基于神经网络的传感器故障诊断方法与应用

Neural Network Based Method for Transducer Fault Diagnostics and Its Application

针对热工过程动态变化的特点 ,提出一种基于神经网络的传感器故障诊断方法。使用神经网络建立传感器的动态预测模型 ,根据当前传感器的测量值来产生下一时刻的预测值 ,分析预测值与实际测量值之间的残差 ,得出传感器故障与否的结论。仿真实例表明该方法能有效判别传感器的故障。

In accordance with the features of dynamic changing process the method based on neural network for transducer fault diagnostics is recommended. By using neural network, the dynamic prediction model of transducer is established. From the current measurement of transducer, the next value is predicted. By analyzing the residual error between predicted value and actual measurement, the conclusion can be obtained to see if transducer is health or not. The simulation test shows that the method can judge the fault effectively. The engineering applicable strategy of this method is also given.