THE ROLE OF MACHINE LEARNING(ML)-DRIVEN-DIGITAL TWINS FOR SMART MANUFACTURING PREDICTIVE MAINTENANCE: A REVIEW

Abstract

Predictive maintenance is one of the approaches that can allow making the industry more sustainable, safe, and profitable.  Considering the implementation of machine learning (ML), digital twin (DT) and the Industrial Internet of Things (IIoT), this paper is going to reflect on the advances of predictive maintenance (PdM) in a connected manufacturing.  Aspects of PdM such as data gathering, sensor connections, and forecasting modelling are discussed, albeit with more focus on condition-based maintenance, forecasting anomalies, and real-time fault detection. The paper discusses various ML mechanisms supervised, unsupervised and reinforcement learning, and the contributions they make to the monitoring of equipment health and decisions. The applicability regarding dynamic digital twin applications is discussed to simulate, predict and optimise the expected system performance without interference with the current process in the system. A comprehensive literature review presents cutting-edge PdM frameworks, including cloud–edge hybrid architectures, multi-agent systems, adaptive AI models. Challenges such as implementation complexity, scalability, and data quality limitations are discussed alongside solutions aimed at interoperability, lightweight modeling, and cross-domain adaptability. According to the results, PdM is an essential component of the industry 4.0 revolution since it can drastically cut down on downtime, increase the lifespan of equipment, and improve operational efficiency through the use of AI, DT, and the Internet of Things.