Machine learning platform for substrate processing
A method includes identifying at least one of historical data associated with historical substrate lots processed by substrate processing tools in a substrate processing facility or simulated data for simulated substrate lots processed by simulated substrate processing tools. The method further includes generating features from the at least one of the historical data for the historical substrate lots or the simulated data for the simulated substrate lots. The method further includes training a machine learning model with data input comprising the features to generate a trained machine learning model. The trained machine learning model is capable of generating one or more outputs indicative of one or more corrective actions to be performed in the substrate processing facility.
1 . A method of operating a substrate processing facility, the method comprising:
identifying historical data associated with historical substrate lots associated with historical sets of substrates processed by substrate processing tools in the substrate processing facility, wherein each historical substrate lot is a corresponding set of historical substrates;
generating, based on perturbations of facility data, simulated data for simulated substrate lots associated with simulated sets of substrates processed by simulated substrate processing tools, wherein each simulated substrate lot is a simulation of a corresponding set of substrates, and wherein the perturbations comprise one or more of moving forward or backward lots in route;
generating features from the historical data for the historical substrate lots and the simulated data for the simulated substrate lots;
training a machine learning model with data input comprising the features to generate a trained machine learning model, wherein the trained machine learning model is trained to generate corresponding output indicative of one or more corresponding corrective actions to be performed in the substrate processing facility;
providing, as input to the machine learning model, current features of the substrate processing facility, wherein the current features are associated with a first substrate yield; and
causing, based on output obtained from the trained machine learning model, performance of one or more current corrective actions associated with processing of corresponding sets of current substrates of current substrate lots in the substrate processing facility to produce a second substrate yield that is higher than the first substrate yield.
2 . The method of claim 1 , wherein the one or more current corrective actions comprise one or more of substrate lot scheduling, substrate lot dispatching, or substrate lot planning.
3 . The method of claim 1 , the corresponding set of historical substrates being processed on a same substrate processing tool with a same process.
4 . The method of claim 1 further comprising storing the features in a data store to be reused for training additional machine learning models without regenerating the features.
5 . The method of claim 1 further comprising receiving user input specifying one or more types of features, wherein the generating of the features is based on the user input.
6 . The method of claim 1 further comprising:
receiving user input specifying a first type of target output; and
generating target output based on the first type of target output and the at least one of the historical data or the simulated data, wherein the training of the machine learning model is further based on the target output.
7 . The method of claim 1 further comprising:
identifying predetermined types of features, wherein the generating of the features is based on the predetermined types of features; and
receiving user selection of the machine learning model from a plurality of predetermined machine learning models associated with predetermined types of target output.
8 . A method of operating a substrate processing facility, the method comprising:
identifying current data associated with current substrate lots of the substrate processing facility;
identifying current features associated with the current data;
providing the current features to a trained machine learning model that was trained based on historical features, wherein the current features are associated with a first substrate yield, the historical features being generated based on historical data for historical substrate lots associated with historical sets of substrates processed by substrate processing tools in the substrate processing facility and simulated data for simulated substrate lots associated with simulated sets of substrates processed by simulated substrate processing tools, wherein each historical substrate lot is a corresponding set of historical substrates, wherein each simulated substrate lot is a simulation of a corresponding set of substrates, the simulated data being generated based on perturbations of facility data, and wherein the perturbations comprise moving forward or backward lots in route, wherein the trained machine learning model is trained to generate corresponding output indicative of one or more corresponding corrective actions to be performed in the substrate processing facility;
obtaining, from the trained machine learning model, an output; and
causing, based on the output of the trained machine learning model, performance of one or more corrective actions associated with processing of corresponding sets of current substrates of the current substrate lots in the substrate processing facility to produce a second substrate yield that is higher than the first substrate yield.
9 . The method of claim 8 , wherein the one or more corrective actions comprise one or more of substrate lot scheduling, substrate lot dispatching, or substrate lot planning.
10 . The method of claim 8 , the corresponding set of historical substrates being processed on a same substrate processing tool with a same process.
11 . The method of claim 8 further comprising retrieving the current features from a data store without regenerating the current features.
12 . The method of claim 8 further comprising receiving user input specifying one or more types of features and generating one or more of the current features based on the user input.
13 . The method of claim 8 further comprising receiving user input specifying a first type of target output, wherein the trained machine learning model is trained based on the first type of target output and the at least one of the historical data or the simulated data.
14 . The method of claim 8 further comprising:
identifying predetermined types of features, wherein the identifying of the current features is based on the predetermined types of features; and
receiving user selection of the trained machine learning model from a plurality of predetermined trained machine learning models associated with predetermined types of target output.
15 . A non-transitory machine-readable storage medium storing instructions which, when executed cause a processing device to perform operations comprising:
identifying historical data associated with historical substrate lots associated with historical sets of substrates processed by substrate processing tools in a substrate processing facility, wherein each historical substrate lot is a corresponding set of historical substrates;
generating, based on perturbations of facility data, simulated data for simulated substrate lots associated with simulated sets of substrates processed by simulated substrate processing tools, wherein each simulated substrate lot is a simulation of a corresponding set of substrates, wherein the perturbations comprise moving forward or backward lots in route;
generating features from the historical data for the historical substrate lots and the simulated data for the simulated substrate lots;
training a machine learning model with data input comprising the features to generate a trained machine learning model, wherein the trained machine learning model is trained to generate corresponding output indicative of one or more corresponding corrective actions to be performed in the substrate processing facility;
providing, as input to the machine learning model, current features of the substrate processing facility, wherein the current features are associated with a first substrate yield; and
causing, based on output obtained from the trained machine learning model, performance of one or more current corrective actions associated with processing of corresponding sets of current substrates of current substrate lots in the substrate processing facility to produce a second substrate yield that is higher than the first substrate yield.
16 . The non-transitory machine-readable storage medium of claim 15 , wherein the one or more current corrective actions comprise one or more of substrate lot scheduling, substrate lot dispatching, or substrate lot planning.
17 . The non-transitory machine-readable storage medium of claim 15 , the corresponding set of historical substrates being processed on a same substrate processing tool with a same process.
18 . The non-transitory machine-readable storage medium of claim 15 , wherein the operations further comprise storing the features in a data store to be reused for training additional machine learning models without regenerating the features.
19 . The non-transitory machine-readable storage medium of claim 15 , wherein the operations further comprise:
receiving user input specifying a first type of target output; and
generating target output based on the first type of target output and the at least one of the historical data or the simulated data, wherein the training of the machine learning model is further based on the target output.
20 . The non-transitory machine-readable storage medium of claim 15 , wherein the operations further comprise:
identifying predetermined types of features, wherein the generating of the features is based on the predetermined types of features; and
receiving user selection of the machine learning model from a plurality of predetermined machine learning models associated with predetermined types of target output.