# US 2025/0156693 A1 · System and method for designing process factor

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0156693 A1
- Document ID: US20250156693A1
- Status: Published application
- Kind: A1
- Application number: 18/920,674
- Filed: 2024-10-18
- Published: 2025-05-15
- Earliest claimed: 2023-11-09
- Assignees: Samsung SDI Co., Ltd.; Seoul National University R&Db Foundation
- Applicants: Samsung SDI Co., Ltd.; Seoul National University R&DB Foundation
- Inventors: Jeemin Yu; Sungroh Yoon; Jaehee Jang; Seungryong Yoo; Jaehyoun Lee; Jee Hoon Han
- Web page: https://patentgrep.com/patent/US20250156693A1

## CPC

- G06N 3/0475: Generative networks
- G06N 20/20: Ensemble learning
- G06N 3/045: Combinations of networks
- G01R 31/3865: {related to manufacture, e.g. testing after manufacture}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)

## Abstract

A system for designing a process factor according to the present disclosure includes: a first generation device configured to generate a plurality of first process factors using a first generative model that estimates a decision boundary of a quality prediction model and generates the process factor based on the estimated decision boundary; a second generation device configured to generate a plurality of second process factors using a deep learning-based second generative model; and a determination device configured to determine a third generative model among the first generative model and the second generative model according to a result of evaluating the plurality of first process factors and the plurality of second process factors using the quality prediction model, and automatically generate at least one process factor applied to a production process using the third generative model. The quality prediction model is an artificial neural network model that receives a process factor of the production process as input data and predicts quality data based on the input data.

## Claims

### Claim 1

A system for designing a process factor, comprising:
a first generation device configured to generate a plurality of first process factors using a first generative model capable of estimating a decision boundary of a quality prediction model and generating the process factor based on the estimated decision boundary;
a second generation device configured to generate a plurality of second process factors using a deep learning-based second generative model; and
a determination device configured to determine a third generative model among the first generative model and the second generative model and automatically generate at least one process factor applied to a production process using the third generative model,
wherein the quality prediction model is an artificial neural network model that receives a process factor of the production process as input data and predicts quality data based on the input data, and
wherein the third generative model is a result of an evaluation of the plurality of first process factors and the plurality of second process factors using the quality prediction model.

### Claim 2 (depends on 1)

The system as claimed in claim 1, further comprising
a quality prediction model construction device configured to construct the quality prediction model by learning the artificial neural network model using a plurality of third process factors and a plurality of quality data collected from the production process as learning data.

### Claim 3 (depends on 1)

The system as claimed in claim 1, wherein the first generation device is further configured to estimate the decision boundary by applying an adversarial attack on the quality prediction model.

### Claim 4 (depends on 1)

The system as claimed in claim 1, wherein the second generation device is further configured to use the quality prediction model as a discriminative model of the second generative model to learn the second generative model.

### Claim 5 (depends on 1)

The system as claimed in claim 1, wherein
the determination device is further configured to:
input the plurality of first process factors and the plurality of second process factors as the input data of the quality prediction model;
analyze output data output from the quality prediction model to calculate a first ratio and a second ratio; and
determine the optimal generative model based on a comparison result of the first ratio and the second ratio, wherein:
the first ratio is a ratio of process factors satisfying a target quality among the plurality of first process factors, and
the second ratio is a ratio of process factors satisfying a target quality among the plurality of second process factors.

### Claim 6 (depends on 1)

The system as claimed in claim 1, wherein:
the production process is a production process of a battery;
the at least one process factor applied to the production process is a factor representing a design condition in the production process of the battery and comprises at least one of a charging voltage, a charging capacity, a charging time, a charging cell temperature, a discharging voltage, a discharging current, a discharging capacity, a discharging cell temperature, and an electrolyte weight; and
the quality data indicates whether a target quality for at least one quality factor indicating a quality of a battery is achieved.

### Claim 7

A method for designing a process factor of a system, comprising:
generating a plurality of first process factors using a first generative model that estimates a decision boundary of a quality prediction model and generates the process factor based on the estimated decision boundary;
generating a plurality of second process factors using a deep learning-based second generative model;
determining a third generative model among the first generative model and the second generative model according to an evaluation of the plurality of first process factors and the plurality of second process factors using the quality prediction model; and
automatically designing at least one process factor applied to a production process using the determined optimal generative model,
wherein the quality prediction model is an artificial neural network model that receives the process factor applied to the production process as input data and predicts quality data based on the input data.

### Claim 8 (depends on 7)

The method as claimed in claim 7, further comprising constructing the quality prediction model by learning the artificial neural network model using a plurality of third process factors and a plurality of quality data collected from the production process as learning data.

### Claim 9 (depends on 7)

The method as claimed in claim 7, wherein the generating of the plurality of first process factors comprises estimating the decision boundary by applying an adversarial attack on the quality prediction model.

### Claim 10 (depends on 7)

The method as claimed in claim 7, further comprising using the quality prediction model as a discriminative model of the second generative model to learn the second generative model.

### Claim 11 (depends on 7)

The method as claimed in claim 7, wherein the determining of the third generative model comprises:
inputting the plurality of first process factors and the plurality of second process factors as the input data of the quality prediction model;
analyzing output data output from the quality prediction model to calculate a first ratio and a second ratio; and
determining the optimal generative model based on a comparison result of the first ratio and the second ratio, wherein:
the first ratio is a ratio of process factors satisfying a target quality among the plurality of first process factors, and
the second ratio is a ratio of process factors satisfying a target quality among the plurality of second process factors.

### Claim 12 (depends on 7)

The method as claimed in claim 7, wherein:
the production process is a production process of a battery;
the at least one process factor applied to the production process is a factor representing a design condition in the production process of the battery and comprises at least one of a charging voltage, a charging capacity, a charging time, a charging cell temperature, a discharging voltage, a discharging current, a discharging capacity, a discharging cell temperature, and an electrolyte weight; and
the quality data indicates whether a target quality for at least one quality factor indicating a quality of a battery is achieved.

### Claim 13

A method for manufacturing a system for designing a process factor, comprising:
providing a first generation device that generates a plurality of first process factors using a first generative model, estimates a decision boundary of a quality prediction model, and generates the process factor based on the estimated decision boundary;
providing a second generation device that generates a plurality of second process factors using a deep learning-based second generative model;
providing a determination device that determines a third generative model among the first generative model and the second generative model and automatically generates at least one process factor applied to a production process using the third generative model, wherein:
the quality prediction model is an artificial neural network model that receives a process factor of the production process as input data and predicts quality data based on the input data, and
wherein the third generative model is a result of an evaluation of the plurality of first process factors and the plurality of second process factors using the quality prediction model.

### Claim 14 (depends on 13)

The method of claim 13, further comprising providing a quality prediction model construction device that constructs the quality prediction model by learning the artificial neural network model using a plurality of third process factors and a plurality of quality data collected from the production process as learning data.

### Claim 15 (depends on 13)

The method of claim 13, wherein the first generation device estimates the decision boundary by applying an adversarial attack on the quality prediction model.

### Claim 16 (depends on 13)

The method of claim 13, wherein the second generation device is further uses the quality prediction model as a discriminative model of the second generative model to learn the second generative model.

### Claim 17 (depends on 13)

The method of claim 13, wherein the determination device inputs the plurality of first process factors and the plurality of second process factors as the input data of the quality prediction model.

### Claim 18 (depends on 13)

The method of claim 13, wherein the determination device analyzes output data output from the quality prediction model to calculate a first ratio and a second ratio.

### Claim 19 (depends on 18)

The method of claim 18, wherein the first ratio is a ratio of process factors satisfying a target quality among the plurality of first process factors, and the second ratio is a ratio of process factors satisfying a target quality among the plurality of second process factors.

### Claim 20 (depends on 13)

The method of claim 13, wherein the at least one process factor applied to the production process at least one of a charging voltage, a charging capacity, a charging time, a charging cell temperature, a discharging voltage, a discharging current, a discharging capacity, a discharging cell temperature, and an electrolyte weight.
