论文部分内容阅读
传统的多元统计过程控制(MSPC)的故障诊断方法要求观测变量数据服从高斯分布,然而实际化工流程中的仪表数据中难以满足这一要求。针对这一问题,提出在仪表数据中提取分离出非高斯信息和高斯信息,并分别利用独立元分析法和主元分析法建立不同的故障诊断模型。在检测到发生故障后,通过改进的贡献度算法定位出发生故障的仪表。通过对Tennessee Eastman(TE)过程数据进行仿真研究,验证了ICA-PCA故障诊断法在化工流程仪表不同故障诊断中的有效性。
The traditional fault diagnosis method of MSPC requires observational data obeying Gaussian distribution, however, it is difficult to meet this requirement in the instrument data of the actual chemical process. To solve this problem, we propose to extract and separate non-Gaussian information and Gaussian information in the instrument data, and to establish different fault diagnosis models by means of independent element analysis and principal component analysis respectively. After detecting the occurrence of a fault, the faulty instrument is located by an improved contribution algorithm. Through the simulation of Tennessee Eastman (TE) process data, the effectiveness of ICA-PCA fault diagnosis in different fault diagnosis of chemical process instruments is verified.