Adaptive Process Estimation for a Grinding System

Precision components require ever increasing quality as well as a reduction in processing time. To achieve high quality precision parts, it is important to control the force, power, position, and the material removal rate (MRR) during grinding. Existing force control systems (such as a PID type controller) have been found to be limited because of process variation, since a fixed-parameter model may possibly lead to instability or part damage. A time varying, parametric model relating the MRR to grinding forces is needed to describe the process. Using this model as a basis, estimation algorithms may be developed for real-time modeling. The real-time model can then be implemented in an adaptive control to maintain a maximum MRR and limit force with changing process parameters, achieving a desired surface finish. Previous research indicates that several measurable sources of data can be correlated to grinding model parameters. Multiple sensor data using Kalman filtering, multi-variable recursion, learning algorithms, or other statistical techniques should achieve a superior grinding model parameter estimation, suitable for adaptive control of force and displacement trajectories. This paper develops multi-sensor input and windowing techniques for estimation of the process model parameters. Simulations, using controlled grinding force experimental data, compare the real-time parameter estimation model results to actual data.