Method and system of using inferential measurements for abnormal event detection in continuous industrial processes

• US 8,285,513 B2
• Filed: 02/15/2008
• Issued: 10/09/2012
• Est. Priority Date: 02/27/2007
• Status: Active Grant

• Current Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Original Assignee: ExxonMobil Research & Engineering Company (Exxon Mobil Corporation)
• Inventors: Emigholz, Kenneth F.
• Primary Examiner(s): KUNDU, SUJOY K
• Application Number:
  US12/070,253
• Publication Number:
  US 20080281557A1
• Time in Patent Office: 1,698 Days
• Field of Search: 702/33, 702/34, 702 57- 59, 702 81- 84, 702/105, 702179-184, 702190-195, 201/1, 703/3, 700108-110, 700209-211
• US Class Current: 702/179
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Abstract

The present invention is a method and system for detecting an abnormal on-line analysis or laboratory measurement and for predicting an abnormal quality excursion due to an abnormal process condition.

Claims

1. An on-line system for calculating inferential estimators of key process conditions from normal process operating data of petro-chemical and industrial processes comprising:

a) real-time data collector,

b) real-time data preprocessor,

c) steady state and/or dynamic inferential models wherein said models estimate values for the current quality value (Y_estimate) and predict abnormal values for the future quality value (Y_predicted), and the inputs to said inferential models for calculating (Y_predicted) are not time synchronized,

d) statistical significance calculator of estimates of the models,

e) means for summarizing multiple model results, and

f) predictive displays for future abnormal events of said continuous petro-chemical and industrial process, and

g) diagnostic displays for event analysis.

2. The system of claim 1 wherein said data collector measures process measurements as inputs to the model.

3. The system of claim 2 wherein said process measurements include temperature, pressure, and flow rates.

4. The system of claim 1 wherein said data preprocessor includes a gross error detection of the data.

5. The system of claim 1 wherein data preprocessor includes mathematical transformation of the data.

6. The system of claim 1 wherein said data preprocessor includes a filter to remove noise.

7. The system of claim 1 wherein said inferential models and said statistical significance calculator determines the current operating parameter (Y_estimate), the future predicted operating parameter (Y_predicted), and long-term drift problems in either the model and/or raw measurements.

8. The system of claim 7 wherein said inferential models and said statistical significance calculator determines an estimate of the value of the operating parameter.

9. The system of claim 7 wherein said inferential estimator determines the predicted future value of the final steady state value of the operating parameter if all the inputs remain at their current level (Y_predict).

10. The system of claim 7 wherein said model and statistical significance calculator further determines the future operating parameter for process control application (Y_control).

11. The system of claim 7 wherein inputs to the model for Y_estimates are time synchronized with the corresponding lab analysis or on-line analysis process measurements.

12. The system of claim 1 wherein said inferential model can be built using Kalman Filters, partial least squares, regression, neural nets and stepwise regression.

13. The system of claim 11 wherein said inferential model is a partial least squares model.

14. The system of claim 11 wherein said model includes a bias term.

15. The system of claim 14 wherein said bias term is updated whenever an on-line analysis or lab analysis of operating parameters is available.

16. The system of claim 1 wherein said statistical significance calculator determines whether a statistically significant difference exists between the output of the inferential estimator and the measurement corresponding to the output of the inferential estimator.

17. The system of claim 1 wherein said model and statistical significance calculator determines an estimate of the value of the operating parameter.

18. The system of claim 9 wherein said model determines future operating limit excursions by comparing the predicted value of the operating parameter to the operating limit.

19. The system of claim 1 wherein said inferential model includes the identification of surrogate measurements for laboratory analyses.

20. The system of claim 1 wherein said inferential model includes the identification of periods of steady operation and the elimination of most steady-state operations.

21. The system of claim 1 wherein said inferential model includes identification of normal ranges for inputs and outputs measurement.

22. The system of claim 1 wherein said inferential model includes data filters to exclude data during times of very unusual operations.

23. The system of claim 1 wherein said inferential model includes single-input, single-output models.

24. The system of claim 1 wherein said inferential model includes the removal of slow trends from the data.

25. The system of claim 5 wherein transformations are used to time synchronize input data.

26. The system of claim 9 wherein said estimator of the future predicted operating parameter (Y_predicted) is used to determine future operating limit excursions by comparing the predicted value of the operating parameter to the operating limits.

27. The system of claim 17 wherein the estimate includes a bias term.

28. The system of claim 27 wherein said bias term is updated with a measurement corresponding to the output of the inferential estimator.