XAI Defect Inspection for Transparent AI Classification Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based defect inspection methods face challenges in providing a clear basis for determining good/defective products, making it difficult to analyze causes of defects and update parameters in response to environmental changes, leading to continuous misdetermination and potential production process damage.
Innovation Solution
A defect inspecting device and method using explainable artificial intelligence (XAI) to generate a category set for determining good/defective products and continuously update parameters, incorporating a pre-trained second classification model for real-time adaptation and flexible countermeasures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deep learning-based classification model is used for defect inspection, then manufacturing precision is improved, but reliability deteriorates due to lack of clear determination basis
Solution Approach 1:
The patent introduces XAI technology as an intermediary component between the deep learning classification model and the inspection system. This intermediary generates determination bases (explanation data) that reveal why the model made its classification decisions, thereby maintaining high inspection accuracy while improving transparency and reliability of the determination process.
Solution Approach 2:
The patent segments the defect inspection system into distinct functional modules: the deep learning classification model for accurate defect detection, and the XAI module for generating explanation data. This segmentation allows each component to perform its specialized function optimally while the explanation data bridges the gap between model output and human understanding.
2Device complexity
If deep learning algorithm parameters are not updated timely, then device complexity is reduced, but productivity deteriorates due to continuous misdetermination
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously monitors determination results and uses XAI-generated explanation data to detect when environmental changes or process issues affect inspection accuracy. This feedback triggers automated retraining of the classification model with updated data, ensuring the algorithm adapts to changing conditions and maintains high productivity without manual intervention.
Solution Approach 2:
The system performs self-updating by automatically collecting new inspection data, retraining the classification model, and deploying updated parameters without requiring external intervention. This self-service capability maintains high inspection accuracy and productivity while minimizing the complexity of manual algorithm maintenance.
3Adaptability or versatility
If explainable AI is implemented for continuous model updating, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent implements a universal framework where the XAI module serves multiple functions: generating determination bases for transparency, detecting environmental changes through analysis of explanation data, and triggering model retraining when needed. This multi-functionality improves adaptability to environmental changes while avoiding the need for separate dedicated systems for each function, thereby limiting the increase in overall system complexity.
Data Source
AI summary
An AI based defect inspecting device and method is disclosed. The present embodiment, in determining the good or defective product using the deep learning-based classification model based on an image of the product, provides a defect inspecting device and method for providing a basis for determining a good/defective product provided by a deep learning-based classification model using explainable AI (XAI) generating a category set for the basis, and continuously updating parameters of the deep learning-based classification model using the category set.