Detecting anomalous events using principal component analysis.

摘要

A control system processes for detecting abnormal events in a process with one or more independent variables and one or more dependent variables, which system comprises means (14-21) for measuring values ​​of the one or more independent and dependent variables, means (22) process control comprising a predictive model for calculating predicted values ​​(28-31) of said one or more dependent variables from the measured values ​​of said one or more independent variables values, means (23) for calculating residual values ​​for the one or more dependent from the difference between the predicted values ​​and measured values ​​of said one or more dependent variables, and means (23) variables for performing a principal component analysis on the residual values, wherein means (22) process control means is a multivariable predictive control and principal component analysis results in the delivery of one or more scores values, values ​​T 2 and Q values, identifying an abnormal event from one or more thereof, characterized in that the contributions variable Q or T 2 (109, 108) of a data point (107) associated with an abnormal event are compared to a database of variable contributions Q or T 2 associated with known abnormal events in order to identify the beginning of a known abnormal event.