Research On Compressor Remote Diagnosis And Fault Prediction

Compressors have always been one of the most important equipment in the well-known home appliance industry and in other industries such as petroleum and chemical industries.Because the reciprocating compressor has a wide distribution and high manufacturing cost,if the compressor manufacturer can monitor the compressor operation status in real time through the Internet,and make fault prediction,it is of great significance for the later maintenance of the compressor.In order to realize remote diagnosis and fault prediction of a compressor,it is necessary to solve the key problems of remote data monitoring,diagnosis and prediction.In this paper,in the process of remote data monitoring,we adopt the method of breakpoint replenishment and table-by-table and database-by-database method to solve the problem of data loss and data storage.We find the diagnostic support vector machine(SVM)suitable for small sample fault data by consulting data and experimental simulation.By studying the SVM theory and influencing factors,we obtain the hybrid kernel least squares support vector machine(LSSVM).According to the existing limitations of hybrid kernel LSSVM,a hybrid kernel LSSVM compressor fault diagnosis model based on Self-adaption Differential Evolution(SADE)algorithm is proposed.The essence of SADE algorithm is adaptive processing of differential strategy selection,scaling factor and crossover probability based on DE algorithm.The SADE algorithm is used to find the optimal parameter combination of hybrid kernel LSSVM,and the final diagnosis model of hybrid kernel LSSVM is obtained.Experimental results show that SADE optimized hybrid kernel LSSVM diagnostic model is better than other optimizations in both accuracy and effectiveness.SVM not only deal with classification problems but also for regression problems.Therefore,the commonly used gray prediction model and the Support Vector Regression(SVR)prediction model are selected in the fault prediction model.Through theoretical research and experimental analysis,it is found that the grey prediction model has better data tracking ability while the single point prediction accuracy is low and the SVR prediction model has better single point prediction accuracy and poor data tracking ability.This paper combines the advantages of the two models to obtain the gray SVR prediction model.Finally,the experimental simulation results also show that the gray SVR prediction model is better than the single prediction model.After the compressor fault diagnosis and prediction model algorithm research,the Browser/Server model was used to construct the remote diagnosis and prediction system of the compressor,which formed a data transmission,data storage,data analysis,fault diagnosis and prediction processing system.