Apparatus for cost-effective conversion of unsupervised fault detection (FD) system to supervised FD system

1. A method, comprising:

monitoring a plurality of runs of a recipe for fabricating one or more semiconductor devices within a manufacturing environment to collect runtime data from a plurality of sensors within the manufacturing environment;

creating a sensor peer group comprising two or more of the plurality of sensors, wherein the values generated by the two or more sensors are correlated, and wherein the two or more sensors have a predefined relationship with one another;

recording, for each sensor peer group value within the collected runtime data, steps of the recipe during which the sensor peer group value was collected;

determining metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe;

characterizing each of the plurality of runs into a respective one of a plurality of groups, based on the metrology data;

generating a data model, based on the characterization of the plurality of runs, that describes, for each of the plurality of groups, (i) a guard band of sensor peer group data for each respective group and (ii) a relative importance of the sensor peer group in each respective group;

performing a multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into a first group of the plurality of groups, by determining which guard band of sensor peer group data specified within the data model best fits the additional runtime data;

wherein classifying the at least one subsequent run of the recipe within the manufacturing environment into the first group is further based on the recorded steps for the collected runtime data and an indication of an additional step during which sensor peer group values in the additional runtime data were collected; and

upon classifying the at least one subsequent run of the recipe into the first group, outputting for display a report depicting a ranking of sensor peer group based on the additional runtime data and the relative importance of the sensor peer group for the first group within the data model.

2. The method of claim 1 , further comprising:

upon classifying the at least one subsequent run of the recipe into the first group:

determining one or more maintenance operations that have historically influenced the classification of runs of the recipe within the manufacturing environment from the first group into a second group; and

performing the one or more maintenance operations within the manufacturing environment, without requiring user interaction, wherein the one or more maintenance operations include at least (i) an auto-calibration operation for one or more pieces of equipment within the manufacturing environment and (ii) initiating a run of a maintenance recipe, distinct from the recipe, within the manufacturing environment.

3. The method of claim 1 , wherein determining the metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe further comprises:

determining an amount of time needed to complete each of the plurality of runs of the recipe; and

analyzing each of the semiconductor devices produced by the plurality of runs of the recipe to determine respective yield and metrology information for the respective semiconductor device, and

wherein characterizing each of the plurality of runs into a respective one of the plurality of groups is further based on the determined amount of time and the analysis of the one or more semiconductor devices fabricated by the respective run.

4. The method of claim 1 , wherein the data model further describes, for each of the plurality of groups, a respective weight for each step of the recipe describing a relative importance of the step in indicating the respective group, and wherein outputting for display the report depicting the ranking of the sensor peer group is further based on the described relative importance of each step of the recipe.

5. The method of claim 1 , wherein the data model further specifies, for one of the plurality of groups, at least one guard band of sensor peer group data, and wherein performing the multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into the first group further comprises determining that the at least one guard band of sensor data for the sensor peer group best fits the additional runtime data.

6. The method of claim 1 , wherein the predefined relationship comprises a physical proximity between the two or more sensors within the peer group within the manufacturing environment.

7. A system, comprising:

one or more computer processors; and

a memory containing computer program code that, when executed by operation of the one or more computer processors, performs an operation comprising:

monitoring a plurality of runs of a recipe for fabricating one or more semiconductor devices within a manufacturing environment to collect runtime data from a plurality of sensors within the manufacturing environment;

creating a sensor peer group comprising two or more of the plurality of sensors, wherein the values generated by the two or more sensors are correlated, and wherein the two or more sensors have a predefined relationship with one another;

recording, for each sensor peer group value within the collected runtime data, a plurality of steps of the recipe during which the sensor peer group value was collected;

determining metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe;

characterizing each of the plurality of runs into a respective one of a plurality of groups, based on an analysis of the metrology data;

generating a data model, based on the characterization of the plurality of runs, that describes, for each of the plurality of groups, (i) a guard band of sensor peer group data for each respective group and (ii) a relative importance of the sensor peer group in each respective group;

performing a multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into a first group, by determining which guard band of sensor peer group data specified within the data model best fits the additional runtime data;

wherein classifying the at least one subsequent run of the recipe within the manufacturing environment into the first group is further based on the recorded steps for the collected runtime data and an indication of an additional plurality of steps during which sensor peer group values in the additional runtime data were collected; and

upon classifying the at least one subsequent run of the recipe into the first group, outputting for display an interface depicting a ranking of the sensor peer group based on the additional runtime data and the relative importance of the sensor peer group for the first group within the data model.

8. The system of claim 7 , wherein determining metrology data characterizing each of the plurality of runs into the respective one of the plurality of groups further comprises:

receiving, by way of a user interface, the metrology data characterizing each of the plurality of runs into the respective group, wherein the plurality of groups include at least a group of good runs of the recipe and a group of faulty runs of the recipe.

9. The system of claim 7 , wherein determining the metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe further comprises:

determining an amount of time needed to complete each of the plurality of runs of the recipe; and

analyzing each of the semiconductor devices produced by the plurality of runs of the recipe to determine respective yield and metrology information for the respective semiconductor device, and

wherein characterizing each of the plurality of runs into a respective one of the plurality of groups is further based on the determined amount of time and the analysis of the one or more semiconductor devices fabricated by the respective run.

10. The system of claim 7 , wherein the generated data model further describes, for each of the plurality of groups, a respective weight for each of the plurality of steps of the recipe describing a relative importance of the step in indicating the respective group, and wherein outputting for display the interface depicting the ranking of the sensor peer group is further based on the described relative importance of each step of the recipe.

11. The system of claim 7 , wherein the data model further specifies, for one of the plurality of groups, at least one guard band of sensor peer group data, and wherein performing the multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into the first group further comprises determining that the at least one guard band of sensor peer group best fits the additional runtime data.

12. A non-transitory computer-readable medium containing computer program code that, when executed by operation of one or more computer processors, performs an operation comprising:

monitoring a plurality of runs of a recipe for fabricating one or more semiconductor devices within a manufacturing environment to collect runtime data from a plurality of sensors within a manufacturing environment;

creating a sensor peer group comprising two or more of the plurality of sensors, wherein the values generated by the two or more sensors are correlated;

recording, for each sensor peer group value within the collected runtime data, a step of the recipe during which the sensor peer group value was collected;

determining metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe;

characterizing each of the plurality of runs into a respective one of a plurality of groups, based on an analysis of the metrology data;

generating a data model, based on the characterization of the plurality of runs, that describes, for each of the plurality of groups, (i) a guard bands band of sensor peer group data for each respective group and (ii) a relative importance of the sensor peer group in each respective group;

performing a multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into a first group, by determining which guard bands of sensor peer group data specified within the data model best fits the additional runtime data;

wherein classifying the at least one subsequent run of the recipe within the manufacturing environment into the first group is further based on the recorded step for the collected runtime data and an indication of an additional step during which sensor values in the additional runtime data were collected; and

upon classifying the at least one subsequent run of the recipe into the first group, outputting for display an interface depicting a ranking the sensor peer group based on the additional runtime data and the relative importance of the sensor peer group for the first group within the data model.

13. The non-transitory computer-readable medium of claim 12 , wherein determining metrology data characterizing each of the plurality of runs into the respective one of the plurality of groups further comprises:

receiving, by way of a user interface, the metrology data characterizing each of the plurality of runs into the respective group, wherein the plurality of groups include at least a group of good runs of the recipe and a group of faulty runs of the recipe.

14. The non-transitory computer-readable medium of claim 12 , wherein determining the metrology data describing each of the semiconductor devices produced by the plurality of runs of the recipe further comprises:

determining an amount of time needed to complete each of the plurality of runs of the recipe; and

analyzing each of the semiconductor devices produced by the plurality of runs of the recipe to determine respective yield and metrology information for the respective semiconductor device, and

wherein characterizing each of the plurality of runs into a respective one of the plurality of groups is further based on the determined amount of time and the analysis of the one or more semiconductor devices fabricated by the respective run.

15. The non-transitory computer-readable medium of claim 12 ,

wherein the generated data model further describes, for each of the plurality of groups, a respective weight for the step of the recipe describing a relative importance of the step in indicating the respective group, and wherein outputting for display the interface depicting the ranking of the sensor peer group is further based on the described relative importance of each step of the recipe.

16. The non-transitory computer-readable medium of claim 12 , wherein the data model further specifies, for one of the plurality of groups, at least one guard band of sensor peer group data, wherein performing the multivariate analysis of additional runtime data collected during at least one subsequent run of the recipe within the manufacturing environment to classify the at least one subsequent run into the first group further comprises determining that the at least one guard band of sensor peer group data.