Time-Series Segment Learning for Defect Detection With Time Shift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect detection technologies struggle with determining similarity between time-series data segments when there is allowable displacement in the time direction, leading to incorrect determination of abnormal segments due to estimated large distances using methods like Euclidean distance.
Innovation Solution
A learning device that collects and processes time-series data from sensors, divides it into segments representing operation states with both rises and falls, classifies similar segments, and generates sample segments showing normal operation regions to improve similarity determination with a margin in the time direction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Euclidean distance is used to determine similarity between time-series data segments, then calculation simplicity is improved, but measurement precision deteriorates when time displacement exists
Solution Approach 1:
The patent segments the time-series data into multiple fixed-time segments and compares corresponding segments between normal and abnormal time periods. By dividing the continuous time-series data into discrete segments, the system can perform localized comparisons that are less sensitive to overall time displacement, thus improving similarity determination accuracy while maintaining computational efficiency through straightforward segment-by-segment Euclidean distance calculations.
2Measurement precision
If fixed-time segment comparison is used, then similarity determination with time margin is improved, but device complexity increases
Solution Approach 1:
The patent divides time-series data into multiple fixed-time segments and compares corresponding segments between normal and abnormal periods. This segmentation approach enables the system to account for time margins by allowing partial overlaps and mismatches in segment alignment, thereby improving similarity determination accuracy while keeping the processing methodology systematic and manageable.
Solution Approach 2:
The patent employs partial segment matching where complete alignment of all segments is not required. By allowing some segments to match partially or with time offsets, the system achieves more accurate defect detection without requiring perfect temporal alignment, thus reducing the complexity of precise synchronization while maintaining high measurement precision.
Data Source
AI summary
A learning device comprising: a training time-series data acquisition unit to collect both training time-series data acquired by a sensor mounted on a target device, and set parameter data of the target device or environment data concerning the target device, while associating the training time-series data with the set parameter data or the environment data; a segment set generation unit to divide the training time-series data into training segments, to generate a segment set containing the training segments; a segment set sort unit to classify the training segments contained in the generated segment set into at least one similar segment set, using either the set parameter data or the environment data; and a sample segment generation unit to generate a sample segment showing a normal region of the operation of the target device from the training segments contained in the at least one similar segment set.