State Observer Abnormality Detection for Variable Environments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection systems face challenges in accurately identifying device malfunctions due to variations in external environments and lack of consideration for device functions or malfunction mechanisms, leading to erroneous determinations and undesired learning directions.
Innovation Solution
An abnormality detection device that utilizes a competitive neural network to generate a state observer reflecting device functions or malfunction mechanisms, calculates an abnormality degree, and determines the presence of abnormalities by comparing thresholds with the calculated degree, using learning and monitoring target data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a competitive neural network is used for abnormality monitoring, then detection accuracy improves, but the system fails to account for variations in external environments and device functions, leading to erroneous determinations
Solution Approach 1:
The patent implements dynamic adaptation by allowing the neural network to learn and adjust to different external environments and device states. The system continuously updates its understanding of normal variations based on input data, enabling it to dynamically adapt to changing conditions while maintaining accurate abnormality detection. This resolves the contradiction by making the system both precise and adaptable through continuous learning mechanisms.
Solution Approach 2:
The system changes its internal parameters (weights and biases of the neural network) based on learned patterns from training data that includes various external environment conditions. By adjusting these parameters to reflect real-world variations, the system maintains high detection accuracy across different environments. This parameter adaptation allows the system to be both precise in detection and versatile in handling environmental variations.
2Reliability
If machine learning is applied to detect abnormalities, then detection capability improves, but the system lacks consideration for device functions or malfunction mechanisms, causing undesired learning directions
Solution Approach 1:
The patent incorporates device function knowledge and malfunction mechanisms into the training process before actual deployment. By pre-loading the neural network with domain-specific knowledge about how devices should function and what malfunctions to look for, the system is guided in the correct learning direction from the start. This preliminary incorporation of expert knowledge prevents the system from learning spurious patterns while maintaining high reliability.
Solution Approach 2:
The system uses feedback mechanisms where the neural network's predictions are compared against known device behavior patterns and malfunction characteristics. This feedback loop allows the system to continuously refine its understanding of device functions and adjust its detection criteria accordingly. By incorporating domain knowledge through feedback, the system maintains reliability while avoiding loss of important device-specific information.
Data Source
AI summary
An abnormality detection device, method, or a storage medium acquires learning target data and monitoring target data, generates a state observer by using a variable in an input variable configuration, generates a threshold, calculates an abnormality degree by combining a second state observation value and the monitoring target data and inputting a combined result to the competitive neural network, and calculates a determination result.