Machine learning (ML) is a field of artificial intelligence that develops algorithms capable of learning patterns from data and making predictions without being explicitly programmed. In earthquake engineering, it is increasingly used for damage detection, fragility modeling, rapid post-earthquake assessment, and surrogate modeling.
Machine learning (ML) is a subfield of artificial intelligence that develops algorithms capable of improving their performance on a task through experience β typically by learning patterns from data. Unlike traditional programming, in which a human specifies the rules that map inputs to outputs, machine learning algorithms infer these rules from examples. The field has exploded over the past two decades, driven by the availability of large datasets, advances in algorithm design, and dramatic increases in computational power (especially GPUs). Applications span from image recognition and natural language processing to scientific discovery, autonomous systems, and increasingly, earthquake engineering.
Machine learning is classified into several paradigms. Supervised learning uses labeled data (input-output pairs) to learn a mapping from inputs to outputs β examples include regression (predicting a continuous value) and classification (predicting a category). Unsupervised learning finds structure in unlabeled data β examples include clustering, dimensionality reduction, and anomaly detection. Semi-supervised learning combines a small amount of labeled data with a large amount of unlabeled data. Reinforcement learning learns through interaction with an environment, receiving rewards or penalties for actions β used in control, robotics, and sequential decision-making. In earthquake engineering, supervised learning is the dominant paradigm, with applications including damage detection (classifying structures as damaged or undamaged), fragility modeling (predicting the probability of damage as a function of intensity), and surrogate modeling (approximating expensive simulations with fast ML models).
Machine learning is transforming earthquake engineering practice and research. In structural health monitoring (SHM), ML algorithms β particularly deep learning and computer vision β are used to detect, locate, and quantify damage from sensor data and images. In seismic hazard and risk analysis, ML is used to develop GMPEs, to classify ground motions, and to generate rapid post-earthquake assessments. In performance-based earthquake engineering, ML surrogate models are used to accelerate nonlinear response history analysis, making it feasible to evaluate thousands of design alternatives. In computer vision, deep learning models are used for crack detection, post-earthquake damage classification from satellite or drone imagery, and structural component recognition. Challenges in applying ML to earthquake engineering include: (1) limited availability of labeled data, especially for rare events like strong earthquakes; (2) the need for physics-informed approaches that respect known physical constraints; (3) the interpretability of complex models, which is critical for engineering decision-making; (4) the transferability of models across different structures, regions, and ground motion characteristics; and (5) the computational cost of training large models. Modern research increasingly combines ML with physics-based modeling β physics-informed machine learning β to leverage the strengths of both approaches. In Iran, ML is increasingly used in research on SHM, damage detection, and rapid post-earthquake assessment, with growing interest in applying ML to the analysis of the large datasets produced by Iran's seismic and strong-motion networks.