Anomaly Detection Models for Limited Industrial Operation Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems in industrial and infrastructure systems face challenges in accurately monitoring operation states with limited data, leading to potential misdetection of malfunctions or anomalies, especially when event signals are absent or mode division is incorrect.
Innovation Solution
An anomaly detection system that collects and processes operation data using a device and information processing apparatus to construct anomaly detection models, specifically employing the k-nearest neighbor method and ensemble learning to minimize variance in anomaly scores, enabling accurate anomaly detection even in conditions with limited data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mode division is performed based on event signals to create normal models for each mode, then anomaly detection accuracy is improved for systems with clear operation state transitions, but misdetection occurs when event signals are absent or mode division cannot be performed correctly
Solution Approach 1:
The patent segments operation data into different density regions (high density, medium density, low density) based on the amount of available data, and applies different anomaly detection methods to each segment. This segmentation allows the system to handle diverse data conditions appropriately, improving both accuracy and reliability across different operation states.
Solution Approach 2:
The patent changes the parameter of data density threshold to dynamically adjust the anomaly detection approach. By monitoring data density parameters, the system switches between different detection methods (model-based for high density, statistical for medium density, and simple threshold for low density), ensuring reliable detection regardless of data availability.
2Adaptability or versatility
If statistical methods are used for anomaly detection based on operation data, then detection capability is provided for systems with various operation states, but misdetection occurs when data density is low or minority operation data is present
Solution Approach 1:
The patent implements a dynamic anomaly detection system that automatically adjusts the detection method based on real-time data density conditions. The system transitions between different detection strategies (model-based, statistical, threshold-based) as data density changes, maintaining both versatility and precision across varying operation states.
Solution Approach 2:
The patent applies partial action by using different levels of sophisticated detection methods appropriate to each data density level. For low density data, a simpler threshold-based method is used rather than attempting to apply complex statistical methods that would require more data, thus avoiding misdetection while maintaining adaptability.
3Reliability
If anomaly detection models are constructed from operation data, then systematic anomaly detection is achieved, but high computational complexity and long learning time occur when processing large volumes of operation data
Solution Approach 1:
The patent segments the operation data processing into different density-based batches, processing only the necessary portions of data for model construction. By dividing data into high, medium, and low density segments and applying appropriate methods to each, the system reduces overall learning time while maintaining systematic detection capability.
Solution Approach 2:
The patent applies local quality by using different processing intensities for different data segments. High density data receives model-based processing with higher computational effort, while low density data uses simpler threshold-based methods with lower computational cost, optimizing the balance between systematic detection and learning time.
Data Source
AI summary
There are provided a device that collects operation data from equipment, and an information processing apparatus that detects an anomaly or an omen of an anomaly of the equipment on the basis of anomaly detection models constructed from the operation data, the information processing apparatus including means for collecting the operation data, means for learning anomaly detection models from the operation data, and means for calculating an anomaly score of respective operation data from the operation data and the anomaly detection models, the means for learning an anomaly detection model in which a dispersion of elements is small among the anomaly detection models. Thus, in an anomaly detection system, when the operation state of equipment is monitored, even if data for performing division of operation states cannot be obtained or even if division cannot be performed correctly, misdetection of an anomaly such as a malfunction or a failure or an omen of an anomaly can be decreased and the state of the system can be evaluated correctly.