Multi-Modal Manufacturing Data Analysis for Semiconductor Defect Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional manufacturing data analysis methods only analyze one type of data at a time, leading to inaccurate defect cause identification and modification suggestions in semiconductor processes.
Innovation Solution
A method and device that transform numerical, image, and text data into vectors, combining them to input into an inference model for simultaneous analysis to accurately determine defect causes and modification suggestions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only one manufacturing data is analyzed at a time, then the analysis process is simple, but the defect cause cannot be obtained accurately
Solution Approach 1:
The patent combines multiple heterogeneous manufacturing data types (numerical, image, and text data) into a unified analysis framework. By merging these different data sources and transforming them into vectors that can be processed together, the system achieves accurate defect cause identification while managing complexity through standardized processing pipelines.
2Measurement precision
If multiple manufacturing data are analyzed simultaneously, then the defect cause can be obtained accurately, but the data processing complexity increases
Solution Approach 1:
The patent transforms heterogeneous manufacturing data (numerical, image, text) into a unified vector representation. This parameter transformation allows different data types to be processed simultaneously through the same inference model, achieving accurate multi-data analysis while managing complexity through standardized vector operations.
Solution Approach 2:
The patent segments the analysis process into distinct functional components: data transformation modules that convert different data types into vectors, a gathering unit that combines vectors, and an inference model that processes the combined data. This segmentation allows complex multi-data analysis to be managed through modular, independent processing stages.
3Reliability
If traditional manual analysis method is used, then the process is easy to operate, but the modify suggestion may not be correct
Solution Approach 1:
The patent implements an automated inference model that processes combined manufacturing data vectors and generates defect causes and modify suggestions. This feedback-driven automated system improves reliability by systematically analyzing multiple data sources together, reducing human error while maintaining ease of operation through automated decision support.
Data Source
AI summary
A manufacturing data analyzing method and a manufacturing data analyzing device are provided. The manufacturing data analyzing method includes the following steps. Each of at least one numerical data, at least one image data and at least one text data is transformed into a vector. The vectors are gathered to obtain a combined vector. The combined vector is inputted into an inference model to obtain a defect cause and a modify suggestion.