提出了一種基于神經網絡的傳感器故障監測與診斷的新方法. 該方法先用BP 網絡的預測輸出和傳感器實際輸出之差來判斷傳感器是否發生了故障,然后用函數型連接神經網絡模擬傳感器的輸出特性函數,通過計算神經元連接權值的變化,確定傳感器哪個輸出特性參數發生了變化最終推斷傳感器發生了哪一類故障. 該方法的特點是只需要知道一個傳感器的信息. 電阻應變式力傳感器故障診斷實驗結果證明了該方法的實用性,為傳感器故障監測與診斷提供了一條新途徑.

A new method for detecting and diagnosing a sensor failure is purposed. A sensor in failure is located through the comparison between the forecast output of back propagation (BP) neural network and the actual one of a sensor. The simulation of sensor output characteristic function with functional2link neural network and the calculation of discrepancy of synaptic2weight are employed to confirm the changes of an individual output characteristic parameter , and then classifies the failure.
This new method needs the signal of one sensor only and has been proved to be unique and practical by the test result of st rain sensor failure detection and diagnosis.