Artificial Intelligence in Performance-Driven Design

Narjes Abbasabadi, PhD, is an Assistant Professor in the Department of Architecture at the University of Washington. Dr. Abbasabadi also leads the Sustainable Intelligence Lab (SIL). Her research centers on sustainability and computation within the built environment. Abbasabadi’s primary focus is advancing design research through the development of data-driven and physics-based methods, frameworks, and tools that leverage digital technologies, including artificial intelligence and machine learning, to enhance performance-based and human-centered design. With an emphasis on multi-scale exploration, her research investigates urban building energy flows, human systems, and environmental impacts across scales—from the scale of building to the scale of neighborhood and city. Abbasabadi’s research has been published in leading journals, including Applied Energy, Building and Environment, Energy and Buildings, Environmental Research, and Sustainable Cities and Society. Abbasabadi earned a Ph.D. in Architecture with a specialization in Technologies of the Built Environment, from the Illinois Institute of Technology, and holds Master’s and Bachelor’s degrees in Architecture from Tehran Azad University. Mehdi Ashayeri, PhD, is an Assistant Professor in the School of Architecture at Southern Illinois University, where he leads the Urban Intelligence and Integrity Lab (URBiiLAB). Ashayeri earned his Ph.D. in Architecture–Technologies of the Built Environment, from the Illinois Institute of Technology. He also holds an M.Sc. in Architectural Engineering and a B.Sc. in Civil Engineering from Tehran Azad University. Dr. Ashayeri’s research is centered on environmental performance and computing, with a strong emphasis on their implications for human health and justice. This involves developing frameworks, tools, and digital platforms using data-driven techniques including artificial intelligence, machine learning, natural language processing, Big data, and sensing, as well as physics-based simulation methodologies. In recent projects, Ashayeri has specifically explored spatiotemporal modeling, energy performance evaluation, assessment of exposure to air pollution, and the integration of human feedback systems across various scales. These studies are designed to facilitate data-informed decision-making for human-centered design, as well as to contribute to the development of sustainable buildings and cities. Ashayeri’s research has been published in high-impact journals, including Environmental Research, Energy and Buildings, Applied Energy, Building and Environment, and Sustainable Cities and Society.

List of Contributors xi

Introduction xiii

1 Augmented Computational Design 1

Introduction 1

Background 2

Relevance of AI in AEC 2

Historical Context 3

Design as Decision-Making 5

AI for Generative Design 7

Framework 9

Design Space Exploration 11

Spatial Design Variables 13

Statistical Approaches to Design 14

Demonstration 15

Case Study 15

Methodology 16

Results 21

BBN Validation Results 21

Toy Problem 22

Discussion 22

Outlook 25

Acronyms 26

Notations 27

References 28

2 Machine Learning in Urban Building Energy Modeling 31

Introduction 31

Urban Building Energy Modeling Methods 32

Top–Down Models 33

Bottom–Up Models 33

Uncertainty in Urban Building Energy Modeling 36

Epistemic Uncertainty 36

Stochastic Uncertainty 36

Addressing Uncertainty 37

Machine Learning in Urban Building Energy Modeling 39

Supervised Learning 39

Unsupervised Learning 44

Reinforcement Learning 46

Machine Learning-Based Surrogate UBEM 47

Conclusion 49

References 50

3 A Hybrid Physics-Based Machine Learning Approach for Integrated Energy and Exposure Modeling 57

Introduction 57

Materials and Methods 59

Data, Data Sources, and Dataset Processing 59

Methodology 61

Results 70

Physics-Based Simulation 70

Data-Driven Computation (Prediction) 70

Discussion 73

Conclusion 74

Acknowledgment 75

References 75

4 An Integrative Deep Performance Framework for Daylight Prediction in Early Design

Ideation 81

Introduction 81

Background 83

Daylight Simulation 84

Deep Learning Models 85

DL-Based Surrogate Modeling 85

Verification Methods 85

Research Methods 86

Data Acquisition 86

Model Training 88

Results and Validation 88

Discussions of Results 90

Conclusions 94

References 94

5 Artificial Intelligence in Building Enclosure Performance Optimization: Frameworks, Methods, and Tools 97

Building Envelope and Performance 97

Artificial Intelligence and Building Envelope Overview 97

Optimization Routes and Building Envelope 98

Optimization Frameworks 99

Optimization Methods 99

Machine Learning and Building Envelope 101

Artificial Neural Network 101

Convolutional Neural Network 105

Recurrent Neural Network 105

Generative Adversarial Networks 106

Ensemble Learning 107

Discussions on Practical Implications 108

Summary and Conclusion 109

References 110

6 Efficient Parametric Design-Space Exploration with Reinforcement Learning-Based Recommenders 113

Introduction 113

Methodology 115

Section 01: Clustering Design Options 116

Section 02: Reinforcement Learning-Based Recommender System 120

Design Dashboard 123

Discussion 124

Conclusion 125

References 126

7 Multi-Level Optimization of UHP-FRC Sandwich Panels for Building Façade Systems 129

Introduction 129

Building Façade Design Optimization 130

Methodology 134

Midspan Displacements and Thermal Resistivity of UHP-FRC Panels 136

Energy Performance of the UHP-FRC Panels at the Building Level 141

Life Cycle Cost Analysis of the UHP-FRC Panels 142

Surrogate Models 145

Multi-objective Optimization Algorithm 147

Results and Discussion 148

Surrogate Models 148

Pareto Front Solutions 151

Conclusion 152

References 153

8 Decoding Global Indoor Health Perception on Social Media Through NLP and Transformer Deep Learning 159

Introduction 159

Literature Review 161

Social Media and Urban Life: Theories, Challenges, and Opportunities 161

Methods for Computing Social Media Data in Environmental Studies 163

Materials and Methods 168

Data Query 168

Text Preprocessing 169

Text Tokenization 169

Text Summarization 170

Generating Co-occurrence Matrix 170

Sentiment Analysis and Classification 170

Visualizations 171

Embedding Visualization 171

Attention Score Visualization (Attention Map) and Interpretation 172

Results and Discussion 173

Conclusion 178

References 179

9 Occupant-Driven Urban Building Energy Efficiency via Ambient Intelligence 187

Introduction 187

Occupancy and Building Energy Use 191

Definitions 191

Occupant Monitoring Methods 193

Occupant Monitoring Via Observational Studies 194

Occupant Monitoring via Experimental Studies 195

Occupant-driven Energy Efficiency via Ambient Intelligence 196

Ambient Intelligence Advancements and Applications 196

AmI-Based Energy Efficiency Feedback (EEF) Systems 197

Energy Efficiency via AmI Systems and Digital Twins Technology 201

Conclusion 202

References 203

10 Understanding Social Dynamics in Urban Building and Transportation Energy Behavior 211

Introduction 211

Methodology 213

Modeling Framework 214

Explanatory Model 214

Data 215

Results and Discussion 219

Effects of Occupancy and Socio-economic Factors 219

Variable Importance (VI) 219

Lek’s Profile 219

Conclusion 226

References 227

11 Building Better Spaces: Using Virtual Reality to Improve Building Performance 231

Introduction 231

Applications of Virtual Reality in Building Performance 233

Virtual Reality for Improving Building Design through Integrated Performance Data 233

Virtual Reality for Building Design Reviews and Education in Architecture and Engineering 236

Virtual Reality for Research on Building Occupant Comfort and Well-Being 240

Conclusion 243

References 245

12 Digital Twin for Citywide Energy Modeling and Management 251

Introduction 251

Urban Building Energy Digital Twins (UBEDTs) 252

Definition and Conceptualization 252

Implications for Citywide Energy Management 254

Enabling Technologies 256

Twining Technologies 256

Urban Digital Twin(UDT) and Data Sources 258

Artificial Intelligence (AI) and Digital Twin 260

Relationship Between IoT, Big Data, AI–ML, and Digital Twins 261

Interoperability Technologies 262

Maturity Levels 263

Architecture 265

Data Acquisition Layer 266

Transmission Layer 266

Modeling and Simulation Layer 266

Data/Model Integration Layer 269

Service/Actuation Layer 269

Challenges in Implementing Citywide Digital Twins 269

Data Quality and Availability 270

Required Smart Infrastructure and Associated Cost 270

Interoperability 270

Data Analysis 271

Cybersecurity and Privacy Concerns 271

Conclusion 272

References 272

Index 277