Model-based predictive control of thermal processing

• US 6,207,936 B1
• Filed: 01/30/1997
• Issued: 03/27/2001
• Est. Priority Date: 01/31/1996
• Status: Expired due to Term

• Current Assignee: ASM America Incorporated (ASM International NV)
• Original Assignee: ASM America Incorporated (ASM International NV)
• Inventors: Lu, Zhimin, Donald, James J., de Keyser, Robin M., de Waard, Henk
• Primary Examiner(s): PASCHALL, MARK H
• Application Number:
  US08/791,134
• Time in Patent Office: 1,517 Days
• Field of Search: 219/494, 219/497, 219/501, 219/506, 219/508, 219/483, 219/486, 219/412-414, 364/151, 364/156, 364/157, 364/165, 364/150, 364/148, 364/149, 364/159, 392/416-420
• US Class Current: 219/497
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

A nonlinear model-based predictive temperature control system is described for use in thermal process reactors. A multivariable temperature response is predicted using a nonlinear parameterized model of a thermal process reactor. The nonlinear parameterized model is implemented using a neural network. Predictions are made in an auto-regressive moving average fashion with a receding prediction horizon. Model predictions are incorporated into a control law for estimating the optimum future control strategy. The high-speed, predictive nature of the controller renders it advantageous in multivariable rapid thermal processing reactors where fast response and high temperature uniformity are needed.

Claims

1. A temperature controlled thermal process reactor comprising:

a reaction chamber enclosing an object to be heated;

a plurality of sources of thermal energy which heat said object;

a plurality of thermal sensors, each sensor configured to measure a sensor temperature, each sensor temperature related to an actual temperature of said object, where each thermal sensor provides an output signal representative of said sensor temperature, and where each of said sources of thermal energy affects each of said sensor temperatures; and

a model-based predictive temperature controller comprising a nonlinear process model, said temperature controller configured to receive said output signals and control said sources of thermal energy in response to said output signals to produce a selected spatial and temporal distribution of heat energy to maintain a relatively uniform actual temperature on said object, said model-based predictive temperature controller comprising a multivariable thermal process model that relates multivariable process input thermal energy to multivariable process output temperature, a prediction calculator that uses said thermal process model to calculate a predicted nominal temperature output over a future time period, and a control calculator that uses said predicted nominal temperature output to calculate an optimum control strategy to control said sources of thermal energy, said prediction calculator configured to calculate said predicted nominal temperature output recursively over a predetermined future time period using a recursive approximation strategy that begins with an unoptimized initial estimate.

2. The temperature controlled thermal process reactor of claim 1, wherein said prediction calculator calculates the predicted nominal temperature output using said nonlinear process model in a parallel recursion, said prediction calculator having a predetermined prediction horizon.

3. The temperature controlled thermal process reactor of claim 2, wherein the prediction calculator assumes a future control strategy.

4. The temperature controlled thermal process reactor of claim 3, wherein said thermal process model is adapted to substantially decouple the influence of system input variables from system input disturbance.

5. The temperature controlled thermal process reactor of claim 1, wherein the control calculator is adapted to compare said predicted nominal temperature output to a desired future temperature output and uses said comparison in a recursive algorithm to compute said optimum control strategy.

6. The temperature controlled thermal process reactor of claim 1, wherein said thermal process model is based on a neural network.

7. The temperature controlled rapid thermal process reactor of claim 1, wherein the nonlinear model-based predictive temperature controller comprises:

a nonlinear multivariable thermal process model which relates multivariable process input thermal energy to multivariable process output temperature;

a prediction calculator which uses said thermal process model, to calculate a predicted nominal temp re output over a predetermined future time period; and

a control calculator which uses said predicted nominal temperature output to calculate an optimum control strategy by which to control said source of thermal energy.

8. The temperature controlled thermal process reactor of claim 7, wherein said prediction calculator calculates the predicted nominal temperature output using a neural network.

9. The temperature controlled thermal process reactor of claim 8, wherein the prediction calculator assumes a future control strategy.

10. The temperature controlled thermal process reactor of claim 9, wherein said neural network is a feed forward network.

11. The temperature controlled thermal process reactor of claim 10, wherein said neural network comprises a hidden layer of neurons.

12. The temperature controlled thermal process reactor of claim 4, wherein said hidden layer of neurons comprises nonlinear sigmoid-type neurons.

13. The temperature controlled thermal process reactor of claim 8, wherein said neural network is trained using a pseudo least squares method.

14. The temperature controlled thermal process reactor of claim 7, wherein the control calculator compares said predicted nominal temperature output to a desired future temperature output to derive said optimum control strategy.

15. The temperature controlled thermal process reactor of claim 1, further comprising a softsensor model.

16. The temperature controlled thermal process reactor of claim 15, wherein said softsensor model is created from a dataset generated by using an instrumented wafer.

17. The temperature controlled thermal process reactor of claim 1, further comprising a setpoint generator, said setpoint generator automatically generating a correction to said recipe inputs into said thermal process reactor, said correction facilitating control of actual wafer surface temperatures.

18. The temperature controlled thermal process reactor of claim 17, said correction facilitating improved control of actual wafer surface temperatures based on measurement of susceptor temperatures.

19. A temperature control system for controlling a thermal process comprising:

a controllable source of thermal energy which comprises multiple independently controllable thermal energy generators to heat an object;

a plurality of temperature sensors, each of which measures a temperature related to an actual temperature of said object and which generates an output signal responsive to said temperature; and

a model-based predictive temperature controller which receives said output signals and which controls said source of thermal energy in response to said output signals, said controller comprising;

a nonlinear thermal process model which relates a process input thermal energy to a process output temperature;

a prediction calculator which uses said nonlinear thermal process model to calculate a predicted nominal temperature output over a predetermined future time period; and

a control calculator which uses said predicted nominal temperature output to calculate an optimum strategy by which to control said source of thermal energy to minimize spatial thermal gradients across said object, said controller generating output signals to said source of thermal energy in response to said optimum strategy, said control calculator configured to compare a predicted nominal temperature output to a desired future temperature output and use said comparison in an algorithm to compute said optimum strategy.

20. The temperature control system of claim 19, wherein said thermal process model substantially decouples the influence of system input variables from system input disturbances.

21. The temperature control system of claim 19, wherein said prediction calculator is adapted to use a postulated future control strategy and a recursive algorithm to optimize said postulated future control.

22. The temperature control system of claim 19, wherein said nonlinear thermal process model comprises a neutral network, said neural network comprising one or more nonlinear sigmoid-type neurons.

23. The temperature control system of claim 22, wherein said thermal process model substantially decouples the influence of system input variables from system input disturbances.

24. The temperature control system of claim 22, wherein said prediction calculator comprises a neural network.

25. A method of controlling a thermal process comprising the steps of:

measuring one or more process output temperatures;

predicting a plurality of future process output temperatures by using a nonlinear thermal process model;

using said one or more measured process output temperatures and said predicted future process temperatures to calculate an optimum process input control strategy by comparing one or more of said predicted future process output temperatures to a desired future temperature and using said comparison in an algorithm to compute said optimum process input control strategy; and

controlling a process input thermal energy using the calculated optimum process input control strategy.

26. The method of claim 25, wherein the step of predicting a future process output temperate comprises:

identifying a nonlinear thermal process model which relates process input thermal energy to process output temperature; and

recursively predicting future process output temperatures using said nonlinear thermal process model, said process output temperature predicted over a predetermined future time period.

27. The method of claim 26, wherein the step of predicting future process output temperatures further comprises periodically updating said predictions in accordance with a receding horizon calculation.

28. The method of claim 25, wherein the step of predicting a future process output temperature comprises postulating a stationary future control strategy.

29. The method of claim 25, wherein the step of calculating an optimum process input control strategy comprises comparing said predicted future process output temperatures to a desired future process output temperature.

30. The method of claim 25, wherein the step of predicting a future process output temperature comprises:

identifying a nonlinear thermal process model which relates process input thermal energy to process output temperature; and

training a neural network to predict fixture process output temperatures using said thermal process model, said process output temperature predicted over a predetermined future time period.

31. The method of claim 30, wherein the step of predicting future process output temperatures firer comprises periodically updating said predictions in accordance with a receding horizon calculation.

32. The method of claim 30, wherein the step of predicting a future process output temperature comprises postulating a stationary future control strategy.

33. The method of claim 30, wherein the step of calculating an optimum process input control strategy comprises comparing said predicted future process output temperatures to a desired fixture process output temperature.

34. A temperature controlled thermal process reactor comprising:

a reaction chamber enclosing an object to be headed;

a plurality of sources of thermal energy which heat said object;

a plurality of thermal sensors, each sensor configured to measure a sensor temperature, each sensor temperature related to an actual temperature of said object, where each thermal sensor provides an output signal representative of said sensor temperature, and wherein each of said sources of thermal energy affects each of said sensor temperatures; and

a model-based predictive temperature controller which receive said output signals and which controls said sources of thermal energy in response to said output signals to produce a selected spatial and temporal distribution of heat energy to maintain a relatively uniform actual temperature on said object said model-based predictive temperature controller comprising a multivariable thermal process model that relates multivariable process input thermal energy to multivariable process output temperature, a prediction calculator that uses said thermal process model to calculate a predicted nominal temperature output over a predetermined future time period and a control calculator that uses said predicted nominal temperature output to calculate an optimum control strategy by which to control said sources of thermal energy said control calculator adapted to compare a predicted nominal temperature output to a desired future temperature output and use said comparison in a recursive algorithm to compute an optimum contort strategy.

35. The temperature controlled thermal process reactor of claim 34, wherein the model-based predictive temperature controller comprises multivariable temperature control.

36. The temperature controlled thermal process reactor of claim 34 wherein said prediction calculator calculates the predicted nominal temperature output using an auto-regressive moving average using a predetermined prediction horizon.

37. The temperature controlled thermal process reactor of claim 36, wherein the prediction calculator calculates an unoptimized initial estimate for a future control strategy.

38. The temperature controlled thermal process reactor of claim 37, wherein said prediction calculator is further adapted to calculate said predicted nominal temperature output recursively over a predetermined future time period using a recursive approximation strategy, said recursive approximation strategy beginning with said unoptimized initial estimate.

39. The temperature controlled thermal process reactor of claim 38, wherein said thermal process model is adapted to substantially decouple the influence of system input variables from system input disturbances.

40. A temperature control system for controlling a thermal process comprising:

a controllable source of thermal energy which comprises multiple independently controllable thermal energy generators to heat an object;

a plurality of temperature sensors, each of which measures a temperature related to an actual temperature of said object and which generates an output signal responsive to said temperature; and

a model-based predictive controller which receives said output signals and which controls said source of thermal energy in response to said output signals, said controller comprising;

a thermal process model that relates process input thermal energy to process output temperature;

a prediction calculator that uses said thermal process model to calculate a predicted nominal temperature output over a predetermined future time period; and

a control calculator that uses said predicted nominal temperature output to calculate an optimum strategy by which to control said source of thermal energy to minimize spatial thermal gradients across said object, said controller generating output signals to said source of thermal energy in response to said optimum strategy said control calculator configured to compare a predicted nominal temperature output to a desired future temperature output and use said comparison in an algorithm to compute said optimum strategy.

41. The temperature control system of claim 40, wherein said thermal process model has parameters selected to substantially decouple the influence of system input variables from system input disturbances.

42. The temperature control system of claim 40, wherein said prediction calculator is adapted to use a recursive algorithm to compute said a predicted nominal temperature output over a predetermined future time period.

43. The temperature control system of claim 40, wherein said prediction calculator comprises means for calculating an auto-regressive moving average, said means for calculating an autoregressive moving average comprising means for computing a receding calculation horizon.

44. The temperature control system of claim 40, wherein said prediction calculator is adapted to use a postulated future control strategy and a recursive algorithm to optimize said postulated future control strategy.

45. A method of controlling a thermal process comprising:

measuring a plurality of process output temperatures;

predicting a plurality of future process output temperatures;

using said measured process output temperatures and said predicted future process temperatures to calculate an optimum process input control strategy by comparing one or more of said predicted future process output temperatures to a desired future temperature and using said comparison in an algorithm to compute said optimum process input control strategy; and

controlling a plurality of process input thermal energy sources using the calculated optimum process input control strategy to reduce spatial thermal gradients.

46. The method of claim 45, when the step of predicting a future process output temperature comprises:

identifying a thermal process model that relates process input thermal energy to process output temperature; and

recursively predicting future process output temperatures using said thermal process model, said process output temperature predicted over a predetermined future time period.

47. The method of claim 46, wherein the step of predicting future process output temperatures further comprises periodically updating said predictions in accordance with a receding horizon auto-regressive moving average calculation.

48. The method of claim 45, wherein the step of predicting a future process output temperature comprises postulating a stationary future control strategy.

49. The method of claim 45, wherein the step of calculating an optimum process input control strategy comprises comparing said predicted future process output temperatures to a desired future process output temperature.