Predicting the remaining useful life (RUL) remains an important part in prognostics and health management (PHM) discipline. But the complexity of machine system and the noise in data make the prediction full of uncertainty and diversity. On the other hand, the traditional statistic methods are not capable for this kind of problems because of the limited number of data and previous knowledge. In this case, the machine learning methods are widely used in PHM field, since they are data-driven approaches which enable the model to determine the states without considering a homogeneous pattern. In this paper, the kernel principal component analysis and sliding window method are introduced for extracting the main features of the raw datasets. Then, support vector machine is used to predict the RUL of aircraft engines using data from C-MAPSS. The real RUL is uncertain, based on the value of features, so the fuzzy theory is introduced to the model to get the prediction interval of RUL. This paper also provides the performance running on dataset FDOOI to compare with prior approaches.
展开▼