基于RBF网络的模拟移动床PH传感器故障诊断

SMB PH fault diagnosis based on RBF neural networks

为了提高模拟移动床控制系统PH传感器的可靠性,提出了一种基于两级RBF神经网络的故障诊断方法。该方法首先利用径向基(RBF)神经网络对传感器的输出序列建立预测模型,通过计算预测输出和实际输出的残差来检测故障的发生,然后对包含故障的残差信号利用小波变换进行特征提取,最后利用RBF诊断网络实现故障诊断。通过把这种方法应用到实际诊断测试中,可达到较准确的诊断结果。

In order to enhance the reliability of PH sensor in Simulation Movable Bed （SMB） control system, a fault diagnostic method based on two-step Radial Basic Function （RBF） neural network is proposed. Firstly RBF neural network is used to built predictor model with a serial of value of sensor, and the residual between output of predictor model and actual value is computed to determine whether sensor is working normal or not. Then wavelet transform is used to extract the sensor feature from fault residual signal, and finally RBF diagnostic neural network is used to diagnose faults. The application to the PH sensor test in SMB proves effective.