Heatmap Encoding for AI Defect Detection in Semiconductor Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current manufacturing processes, particularly in semiconductor fabrication, face challenges in detecting defects in non-image data, such as impedance measurements, which do not benefit from advances in image analysis techniques, limiting the ability to identify patterns and defects effectively.
Innovation Solution
Encoding non-image data as heatmaps, allowing AI models trained for image analysis to detect and classify defects by assigning pixel intensities and colors based on data values, enabling the identification of patterns and flaws that may not be apparent in raw data form.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image analysis techniques are used for defect detection, then detection accuracy is improved, but non-image data such as impedance measurements cannot be analyzed
Solution Approach 1:
The patent introduces heatmaps as an intermediary representation that converts non-image data (impedance measurements, electrical characteristics) into visual image-like formats. This mediator enables AI image analysis models to process and detect defects in non-image data by transforming it into a compatible visual representation that preserves spatial relationships and defect patterns.
Solution Approach 2:
The patent transforms non-image data parameters into visual parameters by mapping numerical values to pixel intensities, colors, and spatial positions in heatmaps. This parameter transformation allows impedance measurements and other non-visual data to be represented in a format that leverages the power of image analysis algorithms for defect detection.
2Measurement precision
If AI models are retrained for different data types, then detection accuracy is maintained, but time and costs increase
Solution Approach 1:
The patent creates a universal defect detection system where a single AI image analysis model can process multiple data types through the heatmap transformation interface. By making the input representation universal (all data types converted to heatmaps), the same trained model can be applied across different manufacturing processes and measurement types without retraining, significantly reducing time and costs while maintaining detection accuracy.
3Device complexity
If traditional defect detection methods are used, then process simplicity is maintained, but defect detection capability is limited
Solution Approach 1:
The heatmap serves as a simple intermediary that bridges traditional straightforward processing pipelines and advanced AI analysis. The transformation from raw data to heatmap is intuitive and maintains process simplicity, while the heatmap format enables deployment of sophisticated AI defect detection models, thereby enhancing capability without significantly complicating the overall system architecture.
Data Source
AI summary
Methods, systems, and apparatuses related to identifying defects in a manufactured product using image recognition are described. Non-imaging data may be converted into an image by encoding the data to a heatmap. The resulting heatmap may be analyzed by one or more artificial intelligence model using image analysis techniques to provide an inference. The inference may indicate a pattern and/or a classification of a pattern of the heatmap.