Controlling a Manufacturing Process with a Multivariate Model

• US 20100191361A1
• Filed: 01/23/2009
• Published: 07/29/2010
• Est. Priority Date: 01/23/2009
• Status: Active Grant

• Current Assignee: MKS Instruments AB (Sartorius AG)
• Original Assignee: MKS Instruments Incorporated
• Inventors: Wold, Svante Bjarne, McCready, Christopher Peter
• Granted Patent: US 9,069,345 B2
• Time in Patent Office: Days
• US Class Current: 700/104
• CPC Class Codes: C12M 41/48 Automatic or computerized control automatic analysis G01N35/00
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: G05B 2219/42061 Stochastic predictive controller spc

Abstract

A method, controller, and system for controlling a manufacturing process (batch-type or continuous-type) with a multivariate model are described. Dependent variable data and manipulated variable data are received. Dependent variable data represents values of uncontrolled process parameters from a plurality of sensors. Manipulated variable data represents controlled or setpoint values of controllable process parameters of a plurality of process tools. A predicted operational value, multivariate statistic, or both are determined based on the received data, and operating parameters of the manufacturing process are determined based on the predicted score, multivariate statistic, or both.

Claims

1. A computer-implemented method for controlling a manufacturing process, the method comprising:

receiving dependent variable data measured during the manufacturing process, the dependent variable data representative of values of a first set of process parameters observed by one or more sensors;

receiving manipulated variable data measured during the manufacturing process from a plurality of process tools and predicted manipulated variable data representative of a second set of process parameters.

determining at least one of a predicted score value, a multivariate statistic, or both based on at least the received data; and

determining operating parameters of the manufacturing process based on at least the predicted score value, the multivariate statistic, or both.

2. The method of claim 1, wherein the second set of process parameters, represented by manipulated variable data, are controlled during the manufacturing process.

3. The method of claim 2, wherein the first set of process parameters, represented by dependent variable data, are not directly controlled during the manufacturing process.

4. The method of claim 1, wherein the operating parameters comprise values for manipulated variables, further comprising providing the manipulated variable values to the plurality of process tools.

5. The method of claim 1, further comprising modifying the present or future values of the manipulated variables based on the past or present values of the manipulated variables and of the dependent variables.

6. The method of claim 1, further comprising receiving predicted values for the dependent variable data.

7. The method of claim 1, further comprising predicting values of the dependent variable data.

8. The method of claim 1, wherein determining operating parameters further comprises satisfying a controller objective.

9. The method of claim 8, wherein satisfying a controller objective comprises optimizing an operational objective function associating values of process data, yield data, result data, or any combination thereof of the manufacturing process.

10. The method of claim 9, wherein the objective function includes one or more constraints on the dependent variable data, the manipulated variable data, the predicted score value, the multivariate statistic, or any combination thereof.

11. The method of claim 10, wherein the one or more constraints are user-specified.

12. The method of claim 10, wherein the one or more constraints are associated with penalties for deviating from a multivariate model.

13. The method of claim 8, wherein the controller objective is a quadratic-type function and satisfying the controller objective further comprises minimizing a parameter of the objective function.

14. The method of claim 1, further comprising determining desirable values of the score or the multivariate statistic.

15. The method of claim 1, further comprising using a dependent variable model that predicts values of predicted dependent variable data based on determined values of the manipulated variable data, the past or present values of the dependent variable data, or any combination thereof.

16. The method of claim 1, further comprising using a score model that predicts future values of the first set of process parameters, represented by dependent variable data, and the second set of process parameters, represented by manipulated variable data.

17. The method of claim 1, further comprising a determining, with a multivariate model, the predicted value for the score, the multivariate statistic, or both.

18. The method of claim 17, wherein the multivariate model receives measured manipulated and dependent variable data and predicted manipulated and dependent variable data.

19. The method of claim 1, wherein the multivariate statistic comprises at least a score, a Hotelling'"'"'s T² value, a DModX value, a residual standard deviation value, or any combination thereof.

20. The method of claim 1, wherein the multivariate statistic comprises a principal components analysis t-score or partial least squares analysis t-score.

21. The method of claim 1, wherein the manufacturing process is a batch-type manufacturing process.

22. The method of claim 1, wherein the manufacturing process is a continuous-type manufacturing process.

23. A multivariate controller for a continuous or batch-type manufacturing process, the controller comprising:

a control module in communication with a plurality of process tools and a plurality of sensors to monitor manipulated variable data from the process tools and dependent variable data from the sensors, the control module including a multivariate model, the dependent variable data representative of values of a first set of process parameters observed by the plurality of sensors and the manipulated variable data representative of expected values of a second set of process parameters; and

a solver module to receive, from the multivariate model, at least one of predicted yield data, predicted manipulated variable data, predicted dependent variable data, a multivariate statistic, or any combination thereof, based on at least the monitored manipulated variable and dependent variable data and to generate values of the manipulated variables for providing to the plurality of process tools and to a prediction model that provides at least predicted statistical data.

24. The controller of claim 23, wherein the controller adjusts one or more parameters of the plurality of process tools based on the generated values of the manipulated variables.

25. The controller of claim 23, wherein the multivariate statistic includes at least one of a score, Hotelling'"'"'s T² value, a DModX value, a residual standard deviation value, or any combination thereof.

26. The controller of claim 25, wherein the prediction model includes a score model to generate predicted values for the solver module of one or more multivariate statistics.

27. The controller of claim 23, wherein the prediction model includes a dependent variable model to generate predicted dependent variable values.

28. The controller of claim 23, wherein the prediction model provides predicted values of dependent variables from the manufacturing process.

29. The controller of claim 23, wherein the prediction model provides the predicted statistical data to the control module and the solver module.

30. The controller of claim 23, wherein the solver module generates values of the manipulated variables based on a controller objective.

31. The controller of claim 30, wherein the controller objective optimizes a quadratic-type function associated with the manufacturing process.

32. The controller of claim 30, wherein the controller objective includes one or more constraints on the dependent variable data, the manipulated variable data, the predicted yield data, the multivariate statistic, or any combination thereof.

33. The controller of claim 32, wherein the one or more constraints are user-specified.

34. The controller of claim 32, wherein the one or more constraints are associated with penalties for deviating from the multivariate model.

35. The controller of claim 23, wherein the solver module is a constrained optimization solver.

36. The controller of claim 23, wherein the control module comprises the solver module.

37. A system for controlling a manufacturing process, the system comprising:

a data acquisition means for acquiring, from a plurality of process tools, manipulated variable data representative of expected values of a set of process parameters and acquiring, from a plurality of sensors, dependent variable data representative of values of a second set of process parameters observed by the plurality of sensors;

a process control means for determining operational parameters of the plurality of process tools of the manufacturing process;

a multivariate control means to determine, based on a multivariate statistical model, values for the manipulated variable data to provide to the process control means, the multivariate statistical model receiving at least the acquired manipulated variable and dependent variable data and providing predicted yield values and statistical information to the process control means based on at least the received data.

38. The system of claim 37, wherein the operational parameters determined by the process control means optimize or satisfy a control objective.