Abnormal event detection using principal component analysis

1. A process control system for detecting abnormal events in a process having one or more independent variables and one or more dependent variables, which system comprises:

means for measuring values of the one or more independent and dependent variables,

a process control means comprising a predictive model for calculating predicted values of the one or more dependent variables from the measured values of the one or more independent variables,

a means for calculating residual values for the one or more dependent variables from the difference between the predicted and measured values of the one or more dependent variables, and

a means for performing a principal component analysis on the residual values,

wherein the process control means is a multivariable predictive control means and the principal component analysis results in the output of one or more scores values, T2 values and Q values, an abnormal event being identified from one or more thereof, and wherein the Q or T2 contributions of a data point associated with an abnormal event are compared to a database of Q or T2 contributions associated with known abnormal events in order to identify the onset of a known abnormal event.

2. A system as claimed in claim 1 , in which the process control means is a dynamic matrix control system.

3. A system as claimed in claim 1 , which additionally comprises a means for applying a correction to the predicted values of one or more of the dependent variables or residuals to account for mismatch between the predicted and measured values of the dependent variables in the absence of an abnormal event.

4. A system as claimed in claim 1 , in which the means for calculating the residual values and/or the means for performing a principal components analysis on the residual values and/or the means for applying a correction to the one or more predicted dependent variable values or residuals values is a programmed computer.

5. A system as claimed in claim 1 , in which the results of the principal components analysis are fed back to the process control means.

6. A system as claimed in claim 5 , in which corrective action associated with an identified abnormal event is automatically fed to the process control means.

7. A system as claimed in claim 1 , in which the process control means has more than one predictive model for different stable modes of operation of the process.

8. A system as claimed in claim 1 , in which the process is a refinery process.

9. A system as claimed in claim 8 , in which the refinery process is the distillation of crude oil.

10. A method for detecting abnormal events in a process, which process comprises one or more independent variables and one or more dependent variables, wherein the method is performed using a process control system comprising means for measuring values of the one or more independent and dependent variables, and a process control means comprising a predictive model for calculating predicted values of the one or more dependent variables from the measured values of the one or more independent variables, which method comprises:

measuring values of the one or more independent variables and the one or more dependent variables,

calculating predicted values of the one or more dependent variables from the measured values of the one or more independent variables,

calculating residual values for the one or more dependent variables from the differences between the predicted and measured values of the one or more dependent variables, and

performing a principal component analysis on the residual values,

wherein the process control means is a multivariable predictive control means and the principal component analysis results in the output of one or more scores values, T2 values and Q values, an abnormal event being identified from one or more thereof, and wherein the Q or T2 contributions of a data point associated with an abnormal event are compared to a database of Q or T2 contributions associated with known abnormal events in order to identify the onset of a known abnormal event.

11. A method as claimed in claim 10 , in which the process comprises a process control system.

12. A method as claimed in claim 10 , in which the process is a refinery process.

13. A method as claimed in claim 12 , in which the abnormal event is associated with the flash point of a fraction derived from a crude oil distillation tower.

14. A method as claimed in claim 13 , in which the fraction is kerosene.

15. A process control system for detecting abnormal events in a process having one or more independent variables and one or more dependent variables, which system comprises:

measuring device for measuring values of the one or more independent and dependent variables,

a process controller comprising a predictive model for calculating predicted values of the one or more dependent variables from the measured values of the one or more independent variables,

a calculator for calculating residual values for the one or more dependent variables from the difference between the predicted and measured values of the one or more dependent variables, and

an analyzer for performing a principal component analysis on the residual values,

wherein the process controller is a multivariable predictive controller and the principal component analysis results in the output of one or more scores values, T2 values and Q values, an abnormal event being identified from one or more thereof, and wherein the Q or T2 contributions of a data point associated with an abnormal event are compared to a database of Q or T2 contributions associated with known abnormal events in order to identify the onset of a known abnormal event.