EXPLAINABLE ARTIFICIAL INTELLIGENCE IN STRUCTURAL ENGINEERING: APPLICATIONS, OPPORTUNITIES, AND CHALLENGES

Abstract

This comprehensive literature review examines the integration of Explainable Artificial Intelligence (XAI) within structural engineering, focusing on its applications, opportunities, and inherent challenges. The rapid advancement of machine learning and deep learning models has revolutionized structural analysis, design, and health monitoring; however, their "black-box" nature has limited widespread adoption in safety-critical applications. XAI techniques such as SHAP (Shapley Additive exPlanations), LIME (Local Interpretable Model-agnostic Explanations), and Grad-CAM address this transparency gap by providing interpretable insights into model decision-making processes. This review synthesizes findings from over 45 peer-reviewed studies published between 2020 and 2025, analyzing XAI applications across concrete strength prediction, structural damage detection, reinforcement design, and real-time health monitoring systems. Results demonstrate that XGBoost and Random Forest algorithms, when coupled with XAI techniques, achieve accuracy levels exceeding 97% while maintaining full interpretability. Key opportunities include enhanced regulatory compliance, improved engineer trust, and accelerated structural optimization. Critical challenges involve computational complexity, dataset heterogeneity, and the need for standardized XAI evaluation metrics. This review advocates for integrated frameworks combining physics-informed neural networks with XAI techniques to develop trustworthy, transparent, and deployable AI systems in structural engineering.