A Fault Diagnosis Method for Rolling Bearings Based on Feature Enhancement and LSTM
Due to the complex and ever-changing working environment of rolling bearings,traditional signal processing techniques are difficult to detect weak early fault features under the interference of noise and other components,and traditional fault diagnosis meth-ods rely on manual feature extraction.To solve the above problems,an improved differential evolution particle swarm optimization multi-point optimal minimum entropy deconvolution adjusted(IDEPSO-MOMEDA)algorithm was proposed based on adaptive local iterative filter(ALIF)and improved differential evolution particle swarm optimization to enhance the fault impact component of rolling bearings.The signal was decomposed using ALIF and reconstructed according to the kurtosis correlation coefficient criterion;IDEPSO was used to optimize the parameters of MOMEDA and enhance the impact of the reconstructed signal;finally,the use of long short term memory(LSTM)networks were used for realizing end-to-end intelligent fault diagnosis of rolling bearings to solve the shortcomings of manual feature extraction.The effectiveness of this method was verified through experimental data of rolling bearings.Compared with LSTM,ALIFLSTM,ALIF-IDEPSO-MOMEDA-RNN,ALIF-IDEPSO-MOMEDA-DBN,the fault diagnosis accuracy of the proposed method ALIF-IDEPSO-MOMEDA-LSTM can reach 99.78%,further proving the superiority of this method.
rolling bearingadaptive local iterative filter(ALIF)multipoint optimal minimum entropy deconvolution adjustedlong short term memory(LSTM)fault diagnosis
NETL
NSTL
万方数据
