Demand-Oriented Optimization of Machine Tools: a Closed Loop Approach for Safe Exploration in Series Production

Abstract

The resource- and energy-efficient operation of machine tools promises significant economic and ecological benefits. However, in the context of series production, optimization of the operating conditions can cause far-reaching consequences for the entire production chain. This paper presents a method for the safe exploration and optimization of new operating parameters on machine tools while ensuring process safety at all times. The method iteratively expands the allowable state space based on predictions of future Overall Equipment Effectiveness, while a Bayesian optimizer identifies the optimal operating points. A statistical verification of clear decision rules further safeguards the optimization and makes risks measurable. The method was tested by optimizing the energy demand on a grinding machine at Mercedes-Benz AG in series production, where it achieved 15% savings without compromising process safety at any point.