Segmented Machine Learning Model for Interpretable Optical Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic optical inspection (AOI) systems for manufacturing, especially in electronics production, face challenges in interpreting complex faults in manifold structured products due to their 'black box' nature, leading to high scrap rates and inefficiencies in adapting to new product types, as they often require redefining filters and rules from scratch.
Innovation Solution
A computer-implemented method generates an interpretable machine learning model for fault detection in manufactured products using optical inspection, which segments images into coherent subsets, applies separate detection models to each segment, and couples them with a fault criterion to provide human-interpretable fault probabilities and locations, allowing for easier adaptation to changing product structures and types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional AOI systems use filters and rules for fault detection, then the system is simple to implement, but it becomes increasingly complex to find specialized filters and rules for complex products with densely placed components
Solution Approach 1:
The patent replaces traditional mechanical filter-based AOI systems with a machine learning model that uses image data and component information to detect faults. The ML model automatically learns detection patterns from training data, eliminating the need for manual filter design and rule creation, thus reducing implementation complexity while improving fault detection capability for complex products
Solution Approach 2:
The patent transforms the AOI system from using fixed filters and rules to using a trainable machine learning model with adjustable parameters. The model's parameters are optimized during training based on input data characteristics, allowing the system to adapt to different product complexities without manually redesigning filters, thereby reducing the difficulty of detecting faults in densely placed components
2Adaptability or versatility
If AOI systems adapt to new product types by redefining filters and rules from scratch, then the system can be customized for new products, but the adaptation process becomes very time-consuming
Solution Approach 1:
The patent performs preliminary training of the machine learning model using historical image data and component information before actual fault detection. This pre-training phase allows the model to learn general detection patterns that can be quickly fine-tuned for new product types, significantly reducing adaptation time compared to creating filters and rules from scratch for each new product
Solution Approach 2:
The patent creates a universal machine learning model that can detect faults across different product types by training on diverse image data. The model learns transferable features and patterns that apply to multiple product configurations, enabling quick adaptation to new products without requiring complete reconfiguration, thus maintaining versatility while minimizing adaptation time
3Difficulty of detecting and measuring
If AOI systems use black box machine learning algorithms, then the system can detect complex faults in manifold structured products, but the system cannot reveal inner working making it hard to comprehend decisions, correct them manually, or improve the model
Solution Approach 1:
The patent segments the fault detection process into distinct components: image data processing, component information integration, and decision generation. By structuring the ML model to process and output information in separable stages, the system maintains complex fault detection capability while enabling intermediate inspection and interpretation of each processing stage, thus reducing information loss about decision-making
Solution Approach 2:
The patent introduces an intermediary layer that connects the black box ML algorithm with human operators. This intermediary provides explanations, visualizations, and confidence scores that make the model's decisions comprehensible without altering the underlying complex detection algorithms, thereby maintaining fault detection capability while reducing information loss about decision rationale
Data Source
AI summary
A computer-implemented method for generating a trained machine learning model for automatically detecting faults in a manufactured product using optical inspection is provided, including receiving a multitude of image data of images taken from manufactured products to be inspected, establishing a segmentation machine learning model which inputs the image data of one image and out-puts a first number of image data segments, establishing a detection machine learning model including a separate detection machine learning model for each of the image data segments, combining each of the separate detection machine learning models with the segmentation machine learning model generating a paired machine learning model for each of the image data segments, coupling the paired machine learning models of all image data segments according to a fault criterion, and generating a trained machine learning model by optimizing the coupled paired machine learning models for all image data segments of the received images.