Rolling Bearing Fault Diagnostics Using SSA-ELM and TVF-EMD Multi-Scale Entropy

In order to address the issues with the traditional intelligent algorithm for optimizing ELM parameters-such as slow optimization speed, many adjustment parameters, and easy fall into the local optimal value-and its potential use in bearing fault classification using extreme learning machines (ELM), we propose a fault diagnosis method that combines time-varying filtering empirical modal decomposition (TVF-EMD) multiscale entropy with support vector machine (SSA-ELM). This research takes a look at rolling bearing defect feature extraction and ELM parameter optimization, and then introduces a novel swarm intelligence optimization approach. To improve the optimization speed and accuracy of bearing fault classification, the Sparrow Search algorithm (SSA) is used to optimize ELM parameters. A number of intrinsic mode functions (IMF) are derived from the signal after its decomposition using the TVF-EMD method. The next step is to recombine the beneficial IMF components by using the appropriate coefficient approach to identify them. Finally, the reconstructed signal's multi-scale entropy is determined using the feature vector, and the fault classification is carried out with the input of SSA-optimized ELM. The results show that the fault information can be obtained accurately and the identification accuracy is high. The feature vector is used to calculate the multi-scale entropy of the reconstructed signal, this method has better fault diagnosis classification performance.