Research On Composite Fault Diagnosis Of Complex Planetary Gear Train Based On Sparse Decomposition

In recent years,the fault detection and diagnosis of gear transmission systems have attracted great attention in order to reduce the downtime and secondary damage caused by mechanical failures.Compared with simple planetary gear systems,the structure of compound planetary gear systems is more compact and the bigger range of transmission ratios can achieved.It is widely used in aviation,aerospace,wind power and other machinery.Therefore,it is necessary to diagnose faults of compound planetary gear systems.Currently,vibration signal-based fault diagnosis technology is the most widely used technique for detecting and diagnosing rotating machinery faults.In practical applications,however,faulty signals are often obscured by healthy component vibration signals and background noise.Therefore,it is necessary to research and develop new signal processing methods based on vibration signal fault diagnosis.In this paper,the research is carried out on the fault diagnosis of vibration signal of compound planetary gear systems,and the content of this work is listed as follows:(1)The common faults of gears in compound planetary gear systems and their causes were analyzed.The vibration signal model of single and compound faults of compound planetary gear systems is established and simulated.The characteristics of local faults and compound fault models in sun gears and external planet gears were analyzed,and the characteristic frequencies of major components that experience local faults in compound planetary gear systems were calculated.(2)The sparse decomposition algorithm was studied.A cyclic dictionary suitable for composite faults using the concept of cyclic matrices was constructed,and the disadvantage of low computational efficiency in the orthogonal matching pursuit algorithm was improved.A fault feature extraction method called Cyclic Sparse Decomposition(CSD)was proposed to enhance the amplitude of reconstructed signals.After combining spectrum analysis and envelope analysis with fault characteristic frequency,fault localization was achieved.The effectiveness of this method was verified through comparative analysis using simulation.(3)An improved random forest algorithm based on cyclic sparsity and slime mold algorithm for fault diagnosis of gears was proposed.Firstly,a cyclic sparse decomposition model was established for envelope analysis,and the envelope harmonic-to-noise ratio was calculated.Then,the feature vector was constructed by combining the envelope harmonic-to-noise ratio with timedomain,frequency-domain and other indicators.The Slime Mould Algorithm(SMA)was introduced to optimize the two parameters of Random Forests(RF)for gear fault identification.At the same time,it was compared with three classification methods: RF,Long Short-Term Memory(LSTM)and Convolutional neural network(CNN).The data results show that the improved RF can improve the accuracy of classification and recognition.The method proposed in this paper can effectively extract the composite fault features of the sun gear and external planet gear from vibration signals with strong background noise and realize fault diagnosis,which can provide theoretical support for the composite fault prediction of the compound planetary gear r systems.