Metrics independent and recipe independent fault classes

• US 20080276128A1
• Filed: 05/04/2007
• Published: 11/06/2008
• Est. Priority Date: 05/04/2007
• Status: Active Grant

• Current Assignee: Applied Materials Incorporated
• Original Assignee: Applied Materials Incorporated
• Inventors: Schwarm, Alexander T., Lin, Y. Sean, Clements, William R.
• Granted Patent: US 8,010,321 B2
• Time in Patent Office: Days
• US Class Current: 714/45
• CPC Class Codes: G05B 19/0428 Safety, monitoring G05B19/0423 takes precedence
• CPC Class Codes: G05B 2219/37519 From machining parameters classify different fault cases

Abstract

A method and apparatus for diagnosing faults. Process data is analyzed using a first metric to identify a fault. The process data was obtained from a manufacturing machine running a first recipe. A fault signature that matches the fault is identified. The identified fault signature was generated using a second metric and/or a second recipe. At least one fault class that is associated with the fault signature is identified.

Claims

1. A method of diagnosing faults, comprising:

analyzing process data using a first metric to identify a fault, wherein the process data was obtained from a manufacturing machine running a first recipe;

identifying a fault signature that matches the fault, wherein the fault signature was generated using at least one of a second metric and a second recipe; and

identifying at least one fault class that is associated with the fault signature.

2. The method of claim 1, wherein the at least one fault class applies to two or more different metrics.

3. The method of claim 1, wherein the first metric is a combined multivariate index (CMI) metric, and the second metric is one of a T2 metric, and a Q metric.

4. The method of claim 1, wherein the at least one fault class applies to two or more different recipes.

5. A machine-accessible medium including data that, when accessed by a machine, cause the machine to perform a method comprising:

analyzing process data using a first metric to identify a fault, wherein the process data was obtained from a manufacturing machine running a first recipe;

identifying a fault signature that matches the fault, wherein the fault signature was generated using at least one of a second metric and a second recipe; and

identifying at least one fault class that is associated with the fault signature.

6. The machine-accessible medium of claim 5, wherein the at least one fault class applies to two or more different metrics.

7. The machine-accessible medium of claim 6, wherein the first metric is a combined multivariate index (CMI) metric, and the second metric is one of a T2 metric, and a Q metric.

8. The method of claim 5, wherein the at least one fault class applies to two or more different recipes.

9. A statistical process monitoring system, comprising:

a fault detector coupled with a manufacturing machine, the fault detector to receive process data produced using a first recipe from the manufacturing machine, and to analyze the process data using a first metric to identify a fault;

a database to store a plurality of fault signatures, each of the fault signatures being associated with at least one fault class; and

a fault diagnoser coupled with the fault detector and coupled with the database, the fault diagnoser to identify a matching fault signature that matches the fault from the plurality of fault signatures, wherein the matching fault signature was generated using at least one of a second metric and a second recipe with the manufacturing machine, and to identify at least one fault class that is associated with the fault signature.

10. The statistical process monitoring system of claim 9, wherein the at least one fault class applies to two or more different metrics.

11. The statistical process monitoring system of claim 10, wherein the first metric is a combined multivariate index (CMI) metric, and the second metric is one of a T2 metric, and a Q metric.

12. The statistical process monitoring system of claim 9, wherein the at least one fault class applies to two or more different recipes.

13. A method of diagnosing faults, comprising:

analyzing process data from at least one of historical data and training data to identify a plurality of faults;

determining one or more process variables that contributed to the plurality of faults;

determining a relative contribution of each of the one or more process variables;

generating a fault signature for each of the plurality of faults, the fault signature having relative contribution ranges that include the relative contributions of each of the one or more process variables;

dividing the plurality of faults into groups based on similarities in fault signatures; and

generating a fault class for each of the groups of faults.

14. The method of claim 13, further comprising:

determining that at least one fault is associated with a product that meets specification requirements; and

classifying a fault class generated for the at least one fault as a false alarm.

15. The method of claim 13, further comprising:

determining that at least one fault is associated with a product that does not meet specification requirements;

identifying a fault class generated for the at least one fault as a real fault; and

receiving user input to classify the generated fault class.

16. The method of claim 15, further comprising:

assigning a severity level to the fault class.

17. The method of claim 13, wherein the plurality of faults include at least two faults having similar relative contributions of at least some of the one or more process variables, further comprising:

generating a single fault signature for the at least two faults, the single fault signature having relative contribution ranges that include the relative contributions of the process variables of each of the at least two faults; and

generating a single fault class for the at least two faults.

18. The method of claim 13, wherein the fault signatures are generated automatically when a multivariate statistical model is generated.

19. A machine-accessible medium including data that, when accessed by a machine, cause the machine to perform a method comprising:

analyzing process data from at least one of historical data and training data to identify a plurality of faults;

determining one or more process variables that contributed to the plurality of faults;

determining a relative contribution of each of the one or more process variables;

generating a fault signature for each of the plurality of faults, the fault signature having relative contribution ranges that include the relative contributions of each of the one or more process variables;

dividing the plurality of faults into groups based on similarities in fault signatures; and

generating a fault class for each of the groups of faults.

20. The machine-accessible medium of claim 19, the method further comprising:

determining that at least one fault is associated with a product that meets specification requirements; and

classifying a fault class generated for the at least one fault as a false alarm.

21. The machine-accessible medium of claim 19, the method further comprising:

determining that at least one fault is associated with a product that does not meet specification requirements;

identifying a fault class generated for the at least one fault as a real fault; and

receiving user input to classify the generated fault class.

22. The machine-accessible medium of claim 21, the method further comprising:

assigning a severity level to the fault class.

23. The machine-accessible medium of claim 21, wherein the plurality of faults include at least two faults having similar relative contributions of at least some of the one or more process variables, further comprising:

generating a single fault signature for the at least two faults, the single fault signature having relative contribution ranges that include the relative contributions of the process variables of each of the at least two faults; and

generating a single fault class for the at least two faults.

24. The machine-accessible medium of claim 21, wherein the fault signatures are generated automatically when a multivariate statistical model is generated.

25. A statistical process monitoring system, comprising:

a database to store a at least one of historical process data and training process data;

a fault detector coupled with the database to analyze at least one of the historical process data and training process data to identify a plurality of faults; and

a fault diagnoser coupled with the fault detector to determine one or more process variables that contributed to the plurality of faults, to determine a relative contribution of each of the one or more process variables, to generate a fault signature for each of the plurality of faults, the fault signature having relative contribution ranges that include the relative contributions of each of the one or more process variables, to divide the plurality of faults into groups based on similarities in fault signatures, and to generate a fault class for each of the groups of faults.

26. The statistical process monitoring system of claim 25, wherein the fault diagnoser to determine that at least one fault is associated with a product that meets specification requirements, and to generate a false alarm fault class.

27. The statistical process monitoring system of claim 25, wherein the fault diagnoser to determine that at least one fault is associated with a product that does not meet specification requirements, to generate a real fault class, and to receive user input to classify the real fault class.

28. The statistical process monitoring system of claim 27, wherein the fault diagnoser to assign a severity level to the fault class.

29. The statistical process monitoring system of claim 25, wherein the plurality of faults include at least two faults having similar relative contributions of at least some of the one or more process variables, the fault diagnoser to generate a single fault signature and a single fault class for the at least two faults, the single fault signature having relative contribution ranges that include the relative contributions of the process variables of each of the at least two faults.