US 12,271,830 B2
Learned model generating method, abnormality factor estimating device, substrate processing device, abnormality factor estimating method, learning method, learning device, and learning data generating method
Asuka Wakita, Kyoto (JP); Naoki Sawazaki, Kyoto (JP); and Takashi Ota, Kyoto (JP)
Assigned to SCREEN Holdings Co., Ltd., Kyoto (JP)
Filed by SCREEN Holdings Co., Ltd., Kyoto (JP)
Filed on Feb. 26, 2021, as Appl. No. 17/185,981.
Claims priority of application No. 2020-052983 (JP), filed on Mar. 24, 2020.
Prior Publication US 2021/0304032 A1, Sep. 30, 2021
Int. Cl. G06N 5/04 (2023.01); G06N 20/00 (2019.01); H01L 21/67 (2006.01)
CPC G06N 5/04 (2013.01) [G06N 20/00 (2019.01); H01L 21/67253 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A learned model generating method comprising:
acquiring learning data; and
generating a learned model for estimating a factor of an abnormality of a processing target substrate after processing using a processing fluid by performing machine learning of the learning data,
wherein the learning data comprises a feature quantity and abnormality factor information,
wherein the abnormality factor information represents a factor of an abnormality of a learning target substrate after processing using the processing fluid,
wherein the feature quantity comprises first feature quantity information representing a feature of a time transition of section data in time series data representing a physical quantity of an object used by a substrate processing device that processes the learning target substrate using the processing fluid,
wherein the first feature quantity information is represented using times,
wherein the first feature quantity information comprises at least one of first information, second information, third information, fourth information, fifth information, and sixth information,
wherein the first information is information representing a state of the physical quantity when the physical quantity increases toward a target value,
wherein the second information is information representing an overshoot of the physical quantity,
wherein the third information is information representing variations of the physical quantity,
wherein the fourth information is information representing the state of the physical quantity when the physical quantity decreases from the target value,
wherein, when the substrate processing device uses at least two different objects, the fifth information is information representing an overlap between the physical quantity of one object out of the two objects and the physical quantity of the other object, and
wherein the sixth information is information representing a time interval on a time axis between the one object and the other object.