Multidimensional Performance Degradation Data Fusion and Residual Life Prediction for Machine Tools

摘要

Manufacturing quality related performance of a machine tool deteriorates with time and usage. The degradation is usually characterized by several indicators, which can be periodically monitored. A challenging issue is how to fuse such multidimensional degradation observations into a composite degradation index (CDI) so as to optimize the preventive maintenance decision of the monitored machine tool. This paper addresses this issue through proposing a data fusion method. The proposed method is a multi-criteria performance evaluation method. It derives the weights of the indicators based on the distribution of time to the degradation limit of each indicator, aggregates redundant [complementary] indicators using a [generalized] weighted mean model, and uses a random failure threshold for the CDI. The parameter estimation methods are developed to estimate the parameters of the generalized weighted mean model and the random failure threshold. The proposed method is illustrated through analyzing a real-world dataset that deals with the degradation of position and orientation accuracy of CNC milling machines. The results show that the CDI is a robust measure to represent the quality performance degradation and hence can lead to a robust remaining life prediction and better maintenance decision.