面向工业平稳/非平稳复杂系统的在线故障监测技术

针对复杂系统中关键性能指标（KPI）相关故障检测方法检测精度低的问题，提出基于双层改进潜结构投影（DL-IPLS）的KPI相关故障检测方法. 利用协整分析和主元分析建立底层模型，对非平稳和平稳变量进行特征提取. 将提取的信息进行融合，建立改进潜结构投影的上层模型，根据融合信息对KPI的贡献进行空间分解. 在2个正交子空间中设计统计量，实现KPI相关故障的在线监测. 田纳西-伊斯曼过程和青霉素发酵过程的仿真结果表明，在面向工业平稳和非平稳复杂工业系统检测时，所提方法有效提高了KPI相关故障的检测率，降低了KPI无关故障的误报率.

On-line fault monitoring technology for industrial stationary/nonstationary complex system

A KPI related fault detection method based on double-layer improved projection to latent structures (DL-IPLS) was proposed in view of the low detection accuracy of key performance indicators (KPI) related fault detection methods in complex systems. The underlying model was established by using cointegration analysis and principal component analysis in order to extract the features of non-stationary and stationary variables. Then the extracted information was fused to establish the upper model of improved projection to latent structures. Statistics were designed in two orthogonal subspaces to realize on-line monitoring of KPI related faults. The simulation results of Tennessee-Eastman process and penicillin fermentation process show that the proposed method can effectively improve the detection rate of KPI related faults for industrial stationary and non-stationary complex industrial systems, and reduce the false alarm rate of KPI independent fault.