Data-Driven Integrated Production and Maintenance Optimization

We propose a data-driven integrated production and maintenance planning model, where machine breakdowns are subject to uncertainty and major sequence-dependent setup times occur. We address the uncertainty of breakdowns by considering various covariates and the combinatorial problem of sequence-dependent setup times with an asymmetric Traveling Salesman Problem (TSP) approach. The combination of the TSP with machine learning optimizes the production planning, minimizing the non-value creating time in production and thus, overall costs. A data-driven approach integrates prediction and optimization for the maintenance timing, which learns the influence of covariates cost-optimal via a mixed integer linear programming model. We compare this approach with a sequential approach, where an algorithm predicts the moment of machine failure. An extensive numerical study presents performance guarantees, the value of data incorporated into decision models, the differences between predictive and prescriptive approaches and validates the applicability in practice with a runtime analysis. We show the model contributes to cost savings of on average 30% compared to approaches not incorporating covariates and 18% compared to sequential approaches. Additionally, we present regularization of our prescriptive approach, which selects the important features, yielding lower cost in 80% of the instances.