Online Parameter Estimation for Model-Based Force Control in Milling Processes

摘要

In order to adapt to a milling machine’s changing behavior while operating due to changing tool-workpiece contact conditions adaptive controllers are an advantageous approach to control the machine’s feed rate. Otherwise inaccuracies can occur leading to poor quality. In this paper Model Predictive Control (MPC) is used to control the feed rate of the machine. By applying a parameterized model to the current state of the milling process its future trend can be predicted. An n -th order system with delay time is used to model the behavior of the milling machine’s velocity control loop. By continually re-estimating the model parameters at run time, the controller adapts to the current machine behavior. We developed solutions of the maximum likelihood estimators of the model’s independent parameters for a state-space representation that can be numerically efficiently calculated. These estimators can be used to estimate the parameters of the n -th order system and the delay time iteratively based on the collected process data. Two validation studies, one emulating an online estimation based on real data from the five-axis machining center Mazak Variaxis 630II-T, and one simulation study with a simulated machine, were carried out to validate the iterative parameter estimation algorithm. In these studies, the time variant model, whose parameters are re-estimated on the basis of the developed estimators, is compared to a time invariant model, for which the controller does not adapt to changing machine behavior.