Method, apparatus, and system with abnormality determination

Patent metadata

Field	Value
Publication number	US-2024201673-A1
Application number	US-202318330589-A
Country	US
Kind code	A1
Filing date	Jun 7, 2023
Priority date	Dec 19, 2022
Publication date	Jun 20, 2024
Grant date	—

Abstract

A processor-implemented method of an apparatus or system includes obtaining an expert classification criterion from a memory of the apparatus or system; converting manufacturing process data associated with a manufacturing process to a test sample in a form of an image; generating, using a machine learning model provided the test sample, a probability value that the test sample corresponds to a target class representing an anomaly occurring in the manufacturing process; adjusting the probability value by reflecting the expert classification criterion for the anomaly; and identifying, by classifying the anomaly based on the adjusted probability value, whether a final abnormality in the manufacturing process has occurred.

First claim

What is claimed is:

1 . A processor-implemented method of an apparatus or system, comprising:
obtaining an expert classification criterion from a memory of the apparatus or system; converting manufacturing process data associated with a manufacturing process to a test sample in a form of an image; generating, using a machine learning model provided the test sample, a probability value that the test sample corresponds to a target class representing an anomaly occurring in the manufacturing process; adjusting the probability value by reflecting the expert classification criterion for the anomaly; and identifying, by classifying the anomaly based on the adjusted probability value, whether a final abnormality in the manufacturing process has occurred.

2 . The method of claim 1 , wherein the adjusting of the probability value comprises adjusting the probability value based on respective distances in an embedded space between the test sample and hard samples classified according to the expert classification criterion.

3 . The method of claim 2 , wherein the hard samples classified according to the expert classification criterion are stored in advance.

4 . The method of claim 2 , wherein the adjusting of the probability value comprises:
mapping the hard samples and the test sample to the embedded space using a mapping function; and adjusting the probability value based on respective distances between first positions corresponding to the respective hard samples in the embedded space and a second position corresponding to the test sample.

5 . The method of claim 4 , wherein the adjusting of the probability value comprises:
adjusting a corresponding probability value such that the corresponding probability value increases, in response to a corresponding one of the distances being less than or equal to a threshold; and adjusting the corresponding probability value such that the corresponding probability value decreases, in response to the corresponding one of the distances being greater than the threshold.

6 . The method of claim 5 , wherein the threshold is adjustable according to a difficulty in classifying the hard samples by the expert.

7 . The method of claim 1 , wherein the machine learning model is trained to predict the anomaly based on the manufacturing process data labeled with the target class.

8 . The method of claim 1 , wherein the machine learning model is trained based on at least one of:
a first loss based on a distance between a training probability value corresponding to a training test sample in an embedded space and a set threshold; or a second loss based on cross entropy between the target class representing a training anomaly predicted by the in-training machine learning model and a ground truth class corresponding to the test sample.

9 . The method of claim 1 , wherein the manufacturing process data comprises at least one of:
sensing data, representing at least one of an operating state of manufacturing equipment, a phenomenon represented by the manufacturing equipment, or an output of the manufacturing equipment and which is generated in a process in which the manufacturing equipment performs the manufacturing process in time series; or virtual metrology data representing the manufacturing process.

10 . The method of claim 1 , further comprising:
storing the classified anomaly of the manufacturing process data in the memory.

11 . The method of claim 1 , wherein the machine learning model comprises at least one of a support vector machine (SVM) or a neural network classifier.

12 . The method of claim 1 , wherein the manufacturing process data comprises semiconductor manufacturing process data from one or more of:
a thin film process of depositing a thin film, on which an electric circuit of a semiconductor device is to be printed, on a surface of a wafer; an implantation process of implanting a dopant into the deposited thin film; a diffusion process of applying heat to the implanted dopant such that the dopant is distributed on the thin film; a lithography process of coating a surface of the thin film with a photoresist to transfer a circuit pattern prefabricated on a reticle or mask to the surface of the thin film, performing an exposure to the photoresist applied onto the surface of the thin film, removing the exposed photoresist, and performing developing; an etching process of removing an exposed film using the photoresist remaining after the removing as a protective film; a cleaning process of removing the remaining photoresist and removing additionally formed foreign materials; a deposition process of forming a thick film to protect and insulate a circuit pattern of the electrical circuit; a planarization process of planarizing the deposited thick film; a metrology process of measuring at least one of a size of the circuit pattern, a thickness of the circuit pattern, or a dopant concentration; or an inspection process of inspecting particles and a pattern defect caused by a fault.

13 . A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform the method of claim 1 .

14 . A processor-implemented method of an apparatus or system, comprising:
obtaining an expert classification criterion from a memory of the apparatus or system; converting manufacturing process data associated with a manufacturing process to a test sample in a form of an image; generating, using a machine learning model provided the test sample and based on the expert classification criterion for an anomaly in the manufacturing process, a probability value that the test sample corresponds to a target class representing the anomaly; and identifying, by classifying the anomaly based on the generated probability value, whether a final abnormality in the manufacturing process has occurred.

15 . The method of claim 14 , wherein the machine learning model is an in-training model, and the method further comprises:
training the in-training model based on:
a first loss based on a distance between a probability value corresponding to the test sample in an embedded space and a set threshold;
a second loss based on cross entropy between the target class representing the anomaly predicted by the machine learning model and a ground truth class corresponding to the test sample; and
a third loss to force a first distance between the test sample and a hard negative sample having a class different from the target class corresponding to the test sample among samples included in the embedded space to be less than a second distance between the test sample and a random negative sample corresponding to one of hard samples classified according to the classification criterion of the expert.

16 . An apparatus, comprising:
one or more processors configured to: convert manufacturing process data to a test sample in a form of an image; generate, using a machine learning model provided the test sample, a probability value that the test sample corresponds to a target class representing an anomaly occurring in the manufacturing process; adjust the probability value based on a classification criterion of an expert for the anomaly; and identify, by classifying the anomaly based on the adjusted probability value, whether the abnormality in the manufacturing process has occurred.

17 . The apparatus of claim 16 , wherein the one or more processors are configured to adjust the probability value based on respective distances in an embedded space

Assignees

• Samsung Electronics Co Ltd

Inventors

• Kim Keunseo
• Kim Gyungmin
• Park Sangdo

Classifications

• H10P74/203 Primary

  Structural properties, e.g. testing or measuring thicknesses, line widths, warpage, bond strengths or physical defects · CPC title

• H10P74/23

  characterised by multiple measurements, corrections, marking or sorting processes · CPC title

• Y02P90/02

  Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS] · CPC title

• G05B19/41875 Primary

  characterised by quality surveillance of production · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title