基于LGRSAE算法的非线性化工过程故障检测  

Fault Detection for the Chemical Process Based on LGRSAE Algorithm

作　　者：杨景超 YANG Jing-chao(CNOOC Petrochemical Engineering Co.,Ltd.)

机构地区：[1]中海油石化工程有限公司

出　　处：《化工自动化及仪表》2025年第1期111-118,共8页Control and Instruments in Chemical Industry

摘　　要：针对常规线性降维方法不能有效提取实际复杂非线性工业数据局部和全局结构特征的问题,提出一种局部和全局保持堆栈自编码器(LGRSAE)以及基于LGRSAE的过程故障检测方法。在自编码器(AE)的目标函数中引入局部保持投影(LPP)算法和主成分分析(PCA)算法的目标约束构造局部和全局结构保持自编码器(LGRAE),以提取数据局部和全局结构相关的特征。为了提取过程数据深层局部和全局结构相关的特征,将LGRAE堆栈构成LGRSAE。在田纳西-伊斯曼(TE)过程数据集的故障检测结果表明,LGRSAE的特征提取方法平均故障检测率高于MRSAE、KPCA算法,且误报率更低。Aiming at the conventional linear dimension-reducing method’s incapability in extracting local and global structural features of the real complex nonlinear industrial data effectively,a local and global stack-retaining auto-encoder(LGRSAE)and a process fault detection method based on LGRSAE were developed.In addition,the objective constraints of the locality preserving projection(LPP)algorithm and the principal component analysis(PCA)algorithm were introduced into the objective function of the auto-encoder(AE)to extract the features related to the local and global structure of the data,including having LGRAE stacks constructed to form LGRSAE to extract features related to deep local and global structure of the process data.The fault detection results on the TE(Tennessee-Eastman)process data set show that,the average fault detection rate of the feature extraction method of LGRSAE is higher than those of the MRSAE and KPCA algorithms and the false alarm rate is lower.

关 键 词：故障检测 堆栈自编码器 特征提取 局部和全局结构保持 数据降维 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]