Time-series Data Processing for Abnormal State Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing time-series data in monitoring systems struggle to accurately and efficiently detect abnormal states, particularly in complex systems with large amounts of data, leading to delayed pattern recognition.
Innovation Solution
The proposed method involves extracting partial time-series data sets, calculating correlation data, and generating coded data based on the time-series and correlation data. This process allows for the creation of models that can quickly identify abnormal states by comparing new data to past patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If correlation destruction patterns are stored without considering time-series behavior, then the system can detect correlation destruction, but it cannot accurately distinguish different states with similar patterns
Solution Approach 1:
The patent segments time-series data into multiple units (e.g., 1-minute intervals) and analyzes correlation destruction patterns for each segment. By dividing the continuous time-series data into discrete units, the system can capture temporal dynamics and distinguish between transient and persistent abnormal states, thereby improving state specification accuracy while managing information effectively.
Solution Approach 2:
The patent performs preliminary analysis by storing correlation destruction patterns with their temporal context before actual monitoring occurs. The system pre-processes historical data to establish baseline correlation patterns and stores them for comparison, enabling faster and more accurate real-time state specification without losing time-series behavior information.
2Reliability
If all correlation destruction information is stored for complex systems, then comprehensive monitoring is achieved, but the search time increases due to enormous data volume
Solution Approach 1:
The patent divides complex systems into multiple subsystems or monitoring units, each with its own correlation destruction pattern database. This segmentation reduces the search space for each query and allows parallel processing, thereby maintaining comprehensive monitoring coverage while significantly reducing pattern search time compared to a monolithic approach.
Solution Approach 2:
The patent applies different storage and processing strategies to different parts of the system based on their characteristics. Critical subsystems with high failure risk receive more detailed correlation destruction pattern storage, while less critical subsystems use simplified approaches. This local differentiation maintains reliability for important components while reducing overall data volume and search time.
Data Source
AI summary
An information processing device 100 of the present invention includes an analysis unit 121 and an encoding unit 122. The analysis unit 121 extracts a partial time-series data set obtained by dividing a time-series data set that is a set of time-series data including a plurality of elements at given time intervals, and calculates correlation data representing a correlation between elements of time-series data included in the partial time-series data set. The encoding unit 122 generates coded data based on the time-series data of the partial time-series data set and the correlation data.