Data-Driven Prediction of Plunge Milling Force Using Stochastic Differential Equations

Plunge milling constitutes a critical procedure in industrial machinery fabrication, with its applications extending to vital sectors such as aerospace, automotive, and maritime industries. As the demand for precision in plunge milling operations grows, precise modelling of cutting forces is essential for the systematic optimization of CNC milling processes and parameters, including but not limited to feed rate, cutting depth, and others. Current cutting force modelling methodologies that utilize spindle drive current and vibration data often assume constant noise levels. Nevertheless, in many mechanical manufacturing systems, process noise is dependent on system states and inputs, suggesting that the dynamics of cutting force are influenced by more complex multiplicative noise. In response to this complexity, this paper introduces an approach that employs stochastic differential equations to more accurately characterize the dynamics of cutting force. Specifically, we propose an innovative Bayesian method for parameter inference within the stochastic framework. When benchmarked against deep learning models using real measured data, our method demonstrates better accuracy and robustness.