Research On Fault Diagnosis Of Driving Motor Based On Machine Learning

As an important tool for urban environmental protection,road sweepers have developed rapidly in recent years.As an important power source in the road sweeper,motor operation directly affects the safety of personnel and equipment in industrial settings.Because the motor of the road sweeper needs to run for a long time and has a heavy burden,and the particularity of the working environment of the road sweeper,the probability of motor failure is very high.Therefore,it is of great significance to use the sensor technology and artificial intelligence methods to detect and diagnose the motor running state,so as to realize the early detection of motor failure,reduce the maintenance cost,improve social and economic benefits,and reduce the incidence of personnel accidents.Using the DC motor for road sweeping vehicle as the research object,this thesis focuses on the application of machine learning algorithms in motor fault diagnosis.Following are the main research contents:Firstly,this thesis analyzes the common drive motors,a research object is determined based on the structure and operation characteristics of different motors,and the types of motor faults and their causes are summarized.thus leads to the methods and principles of motor fault diagnosis,and then,according to the working characteristics of the drive motor of the road sweeper,summarizes three kinds of faults that are easy to appear in the drive motor of the road sweeper as characteristic quantities.Secondly,the working environment and hardware requirements of the road sweeper are analyzed.Based on this,an experimental platform for fault diagnosis of the road sweeper is built.To complete the hardware construction of the experimental platform,select appropriate hardware based on the working environment that the road sweeper may encounter.The experimental platform’s software system is built to ensure its normal operation at the local end.Remote monitoring of operations and cloud data storage can be achieved by connecting the local platform with the cloud platform.Then,a neural network model is studied in machine learning,it is proposed to use Principal Component Analysis(PCA)and Convolutional Neural Networks(CNN)to diagnose the fault of the driving motor of a road sweeper.In this method,to reduce the dimension of data,principal component analysis is used,inputs for training the convolutional neural network are selected based on their ability to represent fault characteristics.Experiments verify the rationality of applying principal component analysis and convolutional neural network to the field of fault diagnosis.Finally,in order to address the problems of long training times and low diagnostic efficiency in the above-mentioned methods,we propose a new method for motor fault diagnosis based on kernel principal component analysis(KPCA)and particle swarm optimization support vector machines.This method constructs a mixed feature set based on the time-domain and frequency-domain indexes of vibration signals,including 30 indexes.Using kernel principal component analysis instead of principal component analysis can reduce the dimension of nonlinear data more accurately.Taking the cumulative contribution rate of 90% as the threshold,select the index that can best represent the fault characteristics to participate in the model training.The improved particle swarm optimization algorithm is used to restrain it and prevent it from premature convergence.Particle swarm optimization is used to optimize the main parameters of support vector machines and diagnose motor faults.Through experimental verification,the fault diagnosis accuracy rate of this method reaches 96.5%.The accuracy rate of this method is slightly higher than the other methods listed above,but the diagnostic efficiency is much higher.It is expected to be applied in the fault diagnosis field.