Apply AI and ML to solve complex problems across sciences
Many problems in physics, engineering, and applied sciences resist traditional modeling approaches. Artificial Intelligence in Science and Engineering: From Porous Materials to Drug Discovery presents AI and ML methods for tackling otherwise unsolvable problems in complex systems. Written by Muhammad Sahimi, who brings over 40 years of research experience to the topic, this reference spans multiple scientific domains.
The book covers AI and ML applications in hydrodynamics, porous media characterization, molecular dynamics simulation, and biological phenomena including protein folding. It addresses environmental applications and drug discovery, connecting computational methods with domain-specific challenges in fluid dynamics, materials science, and biology. Readers gain access to methods that model, predict, and optimize processes difficult to approach through conventional techniques.
Readers will also find:
- Detailed treatment of AI and ML approaches applied to complex systems in fluid dynamics and porous media research
- Coverage of molecular dynamics applications where machine learning accelerates simulation and prediction of material properties
- Methods for protein folding prediction and drug discovery leveraging current artificial intelligence and computational biology techniques
- Environmental science applications demonstrating how AI-driven modeling addresses problems resistant to traditional analytical methods
- Cross-disciplinary frameworks connecting physics, engineering, materials science, and biology through unified computational approaches
Physicists, materials scientists, engineers, computer scientists, and computational biologists will find this volume a substantive reference for applying AI and ML across their research domains. By unifying coverage of diverse complex systems under one framework, the book serves both academics and practitioners working at the intersection of computation and applied science.
Table of Contents:
Contents for Volume 1
Preface xvii
1 Artificial Intelligence and Complex Systems: What It Can and Cannot Do 1
1.1 Introduction 1
1.2 A Glance at History 2
1.3 Complex Media and Systems 4
1.4 Three Types of Complex Systems 5
1.5 Physics-informed and Data-driven Approach to Complex Media and Phenomena 6
1.6 What Artificial Intelligence Cannot Do 7
2 Neural Networks and Other Machine-learning Algorithms 9
2.1 Introduction 9
2.2 Training of Neural Networks: Backpropagation 11
2.3 Classification of Learning 15
2.4 Weak Learners and Boosting Algorithms 21
2.5 Activation Functions 23
2.6 Types of Neural Networks 25
2.7 Regularization of Neural Networks 43
2.8 Training of Large Neural Networks 44
2.9 Other Machine-learning Algorithms 45
2.10 Methods for Minimizing the Loss Function 48
2.11 Challenges and Future Directions 50
3 Solving Differential and Partial Differential Equations 59
3.1 Introduction 59
3.2 Solving Ordinary Differential Equations 61
3.3 Solving Partial Differential Equations 62
3.4 Solving High-dimensional Partial Differential Equations: Deep BSDE Algorithm 66
3.5 Feynman–Kac Solution for Backward Kolmogorov Equation of Stochastic Processes 69
3.6 Data-driven Discretization of Partial Differential Equations 72
3.7 Other Methods 77
3.8 Space-time Fractional Partial Differential Equations 79
3.9 Challenges and Future Directions 80
4 Fluid Mechanics: Single-phase Flow 85
4.1 Introduction 85
4.2 The Microscopic Conservation Laws 85Contents for Volume 1 ix
4.3 A Glance at History 88
4.4 Kinematics of Fluid Flow 89
4.5 Dynamics of Fluid Flow 94
4.6 Modeling Flow Systems of Type I 96
4.7 Data-driven Neural Networks for Flow Systems of Type I 99
4.8 Physics-informed and Data-driven Machine-learning Approach 109
4.9 Turbulent Flows 127
4.10 Control of a Flow Field 139
4.11 Aerodynamic Systems 141
4.12 Machine Learning for Accelerating Direct Numerical Simulations 143
4.13 Challenges and Future Directions 143
5 Fluid Mechanics: Multiphase Flows 155
5.1 Introduction 155
5.2 Physics-informed Simulation of Two-phase Flows 156
5.3 Data-driven Approach to Simulating Two-phase Flows 170
5.4 Multiphase Flow in Heterogeneous Porous Materials and Media 173
5.5 Challenges and Future Directions 174
6 Heat and Mass Transfer Processes 179
6.1 Introduction 179
6.2 Heat and Mass Transfer Processes 180
6.3 Applications of Neural Networks to Heat Transfer Processes 182
6.4 Mass Transfer 224
6.5 Challenges and Future Directions 229
7 Porous Materials and Media 241
7.1 Introduction 241
7.2 Characterization of Core-scale Porous Media 243
7.3 Characterization of Large-scale Porous Media 258
7.4 Reconstruction of Porous Media 261
7.5 Data-driven Neural Networks for Simulating Single-phase Flow and Transport Processes 268
7.6 Physics-informed Neural Networks for Simulating Single-phase Flow and Transport 280
7.7 Two-phase Flow 286
7.8 Thermo-hydro-mechanical Processes 298
7.9 Data-driven Neural Networks for Two-phase Flow 299
7.10 Challenges and Future Directions 302
8 Density-functional Theory and Molecular Simulation 313
8.1 Introduction 313
8.2 Quantum Monte Carlo Method 313
8.3 First-principle Simulation: Density-functional Theory Calculations 316
8.4 Molecular Dynamics Simulation 326
8.5 Active Learning 339
8.6 Other Aspects of Development of Force Fields by Machine-learning Algorithms 341
8.7 Challenges and Future Directions 343
9 Membranes for Separation of Fluid Mixtures 351
9.1 Introduction 351
9.2 Data-driven Neural Networks for Separation Processes 353
9.3 Data-driven Approach for Designing and Screening of Membranes' Materials 374
9.4 Application of Generative Adversarial Networks to Membrane Separation 380
9.5 Data-driven Neural Network for Minimizing Membrane Fouling 382
9.6 Physics-informed Modeling of Flow in Membranes 385
9.7 Challenges and Future Directions 386
10 Catalysis and Reaction Engineering 393
10.1 Introduction 393
10.2 Data-driven Machine-learning Algorithms for Predicting Catalytic Activity and Yield 395
10.3 Data-driven Machine-learning Algorithms for Design and Optimization of New Catalysts 405
10.4 Data-driven Neural Networks for Predicting Potential Energy Surface in Catalysis 414
10.5 Applications of Behler–Parrinello Generalized Neural-network Representation of High-dimensional Potential Energy Surfaces 424
10.6 Machine-learning Approach for Discovering and Designing New Catalysts Using Density Functional Theory Data 426
10.7 Machine-learning Algorithms for Identifying Catalytic Reaction Networks 438
10.8 Black-box, Grey-box, and Glass-box Methods 444
10.9 Challenges and Future Directions 444
Contents for Volume 2
Preface xiii
11 Materials Science 453
11.1 Introduction 453
11.2 Machine-learning Approach for Designing Polymers and other Macromolecules 455
11.3 Machine-learning Algorithms for Crystalline Solids 465
11.4 Generative Approach for Inverse Modeling of Material Discovery 491
11.5 Materials Interface 504
11.6 Challenges and Future Directions 506
12 Protein Structure 519
12.1 Introduction 519
12.2 Molecular Dynamics Simulation 520
12.3 Machine Learning for Coarse-grained Force Fields 522
12.4 Machine-learning Evolutionary Approach to Predicting Protein Structure 526
12.5 Neural Network Approach to Protein Structure 530
12.6 Challenges and Future Directions 563
13 Drug Discovery 573
13.1 Introduction 573
13.2 Machine-learning Methods for Quantitative Structure-property Relationships 575
13.3 Machine-learning Approach for Design of Antibacterial and Antimicrobial Peptides 578
13.4 Recurrent Neural Network Model for Drug Design 586
13.5 Design of Proteins for Neutralizing Lethal Snake Venom 589
13.6 Drugs for Viral Proteins of SARS-CoV-2 592
13.7 Machine Learning for Predicting Drug-target Interactions 594
13.8 Challenges and Future Directions 597
14 Medical Imaging and Anatomical Diagnosis 603
14.1 Introduction 603
14.2 Image Classification 606Contents for Volume 2 ix
14.3 Detection 608
14.4 Segmentation 610
14.5 Registration 614
14.6 Image Enhancement 615
14.7 Extracting Features from a Medical Image 615
14.8 Leveraging Textual Reports to Improve Classification of Medical Images 616
14.9 Textual Description of Medical Images 616
14.10 Anatomical Applications 619
14.11 Commonalities of Images of Porous Media and Biological Organs 639
14.12 Challenges and Future Directions 641
15 Environmental and Climate Sciences 651
15.1 Introduction 651
15.2 Hydrology 652
15.3 Transport of Contaminants in Groundwater 673
15.4 Soil Moisture 677
15.5 Carbon Dioxide Storage in Porous Formations 681
15.6 Climate Models 688
15.7 Challenges and Future Directions 704
16 Learning Governing Equations for Datasets 713
16.1 Introduction 713
16.2 Type-II Systems 715
16.3 Type-III Systems 745
16.4 Kernel Methods 771
16.5 Challenges and Future Directions 772
References 773
Postface 784
Index 785
About the Author :
Muhammad Sahimi, PhD, is a professor of chemical engineering and materials science at the University of Southern California. With over 40 years of experience specializing in porous media, heterogeneous materials, and the application of AI and ML methods, he has published more than 400 peer-reviewed articles and four books.
