基于核主元分析与神经网络的传感器故障诊断新方法

提出综合利用核函数主元分析(KPCA)和神经网络的方法实现非线性系统内传感器故障的检测和识别,克服了以往核函数主元分析法只能给出故障检测结果,却无法对故障进行识别的缺陷,并给出了在特征空间中计算残差的简单方法.最后,对一个简单的多变量过程进行了故障检测和识别,验证了这一诊断策略的有效性.

Novel Sensor Fault Diagnosis Method Based on Kernel Principle Component Analysis and Neural Networks

A novel sensor fault diagnosis method for nonlinear system has been presented, which makes use of kernel principle analysis(KPCA) and neural networks to fulfill the task of fault detection and identification. Compared with conventional KPCA method, it not only gives out of fault detection results, but gives out of fault identification results. A simple approach to calculating the squared prediction error(SPE) in the feature space is also suggested. The method is verified by a simulation test for a simple multivariate process.