Research And Application Of Fault Detection Based On PCA-SVDD

With the rapid development of my country's industry,industrial processes are developing in the direction of intelligence,integration and systematization.Efficient fault diagnosis and prevention in the production process can further ensure production safety.Therefore,fault detection has become a very important step in process automation and a part of core technology,which has considerable practical significance and value.In addition,with the continuous advancement of computer distributed control system technology,large-scale industrial process information is stored in the system.Digging in these massive amounts of information and sorting out valuable information,so as to improve the level of fault diagnosis,so that relying on the data information database to improve the level of fault detection becomes a reality.The detection algorithm currently being applied generally standardizes the original data,and then reduces the dimensionality of the standard data through PCA(Principal Components Analysis,PCA).However,the traditional PCA dimensionality reduction method usually ignores the correlation between sampling batches,which will lead to the elimination of some valid information.In addition,in traditional fault detection methods,the SVDD(Support Vector Data Description,SVDD)algorithm is usually unable to meet The requirements of industrial precision data on detection accuracy.Therefore,in view of the above problems,this paper effectively improves the performance through improved algorithms.The specific work of this article is as follows:(1)Because the traditional PCA algorithm often ignores the correlation between data when processing multiple batches of dimensionality reduction data,this paper based on the autoregressive moving average model can extract the characteristics of time-based independent relationships from the data,and autoregressive The moving average model is combined with the traditional PCA algorithm to improve PCA performance.Experimented with matlab software,selected 50 sets of samples from the automotive crankshaft data set,and used the improved PCA algorithm for feature dimensionality reduction.By comparing with the traditional PCA algorithm,it was found that the improved PCA algorithm can better retain the effective information after dimensionality reduction.(2)Since traditional SVDD cannot regulate negative samples,the detection efficiency is not high enough.This paper proposes a support vector data description algorithm based on an improved support vector description algorithm with negative samples.By using the traditional SVDD algorithm and the improved SVDD algorithm in matlab to learn the original data of 500 car engine idling speeds,and then detecting the fault data,it is found that the improved SVDD algorithm has improved accuracy compared to the traditional SVDD algorithm.(3)This paper proposes a method based on improved PCA-SVDD,combining improved PCA and improved SVDD algorithm to establish fault detection rules and fault diagnosis rules,and simulation experiments of automobile engine idling data prove the effectiveness of the algorithm.Based on the study of traditional fault detection algorithms,the paper improves the traditional detection algorithm detection.This research is helpful to reduce the occurrence of failures in industrial production,thereby improving production efficiency.