinetic Modelling -- 2.1.2 A Catalyst for a Catalyst -- 2.2 Introduction to Machine Learning -- 2.2.1 How Do Machines Learn? -- 2.2.2 Parametric Learning -- 2.2.3 Non-parametric Learning -- 2.3 Basic Practice -- 2.3.1 Data Collection and Databases -- 2.3.2 Data Pre-processing -- 2.3.3 Model Verification -- 2.3.4 Don't Train Hard, Train Smart: The Concern of Overfitting -- 2.3.5 Hyperparameter Tuning -- 2.3.6 Model Performance Diagnostics -- 2.4 Introduction to Optimising Machine Learning Models -- 2.4.1 Optimisation Algorithms -- 2.4.2 Gradient-free Optimisation -- 2.4.3 Optimisation Within Modelling Languages -- 2.5 Conclusion -- References -- Chapter 3 Hybrid Model Construction -- 3.1 Hybrid Model Theory -- 3.1.1 Hybrid Model Motivation -- 3.1.2 Hybrid Modelling in Reaction Engineering -- 3.1.3 Hybrid Modelling in Bioprocess Engineering -- 3.2 Hybrid Model Structure -- 3.2.1 Discrepancy Hybrid Structure -- 3.2.2 Embedded Hybrid Structure -- 3.3 Hybrid Model Construction -- 3.3.1 Discrepancy Hybrid Model Construction -- 3.3.2 Embedded Hybrid Model Construction -- 3.3.3 Multistep-ahead and Uncertainty Analyses
- 7.3 Graph Neural Networks for Molecular Property Prediction -- 7.3.1 Molecular Graph -- 7.3.2 Message Passing -- 7.3.3 Readout -- 7.3.4 End-to- end Learning -- 7.4 Numerical Examples -- 7.4.1 Regression Example: Boiling Point Prediction -- 7.4.2 Classification Example: Biodegradability Prediction -- 7.5 Concluding Remarks -- Acknowledgements -- References -- Chapter 8 Reaction Network Simulation and Model Reduction -- 8.1 Introduction to Reaction Network Simulation -- 8.2 Microkinetic Modelling Approach -- 8.2.1 Microkinetic Model Construction -- 8.2.2 Physics- based Reaction Network Reduction -- 8.2.3 Mathematics-based Reaction Network Reduction -- 8.3 Pseudo-steady-state Modelling Approach -- 8.3.1 Pseudo-steady- state Model Construction -- 8.3.2 Some Remarks on Pseudo-steady- state Models -- 8.4 Summary -- References -- Chapter 9 Hybrid Modelling Under Uncertainty: Effects of Model Greyness, Data Quality and Data Quantity -- 9.1 Introduction -- 9.2 Fermentation Case Study -- 9.3 Three Structures for Three Levels of Greyness -- 9.3.1 Black Hybrid Model Structure -- 9.3.2 Grey Hybrid Model Structure -- 9.3.3 White Hybrid Model Structure -- 9.4 Hybrid Model Construction -- 9.4.1 Mitigating Overfitting Risks During Parameter Estimation -- 9.4.2 Propagated Uncertainty Estimation -- 9.5 Results and Discussion -- 9.5.1 Results of Hybrid Model Construction -- 9.5.2 Influence of Hybrid Model Greyness on Fitting Performance -- 9.5.3 Hybrid Model Temperature- shift Prediction Performance -- 9.6 Conclusion -- References -- Chapter 10 A Data- efficient Transfer Learning Approach for New Reaction System Predictive Modelling -- 10.1 Hybrid Transfer Learning Theory -- 10.1.1 Transfer Learning Motivation -- 10.1.2 Combining Transfer Learning with Hybrid Modelling -- 10.2 Case Study -- 10.2.1 Lutein Production -- 10.2.2 Problem Statement -- 10.3 Hybrid Source Model