Manufacturing Data Analysis System for Semiconductor Yield Root Cause Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In manufacturing processes, particularly in semiconductor production, the vast amount of manufacturing data makes it difficult to identify the root cause of yield loss and key factors effectively, as existing methods lack efficient methods to screen and reserve top key models and their associated factors.
Innovation Solution
An analyzing method and system that screens and ranks models based on correlation values and frequency information, using units such as model generating, screening, factor listing, and reporting to efficiently identify key models and factors by filtering and ranking them layer by layer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If big data technology is used to analyze manufacturing data, then the ability to find root causes is improved, but the complexity of analysis increases due to the huge number of models
Solution Approach 1:
The patent segments the huge number of models into groups based on correlation values with target parameters. By dividing models into different correlation value ranges and analyzing them separately, the system reduces the complexity of analyzing all models simultaneously while maintaining the ability to identify root causes accurately.
Solution Approach 2:
The patent extracts and focuses on the most significant models by filtering out models with low correlation values. This extraction process removes irrelevant models from further analysis, reducing the complexity of the analysis system while preserving the key models that are most likely to contain root cause information.
2Loss of information
If all models are analyzed to find key factors, then the completeness of factor identification is improved, but the time and resources required increase significantly
Solution Approach 1:
The patent performs preliminary screening of models based on correlation values before conducting detailed factor analysis. By pre-filtering models that show strong correlation with target parameters, the system prepares a reduced set of candidate models for further analysis, saving time and resources while maintaining completeness of key factor identification.
Solution Approach 2:
The patent changes the analysis parameter from examining all models to examining models ranked by correlation value. This parameter change allows the system to focus computational resources on the most promising models, significantly reducing analysis time while maintaining the ability to identify all key factors through systematic ranking.
3Productivity
If the number of models is reduced to improve analysis efficiency, then the analysis speed is improved, but the risk of missing key models increases
Solution Approach 1:
The patent implements a feedback mechanism where models are ranked by correlation value and the ranking information is used to guide further analysis. This feedback loop ensures that models with higher correlation values are prioritized, maintaining high reliability in key model selection while improving analysis efficiency by avoiding detailed examination of low-correlation models.
Solution Approach 2:
The patent introduces correlation value as a filtering parameter to reduce the model set. By changing from analyzing all models to analyzing models above a certain correlation threshold, the system achieves both improved efficiency and maintained reliability, as the correlation parameter objectively identifies models most likely to be key models.
Data Source
AI summary
An analyzing method and an analyzing system for manufacturing data are provided. The analyzing method includes the following steps. A plurality of models each of which has a correlation value representing a relationship between at least one of a plurality of factors and a target parameter are provided. The models are screened according to the correlation values. A rank information and a frequency information of the factors are listed up according to the models. The factors are screened according to the rank information and the frequency information. The models are ranked and at least one of the models is selected.