ML2-enabled Condition-based Demand, Production, Inventory, and Maintenance Planning

Abstract

Production planning and control is pivotal to meeting customer demand and maximizing profit. At the same time, machine breakdowns compromise these goals, which can be tackled with a good maintenance strategy. Here, advances in condition-based maintenance and prognostics and health management allow predicting the health state of production machines through sensor data and prescribing optimal demand, production, inventory, and maintenance plans. Here, machine learning (ML) is promising for accurate health predictions using sensor data and decision-making in complex, highly dynamic production environments. Thus, in this work, two ML algorithms are applied. First, a data-driven regression algorithm predicts the health of a machine. This forecast is forwarded to a reinforcement learning algorithm (i.e. proximal policy optimization, recently made famous by its application within ChatGPT) to optimize demand, production, inventory, and maintenance plans. A computational study shows excellent performances of the ML-based health prediction and planning algorithms, which surpass traditional maintenance strategies.