Fault Diagnosis Of Roiling Bearings Based On Dimensionality Of Gaussian And Mixed Kernel SVM Fusion Algorithm

Domestic and foreign research scholars have researched on the problem of mechanical fault diagnosis which utilized support vector machine and neural networks to solve problems of fault diagnosis and obtain some achievements.However,research on fault diagnosis of rolling bearings is few in number where improvements are needed,including: aiming at noise components that can not be avoided in the vibration signal,how to preserve feature of signal during denoising;aiming at conplex vibration signal,how to extract effective features which can reflect status of rolling breaings;aiming at a large number of samples collected of rolling bearings,how to construct valid learning model and achieve the purpose of marking fault types for them.The main idea of this research is: in order to improve diagnosis accuracy,reduce noise of vibration signal to preserve feature component,then reduce dimension and decrease information redundancy;utilize fault samples collected of rolling bearings and train diagnosis model to achieve supervised learning of category labels;construct fault intelligent diagnosis model based on support vector machine and apply to rolling bearings.The main contents include:(1)Aiming at the noise components in vibration signal of main reducer,an empirical mode decomposition and improved wavelet threshold function method is proposed.According to different time scales of signals,by using Empirical Mode Decomposition,a complex signal is decomposed into serveral Intrinsic Mode Functions in descending order of frequency.Using the improved wavelet threshold function to denoise the pseudo component of each frequency IMF to preserve feature of signal.The method effectively solve the problem that signal feature is oscillation over-smoothing during denoising.(2)Aiming at the multiple feature components in vibration signal of rolling breaings,a latent variable model based on local distance preservation is obtained.Firstly,with the help of the locality preserving projection,utilizing the objective function of the locality preserving projection,constraint priori of latent variables can be obtained.Then the posteriori probability of the latent variables can be got according to the Bayes theorem.Finally,the positions of the latent variables in the latent space are determined through maximum a posteriori algorithm.Experimental results and comparison with the traditional LVM prove that our proposed method performs well in maintaining the inherent manifold structure of data in high imensional space.(3)A fault diagnosis intelligent model of support vector machine based on mixture kernels is proposed to solve fault intelligent diagnosis of rolling bearings.Aiming at the selection of kernel function and the optimization of parameters in support vector machine,adapt a new kernel function and uses the genetic algorithm to optimize the parameters in support vector machine.Taking the extracted feature vectors as the input data set,the simulation experiment is carried out by the optimal SVM to achieve the effect of fault pattern recognition.(4)Simulating the operation process of pattern recognition system and the diagnosis process of the actual use of rolling bearing to establish the rolling bearing pattern recognition system.