ML Hardware Module for Diagnosing No Fault Found in Electronic Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex electronic systems often experience intermittent failures that are difficult to diagnose, leading to 'No-Fault-Found' (NFF) conditions, resulting in customer dissatisfaction and increased costs due to the inability to replicate faults in production facilities, necessitating extensive data collection and resource-intensive troubleshooting.
Innovation Solution
A machine learning hardware module integrated into the embedded system's Field Programmable Gate Array (FPGA) monitors I/O interfaces and internal functions in real-time, using neural networks to detect anomalies and record minimal data, thereby identifying failure causes precisely and reducing the need for extensive data collection and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive data collection and resource-intensive troubleshooting are performed to diagnose intermittent failures, then diagnostic accuracy may improve, but cost and time consumption increase significantly
Solution Approach 1:
The system performs preliminary action by continuously monitoring and capturing operational data in real-time during normal system operation. Data capture logic is pre-configured to automatically record relevant parameters when anomalies are detected, so that when intermittent failures occur, the diagnostic information is already available without requiring time-consuming post-failure data collection and analysis
Solution Approach 2:
The patent replaces the mechanical/manual process of extensive data collection and troubleshooting with an automated electronic system. The data capture logic, anomaly detection algorithms, and automated correlation mechanisms substitute for manual troubleshooting procedures, significantly reducing both the time and resources required while maintaining or improving diagnostic accuracy
2Loss of information
If large amounts of data are collected and extensive post-processing is performed to determine fault causes, then diagnostic completeness improves, but computing resource requirements and costs increase
Solution Approach 1:
The system extracts only the essential diagnostic information needed to determine fault causes. The data capture logic is designed to capture specific parameters and operational conditions relevant to anomaly detection, filtering out unnecessary data. This extraction approach maintains diagnostic completeness while significantly reducing the volume of data requiring storage and post-processing
Solution Approach 2:
The patent applies local quality by capturing detailed data only for specific components, functions, or operational conditions where anomalies are detected, rather than uniformly collecting and processing all system data. The system focuses computational resources on analyzing only the relevant local portions of the system where faults are likely to occur, reducing overall computing resource requirements while maintaining diagnostic effectiveness
Data Source
AI summary
A computing method and device for detecting an anomaly event in a hardware- based machine learning anomaly event detector. Data is received from a plurality of input/output interfaces (I/O) or internal functions, wherein the received data of each I/O is associated with a certain sensor or effector. The received data from the plurality of I/O's is analyzed to determine the occurrence of an anomaly event for data from one of the plurality of I/O's. This analysis includes using machine learning techniques dynamically programmed to detect an anomaly event for the I/O data being analyzed by using predetermined parameter values. The predetermined parameter values are retrieved from memory, and are associated with the I/O data being analyzed. Stored in memory for subsequent analysis is at least a portion of the received data from the I/O data being analyzed accompanied with associated neural network output when it is determined an anomaly event occurred for the I/O data being analyzed using the machine learning techniques.