基于RBF时间序列预测器的传感器故障诊断方法研究  

Research on sensor fault diagnosis method based on RBF time series predictor

作　　者：曹正洪[1] 沈继红[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《传感器与微系统》2010年第5期63-65,69,共4页Transducer and Microsystem Technologies

摘　　要：研究了一种基于径向基函数(RBF)神经网络时间序列预测器诊断传感器故障的方法。以压力传感器的过载故障为模型,先用RBF神经网络建立时间序列预测模型,然后利用预测模型对传感器的输出作预测,再和传感器实际输出比较,从而判断传感器是否发生故障,并对发生故障的传感器进行数据重构。仿真实验证实了该方法可以有效地进行传感器故障诊断和数据重构,并可推广到其他传感器中。A sensor fault diagnosis method based on radial basis function （RBF） neural network time series predictor is proposed. RBF neural network is a good feed forward neural network with the best approximation and performance that can be used to overcome the problem of local minimum. The overload fault is given as model of pressure sensor. Time series predictor model of the sensor is set up by RBF neural network. The sensor output is predicted by using the above model. And then compared with the actual output of sensor to determine whether the sensor has fault. Fault data of the sensor is reconstructed. Simulation results confirm that this proposed method is effective for sensor fault diagnosis and data reconstruction. This method can be applied in other sensors.

关 键 词：传感器故障诊断 RBF神经网络 时间序列预测器 数据重构 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]