Gear Fault Diagnosis Model Combined with MED-LMD-Hypersphere Multiclass SVM

In view of the difficulty of features' effective extraction and accurate clustering for different fault gears, a gear fault diagnosis model combined with minimum entropy deconvolution (MED), local mean decomposition (LMD) and hypersphere multiclass support vector machine was proposed. Firstly we processed the collected vibration signal with minimum entropy deconvolution, separating high-low frequency stages and excluding uncertain signal; Secondly the LMD algorithm was applied in the processed signal, the PF components sensitive to the fault feature were selected to reconstruct the signal, and the information entropy was extracted from this signal as fault features; Finally the construction of hypersphere and classification were done by inputting these features to hypersphere multiclass SVM. The validity of the model was verified by analyzing the experimental data.