Fault Detection And Diagnosis Technology Based On SVM And Immune Algorithm

With the development of modern technology, industry scales has been growing and the manufacture infrastructure is more complex than before. The safety in the manufacture process is getting more attention from the people. This situation leads many scholars to research into the process monitoring and fault diagnosis. In recent years, a new machine learning algorithm, support vector machine (SVM), is widely applied in many fields and act a good performance in the case of small samples. Taking the industry process's properties into consideration fault diagnosis methods based on SVM and immune algorithms are deeply researched in the paper.Firstly, several kinds of fault detection and diagnosis methods are summarized and their performances are listed. The principle of statistical learning theory and support vector machine are introduced as well as the principle of SVM for classification. At the same time the impact of SVM core function's parameters for classification are studied in this paper. In order to improve the detection and diagnosis performance of SVM the combination of Immune Algorithms and SVM is suggested in the paper. This paper made some changes to the immune genetic algorithm and clone selective algorithm separately and based on the modified algorithm the MIGA-SVM and MCSA-SVM are brought forward. In order to verify the validity of the algorithms TE process is used as the platform of the simulation. The MIGA-SVM and MCSA-SVM are used to detect and diagnose the fault of TE process. The results shows the efficiency is highly improved using the modified algorithm.