Anomaly detection from aggregate statistics using neural networks

1. A method, comprising:

obtaining a reduced representation of a plurality of sensor statistics representative of data collected by a plurality of sensors associated with a device manufacturing system performing a manufacturing operation;

generating, using a plurality of outlier detection models, a plurality of outlier scores, wherein one or more of the plurality of outlier scores are representative of a degree of presence, in the plurality of sensor statistics, of an anomaly associated with the manufacturing operation, and wherein each of the plurality of outlier scores is generated based on the reduced representation of the plurality of sensor statistics using a respective one of the plurality of outlier detection models; and

processing the plurality of outlier scores using a detector neural network to generate an anomaly score indicative of a likelihood of the anomaly associated with the manufacturing operation.

2. The method of claim 1 , wherein the reduced representation of the plurality of sensor statistics is obtained by processing an initial representation of the plurality of sensor statistics using a reducer neural network, wherein the initial representation comprises a plurality of sets of statistical parameters, wherein each set of the plurality of sets of statistical parameters is obtained by pre-processing raw sensor statistics for a respective one of the plurality of sensors, and wherein the reduced representation has fewer parameters than the initial representation.

3. The method of claim 2 , wherein the initial representation is an aggregate representation of the plurality of sensor statistics.

4. The method of claim 2 , wherein the raw sensor statistics characterizes a plurality of measurements associated with the respective one of the plurality of sensors.

5. The method of claim 2 , wherein the reducer neural network is a feed-forward network.

6. The method of claim 2 , wherein pre-processing of the raw sensor statistics comprises adjusting the raw sensor statistics in view of at least some of one or more preventive maintenance events, one or more changes in settings of the manufacturing operation, or one or more changes in settings of the device manufacturing system.

7. The method of claim 2 , wherein at least some of the plurality of sets of statistical parameters comprise one or more of a mean, a median, a mode, a variance, a standard deviation, a range, a maximum, a minimum, a skewness, or a kurtosis for the pre-processed raw sensor statistics for the respective one of the plurality of sensors.

8. The method of claim 1 , wherein prior to processing the plurality of outlier scores, at least some of the plurality of outlier scores are normalized.

9. The method of claim 1 , further comprising performing an anomaly remediation action for the device manufacturing system in response to the anomaly score indicating a presence of an anomaly.

10. A system comprising:

a memory; and

a processing device operatively coupled to the memory, the processing device to:

obtain a reduced representation of a plurality of sensor statistics representative of data collected by a plurality of sensors associated with a device manufacturing system performing a manufacturing operation;

generate, using a plurality of outlier detection models, a plurality of outlier scores, wherein one or more of the plurality of outlier scores are representative of a degree of presence, in the plurality of sensor statistics, of an anomaly associated with the manufacturing operation, and wherein each of the plurality of outlier scores is generated based on the reduced representation of the plurality of sensor statistics using a respective one of the plurality of outlier detection models; and

process the plurality of outlier scores using a detector neural network to generate an anomaly score indicative of a likelihood of an anomaly associated with the manufacturing operation.

11. The system of claim 10 , wherein the reduced representation of the plurality of sensor statistics is obtained by processing an initial representation of the plurality of sensor statistics using a reducer neural network, wherein the initial representation comprises a plurality of sets of statistical parameters, wherein each set of the plurality of sets of statistical parameters is obtained by pre-processing raw sensor statistics for a respective one of the plurality of sensors, and wherein the reduced representation has fewer parameters than the initial representation.

12. The system of claim 11 , wherein the raw sensor statistics characterizes a plurality of measurements associated with the respective one of the plurality of sensors.

13. The system of claim 11 , wherein pre-processing of the raw sensor statistics comprises adjusting the raw sensor statistics in view of at least some of one or more preventive maintenance events, one or more changes in settings of the manufacturing operation, or one or more changes in settings of the device manufacturing system.

14. The system of claim 11 , wherein at least some of the plurality of sets of statistical parameters comprise one or more of a mean, a median, a mode, a variance, a standard deviation, a range, a maximum, a minimum, a skewness, or a kurtosis for the pre-processed raw sensor statistics for the respective one of the plurality of sensors.

15. The system of claim 10 , wherein the detector neural network is a reduced Boltzmann machine network.

16. The system of claim 10 , wherein prior to processing the plurality of outlier scores, at least some of the plurality of outlier scores are normalized.

17. A non-transitory computer readable storage medium comprising instructions that, when executed by at least one processing device, cause the at least one processing device to:

obtain a reduced representation of a plurality of sensor statistics representative of data collected by a plurality of sensors associated with a device manufacturing system performing a manufacturing operation;

generate, using a plurality of outlier detection models, a plurality of outlier scores, wherein each of the plurality of outlier scores is representative of a degree of presence, in the plurality of sensor statistics, of an anomaly associated with the manufacturing operation, and wherein one or more of the plurality of outlier scores are generated based on the reduced representation of the plurality of sensor statistics using a respective one of the plurality of outlier detection models; and

process the plurality of outlier scores using a detector neural network to generate an anomaly score indicative of a likelihood of an anomaly associated with the manufacturing operation.

18. The non-transitory computer readable medium of claim 17 , wherein the reduced representation of the plurality of sensor statistics is obtained by processing an initial representation of the plurality of sensor statistics using a reducer neural network, wherein the initial representation comprises a plurality of sets of statistical parameters, wherein each set of the plurality of sets of statistical parameters is obtained by pre-processing raw sensor statistics for a respective one of the plurality of sensors, and wherein the reduced representation has fewer parameters than the initial representation.

19. The non-transitory computer readable medium of claim 17 , wherein the instructions are further to cause the at least one processing device to perform an anomaly remediation action for the device manufacturing system in response to the anomaly score indicating a presence of an anomaly.

20. The non-transitory computer readable medium of claim 17 , wherein prior to processing the plurality of outlier scores, at least some of the plurality of outlier scores are normalized.