Design And Implementation Of Vehicle Drive Fault Diagnosis System Based On Unsupervised Learning

Automotive drive fault diagnosis system keeps the stability and safety of vehicle,and it is the most significant component of a vehicle.For more intelligent and more convenient automotive drive fault diagnosis system,the demand is strong,the economy market is large,the policy orientation is good.At present,there are many problems with the On-Board Diagnosis(OBD)system:need a lot of expert knowledge;the ability to diagnose unknown faults is limited;it is inefficient to view fault information with an external OBD diagnostic instrument;the data record of the vehicle drive system is inadequate,which can not reflect the state of the vehicle drive system.The application of remote vehicle drive fault diagnosis system has not bean applied,and there are shortcomings such as high dependence on the OBD system of the vehicle and the problem of limited applicable models.In order to solve the above problems,this paper designed and implemented an vehicle drive diagnosis system based on unsupervised learning.The main contributions of the paper are as follows:1.The architecture of automotive drive fault diagnosis system is composed of distributed data acquisition cluster and real-time fault diagnosis cloud platform.The distributed data acquisition cluster consists of several equipments and a Wi-Fi access point,which can acquire data and transmit data to the fault diagnosis cloud platform.The cloud platform runs the automotive drive fault diagnosis algorithm based on unsupervised learning to achieve real-time fault detection and location.2.Distributed data acquisition equipment:it can form a distributed data acquisition cluster with high-precision synchronized clocks,based on the Wi-Fi LAN clock synchronization algorithm;hardware performance optimization includes preventing power surge,reverse power connection,and power short circuit,combining switching regulated power supplies and linear regulated power supplies to improve energy conversion efficiency,and separating digital circuits,analog circuits,and power circuit areas to improve ADC accuracy.3.Automotive drive system fault diagnosis algorithm based on unsupervised learning:periodic sampler deals with missing data and timestamp misalignment;combined with convolutional neural network,convolutional long and short-term memory and auto-encoder,it can extract the multi-dimensional correlation features and time series features of vehicle drive system;unsupervised learning method to realize vehicle drive system fault diagnosis,without any fault labels and expert knowledge.4.Real-time fault diagnosis cloud platform:JT/T 808 equipment access nodes can be expanded into a super-large-scale cluster,and finally provide large-scale equipment connection capability;Kafka streaming platform and multiple fault diagnosis servers work together to achieve data load balancing and parallel computing,and run the fault diagnosis algorithm in stages,finally provide real-time fault diagnosis capability;the visual interface provides visualization services,such as fault location query and vehicle state restoration.