An Fault Diagnosis Method For Planetary Gear Based On Differential Evolution For Probabilistic Neural Network

Planetary gears play a crucial role in rotating machinery,because of its structure and transmission characteristics,planetary gears are w idely used in precision equipment and heavy machinery.When the gear failure will affect the health of the equipment,serious will cause huge economic losses and casualties.Therefore,it is of great significance to carry out the research on fault diagnosis of planetary gear,identify fault mode accurately and eliminate hidden trouble as early as possible to ensure the healthy operation of equipment.This paper takes planetary gears as the research object and focuses on their key problems in the field of intelligent diagnosis: the signal decomposition,characteristic dimension reduction and fault model training are studied deeply.The main research contents are as follows:Firstly,the vibration characteristics of planetary gear fault are analyzed.Analyze the influence of planetary gear failure on vibration signal and the transmission path of vibration signal in planetary gear box.The influence of gear failure on gear meshing impact is analyzed by dynamic simulation.According to the fault frequency characteristics of different positions of planetary gear to establish the vibration signal fault model of planetary gear,and,to provide simulation experimental data for subsequent experimental verification.Secondly,an empirical wavelet signal decomposition met hod based on binary k-means is proposed.There are some problems in traditional signal decomposition methods,such as window function fixation,mode aliasing and endpoint effect,The Fourier spectrum of the original signal is divided by binary k-means,and the empirical wavelet is used to decompose the signal to overcome the problem of artificial decision factors.The simulation signal is used to verify that the binary k-empirical wavelet transform signal decomposition method has good performance on fault characteristics.Thirdly,a probabilistic neural network(PNN)fault diagnosis model for planetary gear is proposed based on t-distribution random neighborhood embedding feature dimensionality reduction method.In view of the fact that there are different kinds of sample points in the traditional dimensionality reduction method which appear to be aliasing each other after dimensionality reduction,and difficulty in nonlinear mapping of high-dimensional data to low-dimensional data.In this paper,the random neighborhood embedding feature of t distribution is used to reduce the dimension of multi-dimensional feature.This method can keep the probability constant when mapping high-dimensional spatial data to low-dimensional spatial data,and verify the good clustering effect of t-distribution random neighborhood embedding method through simulation experiment data.Then aiming at the problem that the smooth factor in the traditional probabilistic neural network needs to be obtained by experimental adjustment and is easily interfered by human factors,the accuracy is taken as the evaluation index,and the smooth factor is optimized by differential evolutionary algorithm.Through the comparison of the simulation data,it is verified that the fault diagnosis model has high accuracy.Finally,set up planetary gear fault test bed,collect vibration signal,test verification.The normal operation of gears,pitting erosion of sun gears,broken teeth of sun gears,broken teeth of planetary gears and broken teeth of gear rings were taken as the experimental objects to collect vibration signals at different positions and of different kinds,and the effectiveness of the signal decomposition,characteristic dimension reduction and fault diagnosis method proposed in this paper was verified according to their vibration signals.