Process Recipe Recommendation Using Expert Preference Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Optimizing target processes like thin film deposition is hindered by the limitations of data collection, requiring significant time, expense, and manpower, and human expert-based optimization struggles with detailed numerical optimization and identifying nonlinear correlations.

Innovation Solution

A method involving a pre-trained process result predictor and preference predictor is used to select candidate recipes, collect expert preferences, and recommend optimal recipes based on predicted physical properties and preference scores, automating the search process to improve efficiency and accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If human experts manually optimize process conditions through experimentation, then generalization through overall trend analysis is improved, but detailed numerical optimization and identification of nonlinear correlations become difficult

Engineering Contradiction:

Improvegeneralization capabilityVSAvoidnumerical optimization precision

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The optimization system is segmented into two specialized components: a process result predictor for numerical optimization and a preference predictor for capturing expert preferences. This segmentation allows each component to specialize in one aspect (numerical precision or preference generalization) while working together to achieve both goals simultaneously.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The preference predictor acts as an intermediary that translates qualitative expert preferences into quantitative signals that can guide the optimization process. This intermediary bridges the gap between human expert intuition and machine-based numerical optimization, enabling both precision and adaptability.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If extensive experimentation is conducted to collect process data, then comprehensive data for optimization is obtained, but significant time, expense, and manpower are required

Engineering Contradiction:

Improvedata quantityVSAvoidoptimization time

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The process result predictor is pre-trained on existing process data before the optimization process begins. This preliminary action allows the system to leverage historical data without requiring extensive new experimentation, significantly reducing the time and resources needed for data collection during the actual optimization process.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements a feedback mechanism where the preference predictor learns from expert preferences on candidate recipes and continuously improves its predictions. This feedback loop allows the system to refine its optimization recommendations without requiring proportional increases in experimental data collection, reducing both time and resource requirements.

Inventive Principle:

Principle #23Feedback

3Adaptability or versatility

If the search interval for process conditions is widened to cover all possibilities, then comprehensive coverage is achieved, but the actual physically-meaningful interval becomes difficult to identify

Engineering Contradiction:

Improvesearch coverageVSAvoidphysically-meaningful interval identification

Core Design Contradiction:

Adaptability or versatilityVSDifficulty of detecting and measuring

Solution Approach 1:

The system dynamically adjusts the search interval parameters based on the target physical properties and expert preferences. The preference predictor learns the physically-meaningful intervals from expert feedback, allowing the system to automatically adapt the search scope without manual specification, resolving the difficulty of identifying meaningful intervals while maintaining comprehensive coverage.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250307710A1Method of recommending process recipe and training method of process result predictor

Publication Date: 2025.10.02 SAMSUNG ELECTRONICS CO LTD

AI summary

A method of recommending a process recipe is provided. The method includes selecting candidate process recipes for material synthesis corresponding to target physical properties based on a prediction result of a pre-trained process result predictor based on pieces of recipe data corresponding to a target process, collecting preference data of an expert for arbitrary process recipe pairs selected from among the candidate process recipes, training a preference predictor to predict preference of the expert for the arbitrary process recipe pairs, using the preference data, and recommending, among the candidate process recipes, a target process recipe for the target process, based on the target physical properties predicted by the pre-trained process result predictor and a preference prediction value predicted by the preference predictor in response to the candidate process recipes.