Abnormality Detection Using Ensemble State Estimation Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection technologies face challenges in accurately detecting anomalies in real-time data from IoT sensors due to high noise levels and variable patterns, leading to inconsistent prediction errors and delayed or false detections.
Innovation Solution
An abnormality detection apparatus utilizing multiple state estimation models to calculate deviation and abnormality degrees, which includes an estimated data calculator, deviation degree calculator, abnormality degree calculator, and determiner, to determine the presence of abnormalities based on predicted and actual state data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single deterministic model is used for prediction, then the device complexity is low, but the prediction accuracy deteriorates due to noise and variable patterns
Solution Approach 1:
The patent divides the prediction task into multiple independent deterministic models, each handling different aspects or time ranges of the prediction. By segmenting the prediction into multiple models rather than using one complex model, the system achieves better accuracy while keeping individual models simple and manageable.
Solution Approach 2:
The patent combines multiple deterministic models into an ensemble prediction system. The results from multiple models are aggregated (e.g., through averaging or voting) to produce the final prediction, thereby leveraging the strengths of each individual model to overcome noise and variable patterns that would affect any single model.
2Measurement precision
If multiple state estimation models are used to improve prediction accuracy, then the prediction accuracy improves, but the device complexity increases
Solution Approach 1:
The system segments the abnormality detection task by using multiple specialized state estimation models, each optimized for different normal patterns or conditions. This segmentation allows each model to focus on specific aspects, improving overall detection accuracy while maintaining manageable complexity through modular design.
Solution Approach 2:
The multiple state estimation models serve universal purposes in the system - they all contribute to predicting normal patterns and detecting abnormalities. Rather than requiring different complex systems for different detection scenarios, the same framework with multiple models handles various detection needs, achieving multi-functionality with a unified approach.
3Measurement precision
If manual feature value creation for each abnormal pattern is performed, then the detection precision for known patterns improves, but the loss of time for processing and the device complexity increase
Solution Approach 1:
The system employs unsupervised learning models that automatically learn normal patterns from data without requiring manual feature engineering. The models self-adapt to the data characteristics and automatically detect deviations, eliminating the time-consuming manual process of creating feature values for each abnormal pattern while maintaining high detection precision.
Solution Approach 2:
The system changes the approach from manual parameter (feature value) specification to automatic parameter learning through machine learning models. By transforming the problem from manual feature extraction to automated model-based pattern recognition, the system achieves high precision detection without the time penalty of manual processing.
Data Source
AI summary
An apparatus according to one embodiment of the present invention detects an abnormality of a monitoring target on the basis of state data of the target and includes an estimated data calculator, a deviation degree calculator, an abnormality degree calculator, and an abnormality determiner. The estimated data calculator calculates estimated data of a second period on the basis of the state data of the first period. The deviation degree calculator calculates a degree of deviation of the second period on the basis of the state data and the estimated data of the second period. The abnormality degree calculator calculates a degree of abnormality of the second period on the basis of the degree of deviation of the second period. The abnormality determiner determines presence or absence of an abnormality of the target in the second period on the basis of the degree of abnormality in the second period.