Graph-Based Optimal Solution Search for Substrate Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in efficiently searching for optimal process conditions to achieve target results, as existing machine learning models require numerous experiments and lack visualization tools for operators to select optimal solutions effectively.
Innovation Solution
An information processing apparatus that includes a learning trainer to train machine learning models to infer processing results based on process conditions, a graph creator to plot these results on a graph with achievement levels, and an information display to assist operators in selecting optimal process conditions, thereby reducing the need for repetitive experiments and improving inference accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to search for optimal process conditions, then inference accuracy is improved, but the number of experiments required increases
Solution Approach 1:
The patent transforms the search space from a high-dimensional process condition space into a two-dimensional graph display space. The graph creator plots multiple target values on a graph with two axes, allowing operators to visually evaluate multiple process conditions simultaneously. This dimensional transformation enables efficient comparison and selection of optimal solutions without requiring exhaustive experimentation in the original high-dimensional space.
Solution Approach 2:
The patent creates visual copies of process conditions and their results in the form of plotted points on the graph. Each process condition is represented as a point on the graph, and multiple target values are displayed as multiple axes. This visual copying allows operators to evaluate numerous process conditions at once without physically executing each one, significantly reducing the number of required experiments.
2Manufacturing precision
If multiple target values are evaluated to find optimal process conditions, then solution quality is improved, but device complexity increases
Solution Approach 1:
The patent handles multiple target values by plotting them on a graph with two axes, where each axis represents a different target value. This transformation allows the system to evaluate multiple objectives simultaneously in a simplified two-dimensional visual format, avoiding the need for complex multi-objective optimization algorithms while maintaining solution quality.
Solution Approach 2:
The graph serves as an intermediary between the machine learning model's multiple target values and the operator's decision-making process. Instead of directly processing multiple complex target values, the system visualizes them on a graph where operators can easily compare and select optimal process conditions, simplifying the overall system complexity.
3Ease of operation
If visual tools are provided for operators to select optimal solutions, then ease of operation is improved, but information display complexity increases
Solution Approach 1:
The patent simplifies operator interaction by transforming complex multi-dimensional process data into a two-dimensional graph display. Operators can visually evaluate multiple process conditions and target values on a simple two-axis graph, making selection intuitive and easy without requiring complex display interfaces or advanced analytical skills.
Solution Approach 2:
The information display unit highlights the selected optimal solution on the graph, making it visually distinct from other options. This visual emphasis guides operators to the best choice without requiring them to analyze complex data, improving ease of operation while keeping the display relatively simple.
Data Source
AI summary
An information processing apparatus includes: a learning trainer configured to train a machine learning model to train a relationship between a process condition and a processing result of a substrate processing apparatus that has executed a processing based on the process condition; an inferrer configured to infer a plurality of processing results depending on a plurality of process conditions using the trained machine learning model; a graph creator configured to plot the plurality of processing results inferred with the machine learning model on a graph with an achievement level for a plurality of target values of the plurality of processing results as a plurality of axes; and an information display configured to display, on the graph, information used by an operator to select an optimal solution for the process condition, based on the plot of the plurality of inferred processing results.