关键词: 多退化变量, 实时健康度, 相似性寿命预测

Abstract: Since the research on multi-index modeling for the similarity-based residual life estimation were rare and modeling methods were limited to linear regression and thus relatively single,the multi-index modeling method for similarity-based residual life estimation based on real-time health degree was proposed.The principal component analysis,support vector data description,Markov distance and negative conversion function were all hired to fuse multiple degradation variables into a quantitative index namely real-time health degree which could reflect the system degradation state;based on the equipment real-time health degree,the equipment residual life could be predicted by using the similarity-based residual life prediction method with single degradation variable;the recommended method was validated through the case analysis of gyroscope residual life prediction.Results showed that the suggested method was feasible and had certain superiority in terms of statistically more accurate prediction (i.e.smaller prediction error).

Key words: multiple degradation variables, real-time health degree, similarity-based residual life prediction