Badra / Pal / Pei Artificial Intelligence and Data Driven Optimization of Internal Combustion Engines

1. Auflage 2022
ISBN: 978-0-323-88458-7
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: 6 - ePub Watermark

E-Book, Englisch, 262 Seiten

Artificial Intelligence and Data Driven Optimization of Internal Combustion Engines summarizes recent developments in Artificial Intelligence (AI)/Machine Learning (ML) and data driven optimization and calibration techniques for internal combustion engines. The book covers AI/ML and data driven methods to optimize fuel formulations and engine combustion systems, predict cycle to cycle variations, and optimize after-treatment systems and experimental engine calibration. It contains all the details of the latest optimization techniques along with their application to ICE, making it ideal for automotive engineers, mechanical engineers, OEMs and R&D centers involved in engine design. - Provides AI/ML and data driven optimization techniques in combination with Computational Fluid Dynamics (CFD) to optimize engine combustion systems - Features a comprehensive overview of how AI/ML techniques are used in conjunction with simulations and experiments - Discusses data driven optimization techniques for fuel formulations and vehicle control calibration

Autoren/Hrsg.

Badra, Jihad

Pal, Pinaki

Pei, Yuanjiang

Som, Sibendu

Inhaltsverzeichnis

1;Front Cover;1
2;ARTIFICIAL INTELLIGENCE AND DATA DRIVEN OPTIMIZATION OF INTERNAL COMBUSTION ENGINES;2
3;ARTIFICIAL INTELLIGENCE AND DATA DRIVEN OPTIMIZATION OF INTERNAL COMBUSTION ENGINES;4
4;Copyright;5
5;Contents;6
6;Contributors;10
7;Foreword;12
8;Preface;16
9;1 - Introduction;18
9.1;1. Industrial revolution;18
9.2;2. Artificial intelligence, machine learning, and deep learning;19
9.3;3. Machine learning algorithms;20
9.4;4. Artificial intelligence-based fuel-engine co-optimization;21
9.4.1;4.1 Optimization of internal combustion engine;21
9.4.1.1;4.1.1 Design of experiments;22
9.4.1.2;4.1.2 Genetic algorithm;24
9.4.1.3;4.1.3 Machine learning-based algorithms;25
9.4.2;4.2 Optimization of fuel formulation;30
9.4.3;4.3 Mitigation of rare combustion events;32
9.5;5. Summary;33
9.6;References;33
10;1 - Artificial Intelligence to optimize fuel formulation;42
10.1;2 - Optimization of fuel formulation using adaptive learning and artificial intelligence;44
10.1.1;1. Introduction and motivation;44
10.1.2;2. Mixed-mode combustion and fuel performance metrics;45
10.1.3;3. A neural network model to predict fuel research octane numbers;48
10.1.4;4. Optimization problem formulation and description of solution approaches;49
10.1.4.1;4.1 Constrained optimization formulation;49
10.1.4.2;4.2 Genetic algorithm;50
10.1.4.3;4.3 Gaussian process–based surrogate model optimization algorithm;52
10.1.5;5. Numerical experiments and results;54
10.1.6;6. Discussion;57
10.1.7;7. Summary and concluding remarks;59
10.1.8;Acknowledgments;60
10.1.9;References;60
10.2;3 - Artificial intelligence–enabled fuel design;64
10.2.1;1. Transportation fuels;64
10.2.1.1;1.1 Fuel representation;64
10.2.1.2;1.2 Fuel formulation workflow;65
10.2.1.3;1.3 Artificial intelligence modeling approaches;66
10.2.2;2. Application of artificial intelligence to fuel formulation;69
10.2.2.1;2.1 High throughput screening: finding a needle in the haystack;69
10.2.2.2;2.2 Fuel property prediction by machine learning models;71
10.2.2.3;2.3 Reaction discovery;74
10.2.2.4;2.4 Fuel-engine co-optimization;75
10.2.3;3. Conclusions and perspectives;75
10.2.4;Acknowledgments;77
10.2.5;References;77
11;2 - Artificial Intelligence and computational fluid dynamics to optimize internal combustion engines;86
11.1;4 - Engine optimization using computational fluid dynamics and genetic algorithms;88
11.1.1;1. Introduction;88
11.1.2;2. Modeling framework and acceleration strategies;91
11.1.2.1;2.1 Computational fluid dynamics acceleration techniques;91
11.1.2.1.1;2.1.1 Adaptive mesh refinement;91
11.1.2.1.2;2.1.2 Detailed chemistry acceleration strategies;92
11.1.2.2;2.2 Engine geometry generation;93
11.1.2.2.1;2.2.1 Method of splines;93
11.1.2.2.2;2.2.2 Method of forces;94
11.1.2.3;2.3 Virtual injection model;95
11.1.3;3. Optimization methods;96
11.1.3.1;3.1 Fundamentals of genetic algorithms;96
11.1.3.2;3.2 Pioneering investigations;98
11.1.3.3;3.3 Multiobjective framework;101
11.1.3.4;3.4 Convergence acceleration;108
11.1.4;4. Summary and concluding remarks;114
11.1.5;References;115
11.2;5 - Computational fluid dynamics–guided engine combustion system design optimization using design of experiments;120
11.2.1;1. Introduction;120
11.2.2;2. Methodologies;123
11.2.2.1;2.1 Design space construction;124
11.2.2.2;2.2 Response surface model formulation;126
11.2.2.3;2.3 Model-based design optimization and verification;129
11.2.3;3. A recent application;130
11.2.3.1;3.1 Engine and fuel specifications;130
11.2.3.2;3.2 Computational fluid dynamic model setup and validation;130
11.2.3.3;3.3 Design variables;131
11.2.3.4;3.4 Objective variables and evaluation method;133
11.2.3.5;3.5 Data fitting and optimization;134
11.2.4;4. Recommendations for best practice;135
11.2.4.1;4.1 Adequate computational fluid dynamic model validation;135
11.2.4.2;4.2 Efficient geometry and mesh manipulation;136
11.2.4.3;4.3 Sample size;136
11.2.4.4;4.4 Optimization across full engine operation range;136
11.2.4.5;4.5 Computational efficiency;136
11.2.5;5. Conclusions and perspectives;137
11.2.6;Acknowledgments;138
11.2.7;References;138
11.3;6 - A machine learning-genetic algorithm approach for rapid optimization of internal combustion engines;142
11.3.1;1. Introduction;142
11.3.2;2. Engine optimization problem setup;144
11.3.3;3. Training and data examination;146
11.3.4;4. Machine learning-genetic algorithm approach;149
11.3.4.1;4.1 Optimization methodology;149
11.3.4.2;4.2 Repeatability of machine learning-genetic algorithm;151
11.3.4.2.1;4.2.1 Extension of variable domain;153
11.3.4.3;4.3 Postprocessing and robustness;156
11.3.5;5. Automated machine learning-genetic algorithm;158
11.3.5.1;5.1 Hyperparameter selection;159
11.3.5.1.1;5.1.1 Manual selection;159
11.3.5.1.2;5.1.2 Automated strategies for selecting hyperparameters;160
11.3.5.2;5.2 Problem setup;162
11.3.5.3;5.3 Results;163
11.3.6;6. Summary;173
11.3.7;Acknowledgments;173
11.3.8;References;173
11.4;7 - Machine learning–driven sequential optimization using dynamic exploration and exploitation;176
11.4.1;1. Introduction;176
11.4.2;2. Active ML optimization (ActivO);177
11.4.2.1;2.1 Basic algorithm;177
11.4.2.2;2.2 Query strategies;178
11.4.2.3;2.3 Convergence criteria;180
11.4.2.4;2.4 Dynamic exploration and exploitation;181
11.4.3;3. Case study 1: two-dimensional cosine mixture function;182
11.4.4;4. Case study 2: computational fluid dynamics (CFD)-based engine optimization;188
11.4.5;5. Conclusions;196
11.4.6;Acknowledgments;197
11.4.7;References;197
12;3 - Artificial Intelligence to predict abnormal engine phenomena;200
12.1;8 - Artificial-intelligence-based prediction and control of combustion instabilities in spark-ignition engines;202
12.1.1;1. Introduction;202
12.1.1.1;1.1 Artificial intelligence applications to engine controls;202
12.1.1.2;1.2 Dilute combustion instability background;204
12.1.2;2. Case study: artificial-intelligence-enhanced modeling of dilute spark-ignition cycle-to-cycle variability;206
12.1.3;3. Case study: neural networks for combustion stability control;210
12.1.3.1;3.1 Artificial neural networks;210
12.1.3.2;3.2 Spiking neural networks;212
12.1.4;4. Case study: learning reference governor for model-free dilute limit identification and avoidance;216
12.1.4.1;4.1 Constrained combustion phasing control problem;216
12.1.4.2;4.2 Learning reference governor for avoiding misfire events;219
12.1.5;5. Summary;221
12.1.6;References;222
12.2;9 - Using deep learning to diagnose preignition in turbocharged spark-ignited engines;230
12.2.1;1. Introduction;230
12.2.1.1;1.1 Fault detection;230
12.2.1.2;1.2 Optimization and control;231
12.2.1.3;1.3 Predicting combustion parameters (phasing and cycle-to-cycle variation) and emissions;232
12.2.2;2. Preignition detection using machine learning algorithm;232
12.2.2.1;2.1 Feed forward multilayer neural networks;234
12.2.2.2;2.2 Convolutional neural networks;235
12.2.2.3;2.3 Recurrent neural networks;236
12.2.3;3. Activation functions;238
12.2.4;4. Experiments and data extraction;239
12.2.5;5. Machine learning methodology;241
12.2.6;6. Model 1: Input from principal component analysis;247
12.2.7;7. Model 2: Time series input;248
12.2.8;8. Model metrics;249
12.2.9;9. Results and discussion;250
12.2.9.1;9.1 Training and validation losses;250
12.2.10;10. Conclusions;251
12.2.11;References;252
12.2.12;Further reading;253
13;Index;256
13.1;A;256
13.2;B;257
13.3;C;257
13.4;D;257
13.5;E;258
13.6;F;258
13.7;G;258
13.8;H;258
13.9;I;258
13.10;K;259
13.11;L;259
13.12;M;259
13.13;N;259
13.14;O;260
13.15;P;260
13.16;Q;260
13.17;R;260
13.18;S;260
13.19;T;260
13.20;U;260
13.21;V;260
13.22;Z;260
14;Back Cover;262