Machine Learning Approaches for Predictive Maintenance in Infrastructure Systems

Predictive maintenance (PdM) has become an essential set of approaches to the optimization of the work and timeline of the infrastructure systems. Through the application of machine learning (ML) techniques, PdM systems are also capable of estimating equipment failures, which decreases both maintenance costs and downtime and increases system reliability. The chapter presents an aspect of different ML algorithms involved in PdM, such as supervised, unsupervised, and deep learning. Is covered in applications to transportation systems, energy systems, water treatment systems, and building systems, and how ML can be used to infer the future behavior of systems to anticipate failures, enhance efficiency in operations, and reduce unpredictable maintenance. Major complications like data quality, readability of models and scalability, are tackled and highlights of how they could be solved are given. The chapter also introduces the case studies of successful deployment of ML in the real-world infrastructure systems presenting quantitative findings and perspectives of future trends.