Intelligent Diagnosis Of Complex Mechanical And Electrical Equipments Based On Kpcaand Rprop

With the continuous development of science and technology, automotive coating line’s production technology and equipments are constantly renewal and change. The whole production line may be shutdown because of a workstation was in error. Most traditional fault diagnosis performs analysis, investigation after the question happen, which has been unable to meet the growing trend of intelligent diagnosis. This paper studied the KPCA and RPROP in coating line equipment application, and conducted a coating line equipment intelligent early warning research. Thesis’s contents are mainly as follows:1. Further study of the automotive coating line production process and the various subsystems of major equipment failures were analyzed, analysis the failure signs before it happen, pointed out the cause of failure and classified it. Established a data acquisition system in each subsystems, provides the basis for intelligent warning.2. Established the coating line KPCA feature extraction model, analysis the impact of different types of kernel function and parameters on feature extraction. In order to eliminate noise, reduce information dimensions and the correlation, established the coating line KPCA feature extraction model. Studied the polynomial kernel function, Gaussian kernel function and radial Multilayer Perception kernel function, analyses of the impact of characteristic dimension and nuclear parameters in KPCA feature extraction accuracy. 3. Put forward an electromechanical system intelligent early warning method based on KPCA and RPROP neural network. Combined with the advantages of feature extraction in KPCA and RPROP neural network’s strong adaptive capacity, computing power, superior robustness, established a coating line electromechanical system intelligent early warning model. Test model by using numerical simulation method and the result shows the model has a high accuracy.4. Combined with extensive correlation function, improved early warning method of electromechanical systems. The traditional warning method is "measure", for its shortcoming, proposed a new method—Extension intelligence warning. Combined it with RPROP neural network and establish the coating line intelligent early warning model. Through simulation experiments show the model has a high accuracy rate.5. Finally, a summary of the paper was given and some prospects for further research was provided...