CONSTRUCTION OF FAULT DIAGNOSIS SYSTEM BY PRINCIPAL COMPONENT ANALYSIS BASED ON CHANGE OF PROCESS VARIABLES

摘要：

　　It is necessary to discriminate a failure promptly and to do corrective actions when failure occurs in the chemical plant.In previous studies, the feature of failure was clarified by using principal component analysis (PCA) for process variables data of a simulator, but the change of process variables was an unstable factor in 2-dimensional data space by PCA including process variables that there was no much difference between steady state and failure state.This study proposes a method to select the data space which is appropriate for discriminating the failure.Therefore, the process variables which dont feature failure are rejected in turn from analysis based on the change of process variables, and on each data space, selects the space which has small dispersion of same failure and large distance between steady state and failure state.

关键词： Fault Diagnosis System Principal Component Analysis Variable Selection

作者: Toshihiko Banjoya Yoshiomi Munesawa Kazuhiko Suzuki

作者单位: Okayama University, Okayama, Japan

会议类型: 国际会议

会议名称: The 11th International Conference on Industrial Management(第十一届工业管理国际会议)

会议地点: 日本

会议语种:英文

页码: 463-469

在线出版日期: 2012-08-29（万方平台首次上网日期，不代表论文的发表时间）