System, method, and computer program for early event detection

• US 20070088534A1
• Filed: 10/18/2006
• Published: 04/19/2007
• Est. Priority Date: 10/18/2005
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Mylaraswamy, Dinkar, Foslien, Wendy, MacArthur, J., Srinivasrao, Mohan
• Granted Patent: US 7,734,451 B2
• Time in Patent Office: Days
• US Class Current: 703/17
• CPC Class Codes: G05B 13/0295 using fuzzy logic and expert systems
• CPC Class Codes: G05B 17/02 electric
• CPC Class Codes: G06N 7/02 using fuzzy logic computing arrangements based on biological models G06N3/00; computing arrangements using knowledge-based models G06N5/00

Abstract

Various methods, devices, systems, and computer programs are disclosed relating to the use of models to represent systems and processes (such as manufacturing and production plants). For example, a method may include generating a first model and a second model using operating data associated with a system or process. The method may also include using the first and second models to predict one or more events associated with the system or process. The one or more events are predicted by generating one or more initial event predictions using the first model and adjusting the one or more initial event predictions using the second model. The first model may represent a Principal Component Analysis (PCA) model, and the second model may represent a Fuzzy Logic model.

Claims

1. A method, comprising:

generating a first model and a second model using operating data associated with a system or process; and

using the first and second models to predict one or more events associated with the system or process by;

generating one or more initial event predictions using the first model; and

adjusting the one or more initial event predictions using the second model.

2. The method of claim 1, wherein the first model comprises a Principal Component Analysis (PCA) model, and the second model comprises a Fuzzy Logic model.

3. The method of claim 1, wherein adjusting the one or more initial event predictions comprises at least one of:

normalizing, shaping, and tuning an output of the first model using the second model, the output of the first model representing the one or more initial event predictions.

4. A method, comprising:

generating multiple first output signals using multiple first models representing a system or process;

normalizing the first output signals using one or more second models to produce multiple normalized second output signals; and

combining the normalized second output signals.

5. The method of claim 4, wherein the first models comprise Principal Component Analysis (PCA) models, and the one or more second models comprise one or more Fuzzy Logic models.

6. A method, comprising:

identifying at least one invalid data element in data associated with a system or process;

estimating a value for one or more of the invalid data elements using at least some valid data elements in the data associated with the system or process, wherein the estimating comprises performing a rank revealing QR factorization of a 2-norm known data regression problem; and

using the estimated value for one or more of the invalid data elements to predict one or more events associated with the system or process.

7. The method of claim 6, wherein the 2-norm known data regression problem is expressed as:

$\underset{{\mathrm{\beta <br><br>}}_j}{\min }<br><br>=\frac{1}{2}<br><br>{<br><br>{{\mathrm{DP}}^{*}}^T<br><br>{\mathrm{\beta <br><br>}}_j-d_j<br><br>}_2^2$

where;

β

_j represents a j^th column of a matrix β

, which represents an estimate of the data associated with the system or process;

D represents a matrix defined by D=Σ

^0.5, where Σ

represents a singular value matrix having diagonal elements each equal to a singular value or zero;

P*^T represents a transpose of a matrix P*, which contains the valid data elements; and

d_j represents a j^th column of the matrix D.

8. The method of claim 7, wherein performing the rank revealing QR factorization comprises performing a rank revealing QR factorization of DP*^T without determining P*Σ

P*^T.

9. A method, comprising:

presenting a graphical display to a user, the graphical display plotting a range of data values for one or more variables associated with a model of a system or process;

receiving from the user through the graphical display an identification of one or more portions of the data range; and

training the model without using any identified portions of the data range.

10. The method of claim 9, wherein the graphical display further comprises an identification of one or more events associated with the system or process.

11. The method of claim 10, wherein at least one identified portion of the data range corresponds to at least one event associated with the system or process.

12. The method of claim 11, further comprising receiving input from the user defining the at least one event associated with the system or process.

13. A method, comprising:

receiving information identifying multiple first events associated with a system or process;

generating at least one model using operating data associated with the system or process, the operating data not including operating data associated with the identified events; and

testing the at least one model by comparing second events predicted using the at least one model to the first events.

14. The method of claim 13, further comprising generating a graphical display comprising a plot identifying an error associated with the at least one model, the plot comprising:

one or more first indicators identifying the first events; and

one or more second indicators identifying the second events.

15. The method of claim 14, further comprising:

receiving input from a user altering one or more aspects of the at least one model;

updating the second events predicted using the at least one model; and

updating the second indicators in the plot to identify the updated second events.

16. A method, comprising:

presenting a graphical display to a user, the graphical display comprising;

a first plot identifying an error associated with a model representing a system or process; and

a second plot associated with one or more variables contributing to the error associated with the model; and

receiving input from the user identifying how the one or more variables associated with the second plot are selected.

17. The method of claim 16, wherein the input specifies the number of variables associated with the second plot and identifies a technique for selecting the one or more variables, the technique comprising one of:

identifying the specified number of variables that contribute most to the error related to the model; and

identifying up to the specified number of variables when the error related to the model exceeds a threshold specified by the user.

18. The method of claim 16, wherein:

the model comprises a Principal Component Analysis (PCA) model associated with multiple principal components; and

the graphical display further comprises a third plot identifying scores associated with two of the principal components identified by the user.

19. The method of claim 18, wherein the third plot comprises a confidence ellipse identifying which scores do and do not fall within a confidence interval specified by the user.

20. A method, comprising:

configuring a model representing a system or process in an off-line module for on-line use, the off-line module operable to create the model, wherein the configuring occurs automatically and is specified completely within the off-line module; and

providing the configured model to an on-line module for use in predicting events associated with the system or process.

21. The method of claim 20, wherein configuring the model comprises configuring multiple models and combining the multiple models into a single application, the on-line module operable to execute the single application.