中文摘要：

针对动态系统的在线故障诊断问题，将信度分配小脑神经网络CA-CMAC（Credit Assigned Corebellar Model Articulation Controller）应用于主元分析模型，实现多传感器在线故障检测与隔离。首先，应用传感器正常工作时测量的历史数据，由主元分析模型得到所有传感器的预测值；接着计算传感器系统的均方预期误差值SPE（Squared Prediction Error），由SPE值的变化，判定是否发生故障，根据重构单个传感器信号的SPE值来隔离故障传感器；最后应用一个多传感器故障诊断仿真实例说明了该算法的可行性，并通过与误差反传BP（Back Propagation）神经网络和常规小脑神经网络CMAC（Cerebellar Model Articulation Controller）进行比较，说明了基于CA-CMAC的主元分析模型的优越性。

英文摘要：

For the problem of fault diagnosis in dynamic system, a principal component analysis model based on credit assigned cerebellar model articulation controller is proposed to carry out on-line fault detection and isolation for multiple sensor system.Firstly,the forecasting values of sensors are available from historical data measured in fault-free condition based on principal component analysis model.Secondly,the Squared Prediction Error of the system is calculated,the fault occurred when the SPE is suddenly increased. Sensor values are reconstructed respectively to newly calculate the SPE to locate the faulty sensor. Finally, Compared to BP and CMAC,the method proposed is proved feasible and effective by a simulation of multiple sensor fault diagnosis.