Machine Learning for Seismic Vulnerability Assessment: A Review

Assessing seismic vulnerability is essential for reducing the impact of earthquakes on people and structures. The prevalence of machine learning techniques in this domain has been growing. However, there is a need for further research on their efficacy and constraints. This paper provides a systematic review that evaluates the use of machine learning for seismic vulnerability assessment. This study thoroughly examines literature published from 2017 to 2023 to identify the latest machine learning techniques, evaluate their effectiveness, and investigate the challenges and future research areas they present. The review highlights the potential advantages of ML-based approaches, such as decreased computational expenses and reliable earthquake damage predictions, compared to traditional methods. The reliability of ML models is a concern due to their dependence on input datasets. Enhancing the accessibility and standard of annotated datasets is essential in addressing this obstacle. Improving the interpretability of ML models and establishing standardized evaluation frameworks are critical areas for enhancement. The study offers insights into the present and future potential of machine learning in assessing seismic vulnerability. This review can provide researchers and structural engineers valuable insights into earthquake engineering and disaster risk reduction. The potential of ML in seismic vulnerability assessment can be enhanced by addressing research gaps, including improving dataset quality, enhancing interpretability, and establishing an evaluation framework. The research seeks to enhance the field by aiding in developing more efficient strategies for reducing the effects of earthquakes on societies and structures.