Individual machine configuration based on overall process performance and latent metrics
Machine-level metrics for machines of a processing line, as well as latent metrics for results of the processing line, are obtained for the current time period. Machine configuration setting values for the machines during the current time period, as well as at the earlier time period are also obtained. Updates to current configuration setting values of the machines are determined based upon analysis of the machine-level metrics, the process-level metrics for the current time period, the latent metrics for the earlier time period, and the machine configuration setting values for the current and earlier time periods (e.g., the metrics and configuration values may be input to one or more machine learning models used to determine the updates to current configuration setting values for the individual machines). The current values may be changed to the updated values via an automated process or the updates presented as a recommendation report.
1. A system, comprising:
one or more storage systems comprising:
a metrics repository for storing individual non-latent metrics including machine-level metrics data for a plurality of machines used together to implement a process that produces physical products and process-level metrics data for the process as-a-whole;
a configuration data repository for storing historical machine configuration setting values for the plurality of machines;
one or more computers comprising respective processors and memory and configured to:
obtain, during a current time period after a production time period during which processing of a physical product is performed and completed, latent metrics data for the completed physical product produced by the processing, wherein the obtained latent metrics data were unavailable during the production time period of the physical product through to immediately subsequent to the production time period and, if the physical product was tested, unavailable through testing of the physical product during and immediately subsequent to the production time period;
access the configuration data repository to determine previous machine configuration setting values for the plurality of machines during the production time period during which the physical product was processed;
provide the machine-level metrics data and the process-level metrics data for the current time period, the latent metrics data for the physical product produced by the process, and the machine configuration setting values during the production time period to one or more machine learning models to determine updated configuration setting values for the plurality of machines; and
transmit the updated configuration setting values, determined via the one or more machine learning models based at least in part on the machine-level metrics data, the process-level metrics data for the current time period, the latent metrics data for the physical product produced by the process during the production time period, and the machine configuration setting values during the production time period, to enable the plurality of machines to implement the updated configuration setting values.
2. The system of claim 1 , wherein to transmit the updated configuration setting values, the one or more computers are further configured to transmit a recommendation report for the updates to the one or more specified destinations.
3. The system of claim 1 , wherein the one or more computers are further configured to:
access an acceptance history of previous updates to machine configuration setting values determined by the one or more machine learning models;
determine whether the acceptance history meets an acceptance threshold; and
wherein said transmit the updated configuration setting values is performed responsive to the acceptance history meeting the acceptance threshold and causes configuration settings for one or more of the plurality of machines to be updated automatically.
4. The system of claim 1 , wherein the one or more computers are further configured to:
store updates to configuration settings in the configuration data repository, wherein the updates include an indication of a time associated with the updates to the configuration settings.
5. A method, comprising:
performing, by one or more computing devices:
receiving, for a current time period, non-latent metrics including machine-level metrics data for individual ones of a plurality of machines used together to implement a process and process-level metrics data for the process as-a-whole;
receiving one or more latent metrics data for one or more physical products completed by the process during an earlier time prior to the current time period that is subsequent to completion of processing of the physical products, wherein the one or more latent metrics data for the one or more physical products were unavailable during the processing during the earlier period through to immediately subsequent to the processing during the earlier period and, if the physical product was tested, unavailable through testing of the physical product during and immediately subsequent to the processing during the earlier period;
receiving machine configuration setting values for the plurality of machines during the current time period and at the earlier time period;
determining, based on one or more machine learning models, one or more updates to current configuration setting values for the plurality of machines, wherein input to the one or more machine learning models comprises the machine-level metrics data and the process-level metrics data for the current time period, the latent metrics data for the earlier time period, and the machine configuration setting values for the current and earlier time periods; and
transmitting the updates to the current configuration setting values to one or more specified destinations.
6. The method of claim 5 , wherein the machine-level metrics comprise one or more of:
an output rate of one or more of the machines,
an output quality for one or more of the machines,
an operating condition for one or more of the machines, or
an environmental condition at one or more of the machines.
7. The method of claim 5 , wherein the process-level metrics comprise one or more of:
an output rate for the process as-a-whole,
an output quality for the process as-a-whole, or
an environmental condition for the process as-a-whole.
8. The method of claim 5 , further comprising:
receiving, at a data stream storage service, metrics data from a plurality of sensors at respective ones of the plurality of machines or the process; and
parsing the data stream to load the machine-level and process-level metrics data into a metrics repository.
9. The method of claim 8 , wherein the metrics data are received via one or more network connections to the plurality of sensors.
10. The method of claim 8 , wherein the metrics data are received over an Internet connection to a hub device that collects the metrics data from the plurality of sensors.
11. The method of claim 5 , wherein the one or more latent metrics data are received from a source remote from the plurality of machines and process.
12. The method of claim 5 , wherein the one or more latent metrics data describe one or more latent qualities or defects of the one or more physical products of the process.
13. The method of claim 5 , further comprising:
repeating said receiving machine-level metrics data, said receiving one or more latent metrics data, said receiving machine configuration setting values, said determining, and said transmitting for one or more subsequent time periods.
14. The method of claim 5 , further comprising:
varying at least some of the configuration setting values for a subsequent time period;
receiving metrics data for the subsequent time period when the configuration setting values have been varied; and
inputting the received metrics data into the one or more machine learning models to search for improvements to the values of the configuration settings.
15. The method of claim 5 , further comprising:
performing, as a service for individual ones of a plurality of different processes employing different pluralities of machines, each of said receiving machine-level metrics data, said receiving one or more latent metrics data, said receiving machine configuration setting values, said determining, and said transmitting.
16. The method of claim 5 , further comprising:
updating the one or more machine learning models based on previous machine configuration setting values and the latent metrics data to produce one or more updated machine learning models;
determining, based on the updated one or more machine learning models, one or more other updated configuration setting values; and
transmitting the other updated configuration setting values to the one or more specified destinations.
17. One or more non-transitory computer-readable storage media storing program instructions that when executed on or across one or more processors cause the one or more processors to implement a configuration analyzer, configured to perform:
obtaining, for a current time period, non-latent metrics including machine-level metrics data for individual ones of a plurality of machines used together to implement a process and process-level metrics data for the complete process as-a-whole that produces one or more physical products;
obtaining one or more latent metrics data for the one or more physical products of the completed process produced at an earlier time prior to the current time period that is subsequent to completion of the process, wherein the one or more latent metrics data for the one or more physical products of the completed process as-a-whole were unavailable during the processing during the earlier period through to immediately subsequent to the processing during the earlier period and, if the physical product was tested, unavailable through testing of the physical product during and immediately subsequent to the processing during the earlier period;
obtaining machine configuration setting values for the plurality of machines during the current time period and at the earlier time period;
determining, based on one or more machine learning models, one or more updates to current configuration setting values for the plurality of machines, wherein input to the one or more machine learning models comprises the machine-level metrics data and the process-level metrics data for the current time period, the latent metrics data for the earlier time period, and the machine configuration setting values for the current and earlier time periods; and
transmitting the updates to the current configuration setting values to one or more specified destinations.
18. The one or more non-transitory computer-readable media of claim 17 , wherein the program instructions cause the one or more processors to perform:
repeating said obtaining machine-level metrics data, said obtaining one or more latent metrics data, said obtaining machine configuration setting values, said determining, and said transmitting for one or more subsequent time periods.
19. The one or more non-transitory computer-readable media of claim 17 , wherein the program instructions cause the one or more processors to perform:
varying at least some of the configuration setting values for a subsequent time period;
obtaining metrics data for the subsequent time period when the configuration setting values have been varied; and
inputting the obtained metrics data into the one or more machine learning models to search for improvements to the values of the configuration settings.
20. The one or more non-transitory computer-readable media of claim 17 , wherein the program instructions cause the one or more processors to perform:
storing updates to the configuration setting values for one or more subsequent time periods;
receiving metrics data for the one or more subsequent time periods;
determining, based at least in part on the received configuration setting values, to retrain the one or more machine learning models;
retraining the one or more machine learning models;
determining, based on one or more retrained machine learning models, one or more new updates to current configuration setting values for the plurality of machines; and
transmitting the new updates to the current configuration setting values to the one or more specified destinations.
21. The one or more non-transitory computer-readable media of claim 17 , wherein the program instructions cause the one or more processors to perform:
storing updates to the configuration settings in a configuration data repository, wherein the updates include an indication of a time associated with the updates to the configuration settings.
22. The one or more non-transitory computer-readable media of claim 17 , wherein to transmit the updates to the current configuration setting values the program instructions cause the one or more processors to perform transmitting a recommendation report for the updates to the one or more specified destinations.