Hybrid Board Diagnostics Using Rule and Learning Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current board-level diagnostic systems face challenges in handling high design complexity and increasing clock frequencies, often relying on brute-force methods, model-based diagnosis, or artificial neural networks, which lead to ambiguous repair suggestions and increased repair costs due to inefficient root cause identification.
Innovation Solution
A hybrid diagnostic system that generates debug knowledge based on predefined rules and uses a learning-based engine to adjust diagnostic strategies based on feedback, combining rule-based and learning-based techniques for fast and accurate repairs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If rule-based diagnosis is used, then diagnostic speed is improved, but diagnostic accuracy deteriorates due to knowledge acquisition limitations
Solution Approach 1:
The patent combines rule-based diagnosis and learning-based diagnosis into a hybrid system. The rule-based component provides fast initial diagnosis using predefined knowledge, while the learning-based component continuously improves accuracy by learning from historical repair data and feedback, resolving the contradiction between speed and accuracy.
Solution Approach 2:
The system implements feedback mechanisms where repair outcomes are collected and used to continuously update and refine the diagnostic engine. This feedback loop allows the learning-based component to improve diagnostic accuracy over time while maintaining the speed advantages of rule-based initial assessment.
2Measurement precision
If learning-based diagnosis is used, then diagnostic accuracy is improved, but learning process time increases
Solution Approach 1:
The system performs preliminary rule-based diagnosis immediately to provide fast initial results, while the learning-based component operates in the background using historical data. This preliminary action ensures accurate diagnosis is achieved without delaying the diagnostic process.
Solution Approach 2:
The system uses historical repair data from multiple manufacturing facilities as training data for the learning-based diagnosis engine. By learning from copied historical patterns rather than real-time data collection, the system achieves high accuracy without extending the actual diagnostic time.
3Difficulty of detecting and measuring
If brute-force trial-and-error debugging is used, then root cause identification is attempted, but repair cost increases due to ambiguous suggestions
Solution Approach 1:
The patent replaces brute-force trial-and-error mechanical debugging with an intelligent hybrid diagnostic system that uses rule-based reasoning and machine learning to automatically identify root causes. This substitution eliminates ambiguous repair suggestions and reduces repair costs by providing precise, data-driven diagnostic recommendations.
Data Source
AI summary
A method for diagnosing a faulty board includes generating a table of debug knowledge in accordance with predefined debug rules, and configuring a diagnostic engine in accordance with the table of debug knowledge. The method also includes subjecting the faulty board to the diagnostic engine to generate a suggested repair, receiving feedback regarding an effectiveness of the suggested repair, and reconfiguring the diagnostic engine in accordance with the feedback regarding the effectiveness of the suggested repair.