Online Parameterization of a Milling Force Model using an Intelligent System Architecture and Bayesian Optimization

This paper presents an efficient and industry-ready system architecture that enables both the control of machining operations and the high-frequency acquisition of controller data and external sensor signals. Using the recorded data, a dexel-based mechanistic cutting force model, which enables the estimation of cutting forces for complex tool geometries in arbitrary machining operations, is parameterized via an instantaneous cutting force identification method. In the introduced identification method, Bayesian Optimization and Dynamic Time Warping are combined to avoid time-consuming and error-prone synchronization of measurement and simulation. The suitability of this approach was demonstrated by performing cutting experiments at various radial depths of cut and feed rates per tooth. Thereby, a good agreement between simulation and measurement could be observed.