Research Of Fault Diagnosis Of Micro-vehicle Reducer Based On Deep Belief Network

Micro-vehicle reducer is the key equipment to ensure the power and safety of vihicle,and its operating status has a direct impact on the driving and control of microvihicle.At present,many experts and scholars at home and abroad have already started the study of the fault diagnosis of micro-vehicle reducer.Many algorithms such as BP neural network and support vector machine have been widely used in the pattern recognition of mechanical failure,and achieved good results.The fault diagnosis of the micro-vehicle reducer is virtually a process of recognizing certain signals of the machine by machine learning.In general,there are steps such as feature extraction of faulty signals,classification of faults and evaluation of diagnostic results.With the deepening of the study of deep learning algorithms,the classification algorithm based on deep learning is gradually applied to various practical applications,such as image processing,pattern recognition and data mining.Compared to the traditional classification algorithm,deep learning can classify data by digging some of the features or structures within the data.This thesis focuses on the algorithm and application of deep learning,and deeply discusses the Deep Belief Network as the core,and applies it to the fault diagnosis of micro-vehicle reducer.The main research contents include:1.Analyze the common fault types of the micro-vehicle reducer in the production workshop,collect the vibration signal as the basis of diagnosis,and use the kernel principal component analysis method to extract the feature in the vibration signal of the rotating machinery to reduce the dimension of the signal and extract its effective feature.2.Introduced the principle and steps of the Deep Belief Network.In view of the setting of some parameters in the Deep Belief Network,research the effects of the parameters on the performance of the model.The experimental method is carried out by the control variable method to find the optimal parameter setting of the Deep Belief Network.3.In order to solve the problem of crossover or overlap between different types of fault samples,the Deep Belief Network and KNN algorithm are combined to propose a KNN-DBN fault diagnosis model.The KNN algorithm is used in the test phase of the diagnose modal to improved the accuracy of the diagnosis.Discuss the influence of K value in KNN algorithm on the experimental results,then compare the results with those of DBN modal.