Quality Prognostics System for Semiconductor Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods in semiconductor and TFT-LCD manufacturing fail to predict product quality effectively for future product lots, relying on post-production inspection and resulting in defective products and reduced yield, with existing solutions limited in applicability and flexibility.
Innovation Solution
A quality prognostics system utilizing conjecture and prediction modeling means, incorporating neural networks, fuzzy logic, and weighted moving averages, which processes raw data to forecast product quality based on current and historical measurement values, with self-searching and self-adjusting mechanisms to optimize accuracy across various tools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If post-production inspection is used to determine product quality, then measurement accuracy is improved, but defective products are already produced before detection
Solution Approach 1:
The system performs preliminary quality prediction before actual production by analyzing process parameters and historical data. The quality prediction module forecasts product quality in advance, allowing preventive actions to be taken before defective products are manufactured, thus resolving the contradiction between measurement accuracy and time loss.
Solution Approach 2:
The system implements a feedback mechanism where predicted quality results are fed back to the production process. The system continuously monitors process parameters, predicts quality outcomes, and provides feedback for process adjustment, enabling real-time quality control and preventing defective product production.
2Difficulty of detecting and measuring
If process parameters are monitored to judge product quality defects, then defect detection capability is improved, but defective products have already been produced by the time defects are detected
Solution Approach 1:
The system performs preliminary quality assessment by analyzing process parameters during production and predicting future quality outcomes. By using historical data and process parameter trends, the system predicts quality issues before they manifest as actual defects, eliminating the time delay between defect occurrence and detection.
Solution Approach 2:
The system introduces an intermediary prediction layer between process monitoring and defect detection. Instead of directly detecting defects in finished products, the system uses process parameters and historical data as intermediaries to predict quality outcomes, enabling earlier intervention before defects are actually produced.
3Productivity
If batch production is used, then production efficiency is improved, but entire product lots must be discarded when defects are discovered
Solution Approach 1:
The system performs preliminary quality prediction for each product lot before production completion. By predicting quality outcomes in advance, the system identifies potential defective lots early, allowing selective handling of only affected units rather than discarding entire batches, thus reducing material loss while maintaining batch production efficiency.
Solution Approach 2:
The system changes the approach from binary batch acceptance/rejection to continuous quality prediction based on multiple parameters. By analyzing process parameters, historical data, and prediction models, the system determines the actual quality status of each lot, enabling more precise decisions about which specific units need to be discarded rather than entire batches.
4Reliability
If existing quality prediction systems are applied, then prediction capability is improved for current production, but applicability is limited to certain equipment types and cannot predict next product lot quality
Solution Approach 1:
The system implements a universal quality prediction framework that can be applied across different equipment types and production processes. The prediction module uses generalizable algorithms and data structures that adapt to various tool types, enabling the same system to predict quality for different product lots and equipment without requiring complete reconfiguration, thus improving both reliability and versatility.
Data Source
AI summary
A quality prognostics system and a quality prognostics method for predicting the product quality during manufacturing processes are disclosed, wherein the current production tool parameters sensed during the manufacturing process and several previous quality data collected from the measurement tool are utilized to predict the future product quality, and a conjecture modeling step and prediction modeling step are performed respectively. The conjecture modeling step itself also can be applied for the purpose of virtual metrology. Further, a self-searching step and a self-adjusting step are performed for searching the best combination of various parameters/functions used by the conjecture algorithm or prediction algorithm; and meeting the requirements of new equipment parameters and conjecture/prediction accuracy.