Time-Series Segment Extraction Using Waveform Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data analysis devices struggle to accurately extract segments from time-series data for manufacturing processes without event information indicating event occurrence timing.
Innovation Solution
A data processing device that receives waveform data, parameter information, and transition information, calculates similarity between time-series data and waveform data, detects change points, and sets segment start and end times based on device states, allowing for segment extraction even without event information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If event information is used to extract segments, then segment extraction accuracy is improved, but the method becomes inapplicable when event information is absent
Solution Approach 1:
The patent introduces waveform data as an intermediary element that mediates between the time-series data and the segment extraction process. Instead of directly relying on event information, the system uses waveform data representing typical operating modes as a reference to identify segments through similarity comparison, thus resolving the contradiction between accuracy and adaptability
Solution Approach 2:
The patent creates copies of typical operating mode waveforms (waveform data) that serve as reference templates. These copied waveform patterns are then used to match and identify corresponding segments in the time-series data through similarity calculation, enabling segment extraction without requiring original event information
2Ease of operation
If event information is required for segment extraction, then the extraction process is simple and direct, but the system cannot operate when event information is unavailable
Solution Approach 1:
The patent makes the segment extraction system universal by enabling it to function in multiple scenarios: both with and without event information. The waveform data-based approach serves as a universal mechanism that can replace event information when needed, while still maintaining the ability to use event information when available, thus achieving multi-functionality
3Ease of manufacture
If traditional segment division methods are used, then the process is straightforward, but accuracy deteriorates when event information is missing
Solution Approach 1:
The patent performs preliminary action by pre-collecting and storing waveform data that represents typical operating modes before the actual segment extraction process. This pre-prepared reference data enables accurate segment identification even when event information is unavailable, maintaining both process simplicity and extraction accuracy
Data Source
AI summary
The data processing device includes: an extraction condition input unit receiving waveform data including a change point of a state of a device, parameter information about the waveform data, and transition information of the device; a similarity calculation unit calculating similarity between time-series data of the device and the waveform data; an operating mode determination unit setting the state of the device on the basis of the transition information of the device; a change point detection unit detecting the change point from the time-series data of the device on the basis of the calculated similarity and the determined state of the device, and setting a start time of a segment which is a subsequence of the time-series data and an end time of the segment; and an information output unit outputting, as segment information, the state of the device, the start time and the end time of the segment.