Intelligent Diagnosis Of Mechanical And Electrical Equipments In Automobile Coating Line Based On PCA And SVM

Due to the rapid development of automobile coating line both in newly painting technology and production processes, even in greater and more complicated equipments, it's given a higher requirement to make an exact diagnosis for electrical and mechanical equipments on condition that the process in production and the efficient and reliable equipments operation in line is a sure thing. As the intelligent theory brought into and developed, the equipment status recognition will be getting into a stage of intelligence. This paper gave a detailed description about the application in equipment diagnosis of coating line with principal component analysis and support vector machine. In addition, the monitoring and intelligent diagnostic system in automobile coating line was designed under the circumstance of virtual instrument. Thesis's contents are mainly as follows:1. According to the production process in an automobile coating line, the failure about formation and occurrence of major equipments from various subsystems was analyzed and studied, pointing out the frequent failure types and signs of equipment within each subsystem. The data acquisition system was established based on that.2. The methods of fault symptoms extraction were studied. Under the real system and working environment, various signals picked up by sensors, on the one hand, were not entirely reflect the significant information on device states, some variables can interfere with diagnosis. On the other hand, the outputs of equipment signals feature had some certain relevance. Therefore, the application of both principal component analysis and principal component analysis improvement model is mainly discussed in paper. Meanwhile, the advantages of this method were verified through analysis to equipments in oven heating system of coating line.3. The influence the kernel function types and kernel parameters brought on classification accuracy was deeply studied. According to the simulation samples data from double helix, the effect Gaussian and polynomial kernel functions imposed on was analyzed respectively. So did the Gaussian kernel width and punishment parameters. The results showed that the best classification accuracy can be found when the Gaussian kernel function parameter and penalty parameter was at a range.4. A method of equipment status recognition was proposed based on principal component analysis and support vector machine. Combined with feature extraction by principal component analysis and better recognition function by vector machine, the vector machine training model was created with the best kernel parameters optimizing by using a grid search and cross-validation method. That was verified through the 12 different states from four kinds of equipments in heating system. The classification accuracy was analyzed on basis of before and after the principal component analysis improvement, and the rate up to 85% basically or more.5. The monitoring and intelligently diagnostic system in equipments of automobile coating line was designed accompanied with virtual instruments. At the same time, the modules were introduced including data acquisition, time and frequency domain analysis, feature extraction and intelligent diagnosis.6. Finally, a summary of the paper was given and some prospects for further research was provided.