Jixie qiangdu (Jan 2020)
FAULT DIAGNOSIS OF PLANETARY GEAR BOX BASED ON SENSITIVE TRANSIERT IMPACT FEATURE EXTRACTION AND HIERARCHICAL EXTREME LEARNING MACHINE
Abstract
In the intelligent diagnosis of planetary gearbox faults,the issue of "difficult extraction"of the vibration signal characteristics,the"quality difference"of the constructed eigenvector set and the"low precision"of the fault classification model based on the extreme learning machine. Put forward a state identification method for a planetary gearbox solar wheel for how to capture sensitive transient impact feature in the vibration signal and construct a high-dimensional eigenvector set and improve the fault classification accuracy of the extreme learning machine. Firstly,the vibration signals are respectively solved by fast kurtosis diagram and variational mode decomposition and several intrinsic mode function matching the center frequency fωcorresponding to the maximum kurtosis value are selected,find the value of improve multi-scale permutation entropy to construct the highdimensional eigenvector set. Secondly,the de-noising automatic encoder is used to make the input weight and threshold of the implicit learning node of the extreme learning machine satisfy the orthogonal condition to realize the layering of its hidden layers.Finally,the above eigenvector set is used as the input of the hierarchical extreme learning machine,and the fault classification model of the planetary gearbox solar wheel is established through training. The results show that the proposed method achieves the effective extraction of sensitive transient impact feature in the vibration signal of the solar wheel and the high quality construction of the eigenvector set,and also improves the classification accuracy of the intelligent diagnosis model.
