基于改进核主元分析的TE过程故障诊断

Fault diagnosis of Tennessee Eastman process based on improved kernel PCA

对于复杂非线性化工过程,传统的核主元分析(KPCA)方法在故障检测方面明显优越于普通的PCA方法,但存在故障辨识效果差的问题,而且实际得到的数据不可避免地带有噪声、随机干扰。针对此,提出一种改进的核主元分析方法,对数据进行小波去噪预处理,利用核主元分析方法进行故障检测,并利用计算核函数的偏导方法求取KPCA监控中每个原始变量对统计量T2和SPE的贡献率,根据每个变量对监控统计量贡献程度的不同,可以辨识出故障源。把上述方法应用到TE(Tennessee Eastman)化工过程,仿真结果表明该方法不仅能够去噪、抗干扰和准确检测故障,而且能够有效辨识故障。

For complex chemical industry process,the traditional kernel principal component analysis（KPCA） method is superior to the common principal component analysis（PCA） method obviously in fault detection.But the original fault sources are difficult to identify by KPCA,and the received data inevitably have noise,random disturbance.So an improved KPCA method is developed.The wavelet denoising method is used for data processing,the KPCA method is used for fault detection,once fault was detected,the gradient algorithm of kernel function is used to calculate the contribution of each original variable for Hotelling T2 and SPE,the fault variables might be identified from these correlative variables according to the degree of contribution.The proposed method is applied to the TE（Tennessee Eastman） process.The simulation results demonstrate that the proposed method could not only denoise and remove disturbances,but also effectively identify various types of fault sources.