Time-Series Analyzer Using Sliding Windows for Multi-Signal Relations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to effectively analyze the relationship between multiple pieces of time-series data from industrial machines, especially when the data volume is large or the acquisition period is long, making it difficult for operators to determine trends and relationships.
Innovation Solution
An analyzer that employs association analysis by extracting interval data using a sliding window approach, simplifying change trends, and generating combination data for easier analysis, allowing for the application of known data analysis methods to time-series data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the number of pieces of data acquired from the industrial machine is large or when the data acquisition period is relatively long, then the data volume increases, but it becomes difficult for operators to visually grasp the relationship between the pieces of data
Solution Approach 1:
The patent divides time-series data into multiple intervals using a sliding window approach, where each interval is analyzed separately. This segmentation transforms a large, complex dataset into manageable segments that can be visually analyzed, while still capturing the overall relationships through systematic processing of each segment
Solution Approach 2:
The patent introduces a new dimension by converting time-series data into interval-based representations with extracted features (trend, amplitude, frequency characteristics). This dimensional transformation allows complex temporal relationships to be visualized and analyzed in a different space, making large datasets more comprehensible
2Measurement precision
If traditional methods such as Euclidean distance, cross-correlation function, or dynamic time warping are used to calculate similarity between time-series data, then similarity can be calculated, but it is difficult to simultaneously grasp the relationship between multiple pieces of data by paying attention to characteristic parts
Solution Approach 1:
The patent extracts characteristic features (trend, amplitude, frequency) from each interval of time-series data. By taking out these essential characteristics and representing them in a simplified format, the system can analyze relationships between multiple datasets without being overwhelmed by the complexity of raw time-series comparisons
Solution Approach 2:
The patent transforms time-series data by changing parameters - converting continuous temporal data into discrete interval representations with specific feature parameters. This parameter transformation enables the application of association analysis rules to time-series data, simplifying the analysis of relationships between multiple datasets while maintaining measurement precision
Data Source
AI summary
An analyzer includes a preliminary analysis unit that extracts interval data obtained by cutting out time-series data with a predetermined sliding window width, and analyzes simplicity of a change trend of the extracted interval, a data division unit that divides the data into pieces of division data with a sliding window width set based on an analysis result performed by the preliminary analysis unit, a data generation unit that generates combination data which is a text indicating a change trend in the division data based on the division data, and a data analysis unit that analyzes the combination data.