Research On Fault Diagnosis Of Variable Air Volume Air Handling Unit Based On Data Drive

The internal failures of variable air volume air conditioning systems not only bring about a decrease in the service life of the equipment,but also cause a decrease in indoor air quality.Therefore,it is necessary to conduct fault detection and diagnosis research on variable air volume air conditioning systems.Quick and effective diagnosis of the source of the fault can ensure stable system operation,increase system safety,and reduce energy consumption and operating costs.The use of 5G and Io T technology enables online monitoring of sensor data,which lays the foundation for fault detection and diagnosis research using a data-driven approach.In recent years,data mining and deep learning methods have been rapidly developed in the field of artificial intelligence,so the data-driven analysis method has become an effective way to solve the fault diagnosis problem of variable air volume air conditioning system.In this paper,we firstly build an experimental platform for AHU of variable air volume air conditioning system,connect various types of sensors with EMCP IOT platform through PLC and wireless gateway,and realize the system operation data collection through mobile terminal and PC terminal.The experimental study was also carried out to obtain first-hand experimental data for fault detection and diagnosis,while the mathematical model of sensor faults in AHU was clarified.To address the problem of correlation between AHU sensor variables,principal component analysis is used to detect sensor faults in AHU units,and a principal component analysis model is established using the collected historical normal operation data,through which sensor deviation faults and drift faults are detected separately,and the results show that the fault detection based on principal component analysis is effective,but for sensor small-amplitude deviation faults,the fault detection The rate is low.Considering the problem that the correct rate of the principal component analysis method is low in detecting the small-amplitude deviation fault of the sensor,an improved principal component analysis method,the kernel principal component analysis method,is proposed.The kernel principal component analysis method maps the sample data to a high-dimensional space through the kernel function and then performs linear decomposition,which can extract features while preserving the nonlinear relationship of the original data.The kernel principal component analysis method is used to detect sensor deviation faults and drift faults.The results show that the improved method can effectively improve the detection rate of overall faults and small magnitude faults.The principal component analysis and kernel principal component analysis can only detect the occurrence of system faults but cannot identify the location of faults.To further diagnose the source of faults,the contribution rate method,BP neural network and PCA-BP neural network method are used for fault diagnosis of the system,and the results show that the sensor drift fault can be successfully identified using the contribution rate method but not the source of sensor deviation fault.Therefore,using BP neural network method and PCA-BP neural network method for sensor deviation fault diagnosis,the results show that compared with the traditional BP neural network,the fault diagnosis rate of BP neural network using PCA after dimensionality reduction is 95.39%,which is 4.09%higher than the 91.3% fault diagnosis rate of BP neural network,and can identify all types of sensor fault sources.The research in this paper can provide a corresponding basis and reference for fault detection and diagnosis of equipment in variable air volume air conditioning systems and similar fields.