Approach Research of Sensor Fault Detection and Diagnosis in HVAC System |
|
Author | ZhangXiaZuo |
Tutor | WangXiaoZhe |
School | Northeastern University |
Course | Pattern Recognition and Intelligent Systems |
Keywords | Air-conditioning system Fault Diagnosis Sensor Principal Component Analysis The nuclear principal component analysis Neural Networks Genetic Algorithms |
CLC | TP212 |
Type | Master's thesis |
Year | 2009 |
Downloads | 92 |
Quotes | 1 |
With HVAC (HVAC) system has been great development in various occasions more widely used , the HVAC system , the control system of the air-conditioning system has become increasingly complex . Whether it is a commercial office building or industrial processing sites, and even civilian residential air conditioning system put forward higher requirements , requires the stable operation of the air conditioning system , comfort and energy efficiency . Increasingly complex and large HVAC systems tend to generate a variety of fault , to the quick and timely detection , identify system faults , has been far from the operator whatever , which makes fault detection diagnosis (Fault Detection and Diagnosis, FDD) system has increasingly become necessary . The fault diagnosis method based on multivariate statistical process control is an important branch of research in the field of fault diagnosis . In recent years, principal component analysis of the different forms of fault diagnosis in various fields in the wide range of applications , and achieved good results . Firstly, principal component analysis (PCA) of the HVAC system sensor fault detection and diagnosis . The principal component analysis method using measurement data to create a model of the system under normal operating conditions , in order to avoid direct the establishment of a system of analytical model . By using this method to create a model of the system , the selection of the main metadata is a critical issue . This selection of the principal component contribution rate method to select the optimal principal component scores . Traditional principal component analysis method can not be used to analyze the nonlinear system , this paper presents a comprehensive utilization of nuclear function principal component analysis and neural network predictor nonlinear systems fault diagnosis method ; same time , to improve the performance of the neural network predictor , in this paper, a genetic algorithm applied to the neural network weights selected , and achieved good results . MATLAB simulation tool to the the a building simulation air conditioning system as the research object , the application of the above method to design a sensor fault diagnosis system , proved by a large number of simulation tests , the above method has satisfactory fault diagnostic capabilities for perfect air conditioning system sensor fault diagnosis method has a specific meaning . Finally, the thesis is summarized and put forward recommendations for future air conditioning system fault diagnosis .