Optimal scheduling of manufacturing processes across multiple production lines by polynomial optimization and bagged bounded binary knapsack

Key objectives of ‘Industry 4.0’ methodologies in manufacturing include improved quality control, predictive maintenance and tracking. We consider whether a further objective that can easily and cost-effectively be achieved is the implementation of aggressive optimization of production schedules across multiple manufacturing lines. The present contribution describes how this objective can be achieved by exploiting operations research techniques for the supporting of plant operation scheduling. We focus on methodologies for optimizing energy and time consumption in particular. The data acquisition from networked sensors ends with the aggregation and insertion of measures in a relational database of multi-variate sample records representing the state of separate production lines. Starting from these records, production plans can be generated automatically. Plans can be represented as tables reporting sequences of the numbers of pieces of a specific type to be produced by each line. We describe schedule optimization methodologies in detail, reporting in particular on the costs of their computation. We finally discuss issues related to the design of the end user interface, which can usefully be based on a web service-oriented architecture meant to allow communicating the obtained results to human operators, finally allowing them to monitor and implement the resulting schedules. Results are evaluated on real data acquired by sensors installed in a metal injection molding plant in Bizkaia, Spain.