Maintenance 4.0 Model Development for Production Lines in Industry 4.0 Using a Deep Learning Approach and IoT Data in Real-Time: an Experimental Case Study

Abstract

The current buzz in research and industry circles revolves around digitalization and digitization. By incorporating new technologies, companies can now collect data in real-time, which can be processed at a later stage. The manufacturing sector, in particular, is showing a keen interest in adopting maintenance 4.0 to enhance energy efficiency, optimize machine availability and utilization rate, extend equipment life, and forecast equipment failures before they occur. Such proactive measures help prevent breakdowns and minimize downtime, thereby improving productivity. This study focuses on a data-driven predictive maintenance system developed to implement effective maintenance 4.0. The system aims to first acquire real-time data from various sensors using an IoT-based system, second to detect failures before they occur, and finally to estimate the remaining useful life (RUL) using artificial neural networks of the LSTM type. Therefore, our methodology facilitates the prediction of defects in rotating machines, enabling timely and preventive interventions to address issues and prevent production interruptions. We conducted a case study evaluation to assess the efficacy of our proposed approach. The outcomes demonstrated the system's proficiency in recognizing failure indicators and its role in anticipating production halts resulting from defects. Moreover, experimental results highlighted its superior accuracy in Remaining Useful Life (RUL) estimation and reduced prediction errors in comparison to various algorithms explored in the literature, including RNN models, XGBoost, and boosting techniques.