Research On Equipment Fault Diagnosis Algorithm Based On Support Vector Machine

With the rapid development of industrial intelligence,the operation and maintenance of various equipment is becoming more and more important.In order to reduce the impact of equipment faults on production,timely diagnosis of equipment faults has become an important link of industrial intelligent production.Fault diagnosis algorithm uses machine learning to carry out real-time diagnosis of the running state of equipment,ensuring the normal operation of equipment.In this paper,the main algorithms of fault diagnosis are studied,especially the fault diagnosis algorithm based on support vector machine.In view of the characteristics of fault data sets,isolated kernel support vector machine is used as the basic algorithm.Isolated kernel is a kind of data dependent kernel,which has better adaptability to complex data sets.In this paper,kernel function and algorithm are improved to improve the performance of fault diagnosis algorithm.The main work includes:(1)The existing partition method of isolated kernel is only sensitive to global anomalies and not good at dealing with local anomalies.To improve the partition method of isolated kernel,this paper proposes the PNS-NN(Partial Normal Sample-Nearest Neighbor)algorithm.The Euclidean distance is used to screen the normal data around the abnormal data and reduce the coincidence density of positive and negative samples.The simulation shows that the algorithm can increase the isolation area of abnormal data in the partition,and improve the detection performance of the algorithm model.(2)Aiming at the problem of high proportion of abnormal data in misclassified samples of SVM,the ISO-SVM-AdaBoost(Isolation KernelSupport Vector Machine-AdaBoost)algorithm is proposed.The weight parameter of misclassified samples is introduced to increase the attention of the algorithm to the wrong samples,improve the classification accuracy of this part of data,and effectively reduce the proportion of abnormal data.The data of photovoltaic power generation equipment are selected for simulation experiments,and it is verified that the algorithm proposed in this paper has a good improvement in F1 value and AUC value two evaluation indexes.(3)Based on the algorithm proposed in this paper,the overall architecture of the fault diagnosis system is designed,the required functional modules are realized,the fault diagnosis system is built,and the feasibility of the algorithm is verified.