Diagnosing planetary gear faults using the fuzzy entropy of LMD and ANFIS

The small size, low weight, and large transmission ratio of planetary gear have resulted in large-scale use, low speed, and heavy-duty mechanical systems. Poor working conditions of planetary gear lead to frequent occurrence of faults. A method is proposed for diagnosing faults in planetary gear based on fuzzy entropy of Local mean decomposition (LMD) and Adaptive neuro-fuzzy inference system (ANFIS). The original vibration signal is decomposed into six Product function (PF) components and a residual using LMD. Given that decomposed PF components contain the main fault feature information, fuzzy entropy is used to reflect the complexity and irregularity of each PF component. The fuzzy entropies of each PF component are defined as the input of the ANFIS model, and its parameters and membership functions are adaptively adjusted based on training samples. Finally, fuzzy inference rules are determined, and the optimal ANFIS model is obtained. Testing samples are used to verity the trained ANFIS model. The overall fault recognition rate reaches 88.8%, and the fault recognition rate for gear with wear reaches 96%. Therefore, the proposed method is effective at diagnosing planetary gear faults.