System and method for real-time industrial process modeling

• US 8,457,767 B2
• Filed: 12/31/2010
• Issued: 06/04/2013
• Est. Priority Date: 12/31/2010
• Status: Active Grant

• Current Assignee: Radl, Brad
• Original Assignee: Radl, Brad
• Inventors: Radl, Brad
• Primary Examiner(s): BARNES-BULLOCK, CRYSTAL JOY
• Application Number:
  US12/983,054
• Publication Number:
  US 20120173004A1
• Time in Patent Office: 886 Days
• Field of Search: 700/29, 700/31, 700/49, 700/73, 703/2, 706 14- 16, 706/18, 706/20, 706/23, 706 45- 48
• US Class Current: 700/49
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

The present invention presents two new model types and a new method for evaluating a model used in the control application. These include a compound model, a hybrid model and a directional change coefficient model. The present invention allows the mixing of models with different inputs and outputs and the switching between these models based criteria for measuring optimization accuracy. The present invention allows switching between these models. The compound model is a model type that allows zooming in on the process to model parts of the data space with higher fidelity or resolution without loosing the capability to model the complete data space. The modeler does not loose any functionally over a regular neural network, but instead gains the ability to define the conditions when the model should use network weights best matched to the defined local conditions. The hybrid model is an extended version of a compound model. A hybrid model allows the combining of one or more models into a single model for purposes of interrogation or optimization. Within the hybrid model may reside a compound model itself. The directional change model (DCC) allows better evaluation of the predictive capability of Compound Models. It may also be used with any other model type.

Claims

1. A modeling system adapted to control a particular industrial process comprising:

a plurality of sensors coupled to said industrial process;

a plurality of control lines coupled to said industrial process;

a computational engine configured to read said plurality of sensors and command said control lines;

said computational engine adapted to run a plurality of predetermined models relating to a data space within said industrial process, said computational engine also adapted to create sub-models from said predetermined models based on predetermined data thresholds, wherein a particular sub-model is created and trained if a number of data patterns exceeds one of said thresholds and said particular sub-model is not created if said number of data patterns does not exceed said threshold.

2. The modeling system of claim 1 further comprising a default model adapted to model an entire data space.

3. The modeling system of claim 1 adapted to receive user data defining said data space.

4. The modeling system of claim 1 adapted to receive user input determining an acceptable amount of data for said data space.

5. The modeling system of claim 1 wherein said computational engine sorts incoming data for said data space and saves said data as clusters.

6. The modeling system of claim 1 wherein outlying data is removed from said data space.

7. The modeling system of claim 1 wherein older data is removed from said data space according to predetermined age criteria.

8. The modeling system of claim 1 wherein said computational engine creates a hierarchy of sub-models of increasing resolution.

9. The modeling system of claim 1 adapted to combine of one or more models into a single model for interrogation or optimization.

10. The modeling system of claim 1 wherein at least one of said models is a evaluated by a directional change coefficient model (DCC).

11. The modeling system of claim 1 wherein said computational engine can use any combination of available inputs and outputs.

12. The modeling system of claim 1 wherein model types may be mixed.

13. The modeling system of claim 1 which allows pruning or addition of sensor data.

14. A computational modeling engine directly attached to an industrial process comprising:

at least one input sensor directly attached to said industrial process, wherein said input sensor provides real-time data from said industrial process, and wherein said computational modeling engine can read said data from said input sensor;

said computational modeling engine adapted to run at least one predetermined model relating to a data space within said industrial process, said computational modeling engine also adapted to create sub-models from said predetermined model based on data thresholds, wherein a particular sub-model is created if a number of data patterns exceeds one of said thresholds and said particular sub-model is not created if said number of data patterns does not exceed said threshold.

15. The computational modeling engine of claim 14 wherein said computational engine creates a hierarchy of sub-models of increasing resolution.

16. The computational modeling engine of claim 14 adapted to combine of one or more models into a single model for interrogation or optimization.

17. The computational modeling engine of claim 14 wherein at least one of said models is a evaluated and/or ranked directional change coefficient (DCC).

18. The modeling system of claim 14 which allows pruning or addition of sensor data.

19. A computational modeling system comprising:

a computational engine adapted to be attached to a plurality of independent control systems, each controlling an industrial process, wherein each of said independent control systems receives data from one or more sensors and controls one or more actuators, and wherein said independent controls systems have no exchange of data between systems;

said computational engine adapted to run at least one predetermined model relating to a data space within said industrial process related to at least some of said independent control systems, said computational engine also adapted to create sub-models from said predetermined model based on data thresholds, wherein a particular sub-model is created if a number of data patterns exceeds one of said thresholds and said particular sub-model is not created if said number of data patterns does not exceed said threshold.

20. The computational modeling system of claim 19 wherein said computational engine allows both independent actions of said independent control systems or dependent actions of said independent control systems.