Shapelets-Based Research on Fault Classification of Rotating Machinery

In order to improve the accuracy and operational efficiency of the classification of rotating machinery faults, this paper proposes a multi-time sequence classification method based on FDTSC (F-MTSC). The algorithm first captures the sequence features of the time series on a one-dimensional time series, obtains candidate shapelets, obtains the most representative k shapelets after diversified top-k query, and then applies the traditional classification method to classify the shapelets converted data set to produce the result. In the multivariate time series, integrated learning is introduced. In each dimension, the classification results of different traditional classifiers are used to obtain the classification result of the current dimension, and then the relative majority voting method is used to classify the on-site rotating machinery fault data. Experimental results show that the comparison among using traditional classification methods alone, traditional classification methods combined with F-MTSC algorithm and neural network model turns out that Traditional classification methods combined with F-MTSC algorithm have the best effect. Applying this algorithm to the classification of rotating machinery faults is of great significance for better solving industrial practical problems.