Improving production and maintenance planning with meta-learning-based failure prediction

Abstract

Digital manufacturing systems allow for dynamic production and maintenance planning adaptation based on internal and external evolving changes. In this context, data-driven predictive techniques can improve maintenance planning. However, dynamic production environments require the development of novel methods capable of efficiently managing the extensive data generated by manufacturing systems and processes while providing valuable insights for decision-making. Thus, this research proposes an integrated production and maintenance planning (IPMP) approach for digital manufacturing systems featuring a dynamic simulation-based optimization and a meta-learning-based failure prediction, which selects the best prognostic method to predict machine failures in real-time. The proposed model was applied to the Minifab case, in the semiconductor industry, providing a reduction of 20.9 % in lead time and 25.9 % in tardy jobs compared to setups with static dispatching rules and single-prognostic-method failure prediction. Such findings support the pertinence of the proposed novel approach in the area of production planning and control, contributing to the improvement of business performance, reliability, and competitiveness.