Fault Diagnosis Of Air-conditioning Refrigeration System Based On SAE

The application of air conditioning and refrigeration systems has penetrated into all areas of society,the scale of the system and its automatic control system is increasing year by year,the types of the main components,auxiliary accessories and control components are various,the recycled substances are complexed,faults are inevitable during running the system,the output does not match the expected,lost the original design of the functions specified,which will affect the reliability and security of system operation,real time fault diagnosis can not only ensure the normal operation,but also can timely detected and maintain,which can avoid unnecessary losses with certain practical significance.To overcome the drawbacks of using supervised learning to extract fault features for classification,high degree of artificial dependence and low nonlinearity of the features in most of current fault diagnosis of air-conditioning refrigeration system.Sparse auto encoder(SAE)is presented to extract fault features.Briefly analyze the principle of Artificial Neural Network classification,Support vector machine,Principle component analysis(PCA)and SAE.Analyze air conditioning and refrigeration system and the common faults,introduce the principle of the typical fault occurrence.Briefly introduce ASHRAE experiments and use the data as historical data for the fault diagnosis model establishment of the air conditioning refrigeration system.The features extracted by the SAE and the PCA are used as the input to the classifier,to achieve fault diagnosis for air-conditioning refrigeration system.Results indicate that the indexes of the model combined with SAE,such as Accuracy,Precision and Recall,are all improved,especially for the faults with high complexity(fault with low fault deviation degree,System-wide fault,such as,Excessive or insufficient of refrigerant and lubricating oil and normal condition),the fault feature sensitivity is better extracted by the SAE than the PCA.The SAE structure is tuned by adjusting the number of hidden layers and nodes to build the optimal model.The SAE feature self-learning ability turns out to be good by adjusting the quantity of the training sample,SAE shows high generalization ability with small scale sample data and high efficiency with large scale data.Obviously,the use of SAE could effectively optimize the diagnosis performance of the classifier.Compare characteristics of the SAE,PCA,ANN and Artificial feature extraction.As for SAE feature extraction,no labeled data and manual experience are required,good for non-linear and complex data,automatically learn multi-dimension feature with multi-hidden layer,less training sample data requirements and higher feature fault sensitivity,it can decrease the data dimension to a certain extent with high feature extraction speed.