Advance failure prediction

• US 6,915,173 B2
• Filed: 08/21/2003
• Issued: 07/05/2005
• Est. Priority Date: 08/22/2002
• Status: Expired due to Fees

• Current Assignee: IBEX Process Technology
• Original Assignee: IBEX Process Technology
• Inventors: Reitman, Edward A., Card, Jill P., Chan, Wai T.
• Primary Examiner(s): PATEL, RAMESH B
• Application Number:
  US10/645,209
• Publication Number:
  US 20040168108A1
• Time in Patent Office: 684 Days
• Field of Search: 700/21, 700/28, 700/44, 700/45, 700/79, 700/80, 700/81, 700/262, 714/1, 714/2, 714/20, 714/26, 714/57, 361/115, 361/127, 702/183, 702/185
• US Class Current: 700/44
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0283 Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06

Abstract

Failure prediction for complex processes is performed utilizing one or more nonlinear regression models to relate operational variable values measured at two or more times to predicted process metric values and maintenance variable values.

Claims

1. A method of predicting failure in a process having associated metrics and a plurality of operational variables, the method comprising the steps of:

using non-linear regression to predict values for a first set of operational variables based on two or more prior values thereof;

using non-linear regression to predict a plurality of process metric values based on the first set of predicted values and prior values of two or more operational variables; and

determining a likelihood of a process failure based on one or more of the predicted process metric values.

2. The method of claim 1, wherein the first set of predicted values is predicted by applying a separate non-linear regression model to each of the process operational variables, wherein each of the separate non-linear regression models has been trained in the relationship between a single process operational variable and prior values of two or more process operational variables.

3. The method of claim 2, further comprising repeating the steps of the method for at least one sub-process of the process.

4. The method of claim 2, further comprising repeating the steps of the method for a higher-level process comprising the process.

5. A method of predicting failure in a process having a plurality of operational variables associated therewith, the method comprising the steps of:

using non-linear regression to predict values at a first time for a first set of operational variables based on two or more prior values thereof;

using non-linear regression to predict values at a second time for a second set of operational variables based on two or more prior values thereof; and

determining a likelihood of a process failure based on one or more of the predicted values for the process operational variables at the first time and the second time.

6. The method of claim 5, wherein the second set of process operational variables are at least a subset of the first set of process operational variables.

7. The method of claim 5 further comprising using non-linear regression to predict values at a third time for a third set of operational variables based on two or more prior values thereof.

8. The method of claim 7 further comprising using non-linear regression to predict values at a fourth time for a fourth set of operational variables based on two or more prior values thereof.

9. A method of predicting the need for maintenance activities for a process having a plurality of operational variables associated therewith, the method comprising the steps of:

using non-linear regression to predict values at a first time for a first set of operational variables based on two or more prior values thereof;

using non-linear regression to predict values at a second time for a second set of operational variables based on two or more prior values thereof; and

determining the need for a maintenance action based at least in part on a comparison of the first set of predicted values with the second set of predicted values.

10. The method of claim 9 wherein the determining step is also based on process yield metrics.

11. The method of claim 9 wherein the step of predicting the second set of values at the second time is also based on process yield metrics.

12. The method of claim 9 wherein the first set of predicted values at the first time is constrained by cost data.

13. The method of claim 9 wherein the second set of predicted values at the second time is constrained by cost data.

14. A system for predicting events of a process having associated operational variables, the system comprising:

(a) a process monitor for monitoring operational variables; and

(b) a data processing device for receiving, from the process monitor, data indicative of values of the operational variables, and predicting events based on (i) a relationship between a first set of predicted values for a first set of process operational variables and two or more prior process operational variable values thereof, (ii) a relationship between a second set of predicted values for a second set of process operational variables and two or more prior process operational variable values thereof, and (iii) a relationship between a predicted process event, the first set of predicted values, and the second set of predicted values.

15. The system of claim 14 wherein the process event is a process failure.

16. The system of claim 14 wherein the process event is a maintenance activity.

17. A system for predicting failure in a process having associated operational variables, the system comprising:

(a) a process monitor for monitoring operational variables; and

(b) a data processing device for receiving, from the process monitor, data indicative of values of the operational variables, and predicting process failure based on (i) a relationship between a first set of predicted values at a first time for a first set of process operational variables and two or more prior process operational variable values thereof, (ii) a relationship between a second set of predicted values at a second time for a second set of process operational variables and two or more prior process operational variable values thereof, and (iii) a relationship between a process failure, the first set of predicted values at a first time, and the second set of predicted values at a second time.

18. A system for predicting the need for maintenance activities in a process having associated operational variables and process metrics, the system comprising:

(a) a process monitor for monitoring operational variables; and

(b) a data processing device for receiving, from the process monitor, data indicative of values of the operational variables, and predicting process failure based on (i) a relationship between a first set of predicted values at a first time for a first set of process operational variables and two or more prior process operational variable values thereof, (ii) a relationship between a second set of predicted values at a second time for a second set of process operational variables and two or more prior process operational variable values thereof, and (iii) a relationship between a need for a maintenance activity, the first set of predicted values at a first time, and the second set of predicted values at a second time.

19. The system of claim 18 further comprising a process controller, responsive to the data processing device, for performing maintenance activities based on the predicted process event.

20. The system of claim 18 further comprising a data storage device for storing one or more of maintenance activity records and maintenance activity costs.