Infrastructure damage assessment via machine learning approaches: a systematic review

Monitoring civil infrastructures to detect early damage and extract the data required for urban management can prevent sudden infrastructure collapse, increase infrastructure sustainability and service life, and facilitate the management of smart cities. Machine learning (ML) approaches have sparked a great interest in structural health monitoring in recent years due to their superior potential to damage and deficiencies diagnosis in civil engineering infrastructures. ML can efficiently perform several analyses of regression, clustering, and classification of damage in diverse infrastructures, including pavements, bridges, dams, railways’ tunnels, and wind turbines. In this systematic review, the diverse ML algorithms used in this domain have been discussed. The efficacy of deploying ML approaches in infrastructure health monitoring systems has been assessed and a detailed critical analysis of ML applications in damage diagnosis has been provided. Furthermore, the knowledge gaps, challenges and future trends have been outlined to integrate independent ML approaches into systems and develop field-applicable ML algorithms that convince both researchers and end-users to adopt them. This critical review also paves the researchers’ way for decision-making regarding the adoption and development of applicable ML and deep learning (DL) approaches in the domain of infrastructural monitoring, particularly, for smart structures/systems and detection methods.