A Fault Degree Classification Method for AE Signal Based on VAR-DBN

YONG ZHOU, TIAN-SHU LI, LI LIN

Abstract


It is difficult to find the train axle fault earlier by traditional calculation method. In this paper, a time series analysis and pattern recognition analysis based model is presented to improve the performance on diagnosing fault of AE signal. The suggested methodology mainly involves three parts. First, empirical mode decomposition (EMD) is used to transform the AE signals into stationary signals by decomposing the original signal into several intrinsic mode functions (IMFs) and one residue. Then, vector autoregressive (VAR) model is selected to reflect the characteristic included in IMFs by establishing feature vector comprised by the coefficients of the VAR model. In the end, a Deep Belief Network(DBN) is introduced to classify the AE signals and detect the fault signals caused by faults. The result of the experiment shows that compared with other time series parameter analysis based classification study, the proposed method has better behavior in accuracy when identifies different conditions of railway vehicle axle on AE signal.

Keywords


Empirical mode decomposition, Vector autoregression, Deep belief network, Feature extraction, Fault diagnosis.Text


DOI
10.12783/dtcse/ceic2018/24536

Full Text:

PDF

Refbacks

  • There are currently no refbacks.