CNN Root Cause Analysis for Genotyping Process Cycle Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current genotyping processes face challenges in identifying the root cause of failures due to mechanical and chemical processing errors, as the quality scores generated do not provide insight into the underlying causes of low-quality processes.
Innovation Solution
The implementation of a system that uses deep learning-based image classification, specifically convolutional neural networks (CNNs), to analyze process cycle images from genotyping instruments. This system classifies images as good or bad and further determines the root cause of failure by identifying specific defects such as hybridization failures, spacer shift errors, and reagent flow issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional quality scoring methods are used to evaluate process cycles, then the evaluation is simple and fast, but the method cannot provide insight into the root cause of failure
Solution Approach 1:
The patent segments the image analysis into multiple specialized components: a quality scorer that evaluates overall image quality, and multiple root cause classifiers that detect specific failure modes (hybridization failure, spacer shift, offset failure, surface abrasion, reagent flow failure). Each component focuses on a specific aspect, enabling comprehensive root cause analysis without requiring a single complex system to handle all possibilities simultaneously.
Solution Approach 2:
The patent introduces an intermediary processing layer that takes the original process cycle image and generates both a quality score and multiple root cause assessments. This intermediary analysis layer acts as a mediator between the raw image data and the final diagnostic conclusions, enabling detailed root cause identification while maintaining a separate quality evaluation pathway.
2Measurement precision
If deep learning-based image classification is implemented to identify root causes, then the accuracy of failure analysis is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent implements preliminary quality filtering before detailed root cause analysis. The quality scorer first evaluates the overall quality of process cycle images, and only images that meet certain quality thresholds undergo the more computationally intensive root cause classification. This preliminary action filters out low-quality images that would waste computational resources regardless of root cause analysis.
Solution Approach 2:
The analysis process is segmented into two distinct stages: quality assessment and root cause classification. By dividing the computational workload into these separate stages, the system can quickly eliminate obviously poor-quality images at the first stage, reducing the number of images that require the more time-consuming deep learning-based root cause analysis in the second stage.
3Adaptability or versatility
If multiple failure categories are detected and classified, then the comprehensiveness of failure analysis is improved, but the complexity of the classification system increases
Solution Approach 1:
The patent divides the comprehensive failure detection task into multiple specialized classifiers, each trained to detect a specific failure mode (hybridization failure, spacer shift, offset failure, surface abrasion, reagent flow failure). This segmentation allows each classifier to be simpler and more specialized, while collectively they provide comprehensive coverage of all possible failure types through a modular architecture.
Solution Approach 2:
The system employs a universal deep learning framework that can handle multiple failure categories through a common architecture. The same base neural network model serves multiple functions by being configured with different classification heads or output layers for different failure types, reducing the need for separate specialized systems for each failure mode.
Data Source
AI summary
The technology disclosed relates to training a convolutional neural network (CNN) to identify and classify images of sections of an image generating chip resulting in process cycle failures. The technology disclosed includes creating a training data set of images of dimensions M×N using labeled images of sections of image generating chip of dimensions J×K. The technology disclosed can fill the M×N frames using horizontal and vertical reflections along edges of J×K labeled images positioned in M×N frames. A pretrained CNN is further trained using the training data set. Trained CNN can classify a section image as normal or depicting failure. The technology disclosed can train a root cause CNN to classify process cycle images of sections causing process cycle failure. The trained CNN can classify a section image by root cause of process failure among a plurality of failure categories.