Machine Tool Prognosis for Precision Manufacturing

Increasing demand for precision-manufactured parts for high-tech applications in aerospace, nuclear power, transportation, etc. continually drives the advancement of precision manufacturing technologies. As the precision and quality of manufactured parts are significantly affected by the performance of the machine tools, accurate and reliable condition monitoring, performance prediction, and maintenance of machine tools become one important part of precision manufacturing. In this chapter, a stochastic modeling technique is presented for prediction of machine tool performance degradation based on sensing data from the tool wear. Specifically, to account for the nonlinear and non-Gaussian characteristic of operating and environmental conditions on the wear propagation and machine performance degradation, particle filter (PF) that approximates probability distributions through a set of weighted particles is investigated. To improve the reliability of time-varying degradation tracking and prediction, an adaptive resampling particle filter method is developed. Specifically, particles are recursively updated and resampled from the neighborhoods that are determined by particles’ estimation performance from the last iteration, to characterize the temporal variation in the tool degradation rates. This leads to improved tracking and prediction accuracy with progressively narrowed confidence interval. The developed method has been experimentally evaluated using a set of benchmark data that were measured on a high-speed CNC machine.