A method for predicting aviation equipment failures based on degradation-track similarity

Abstract

The framework of similarity-based prognostics was presented, which takes advantage of system 's training instances' degradation trajectory and run-to-failure time to predict the remaining useful life(RUL) of test instances. Degradation models are extracted from time series data of training instances. Similarity between time series data of test instance and degradation model is calculated by likelihood function. RUL value according to the degradation model is then estimated at the best matched time stamp. RUL values weighted by similarities are fused by kernel density estimation to form the final probability density of the RUL of test instance. Results of aviation equipment simulation experiments show that the similarity-based RUL prediction performs better in accuracy and convergency.