Research On Sensor Fault Diagnosis Based On Autoassociative Neural Networks

In modern control engineering,the function of sensor is very important. Specially,the operating status of sensors in the feedback control directly decides if the equipment can work normally and safely.The number of sensors applyed is more and more,and sensors are easily destroyed.for improving system reliability,So the research of Sensor Failure Detection, Identification and Accommodation(SFDIA)has its practical value.This work is aimed towards the development of an artificially intelligent search algorithm used in conjunction with an Auto Associative Neural Network(AANN) to help locate and reconstruct faulty sensor input in control systems based on domestic and foreign theory. A novel algorithm based Auto Associative Neural Networks is given for sensor failure diagnosis.The structure and algorithm of AANN are presented.The approach is proposed for searching two faulty sensors and estimate their real value. Auto Associative Neural Networks is used to diagnose sensor drift failure and recover their signals based on the inherent behavior of the AANN, SSE(Sum Squared Error) gives an estimate of the utility of a node.The algorithm uses a decremental step sizing procedure to ensure that the SSE falls to the lowest value.A cut off test is used during the search to discard those combinations under test that highly unlikely to be the correct faulty sensor pair based on their response and" SSE trend.a preliminary test is used to capture the SSE trend.Based on a model of a chiller ,Common sensor errors such as drift, shift errors, the algorithms response to them have been studied. The issue of noise has also been investigated.MATLAB is used to test the algorithm's performance under different types of condition. The method help locate the faulty sensors and reconstruct their actual values.The method is easy to realized and has simple structure.The simulation results show the method works well.The software using JAVA technology is developed to diagnose sensor faults such as drift error and shift error.