Multi-Scale Time Series Encoding for Machine Condition Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing and processing large volumes of operational data from industrial machines is challenging due to storage limitations and the inability to accurately capture features like slope, highs, and lows, especially when snapshots are not perfectly aligned.
Innovation Solution
A system and method using autoencoders to generate multi-scale encodings of operational time series data, allowing for efficient indexing and processing, enabling similarity searches, anomaly detection, and signal prediction by representing different time resolutions or scales in a compact form.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If snapshots of operational data are created and stored at different points in time, then the ability to review machine performance over time is improved, but storage requirements increase significantly
Solution Approach 1:
The patent extracts only the essential features and patterns from raw operational data using autoencoders, storing only the encoded representations rather than complete data snapshots. This extraction approach maintains the ability to review machine performance while dramatically reducing storage requirements by keeping only the most informative aspects of the data.
Solution Approach 2:
The patent creates compressed digital copies of operational data through encoding, where the encoded representation serves as a surrogate for the original data. These encoded copies retain the essential information needed for performance review and anomaly detection while occupying minimal storage space compared to full data snapshots.
2Measurement precision
If complete operational data snapshots are stored for accurate feature capture, then measurement precision is improved, but processing requirements increase
Solution Approach 1:
The autoencoder architecture extracts only the most salient features and patterns from operational data, discarding redundant information. This extraction process maintains measurement precision for critical features like slope, highs, and lows while significantly reducing the computational power needed for processing and comparing data snapshots.
Solution Approach 2:
The encoding process applies different levels of abstraction to different aspects of the data, focusing computational resources on capturing locally important features such as critical operational patterns and anomalies while using less processing power for less critical data elements.
3Reliability
If snapshots are compared directly to identify patterns, then anomaly detection capability is improved, but the ability to handle misalignment between snapshots deteriorates
Solution Approach 1:
The patent transforms the data from the time domain to an encoded feature space, creating a new dimensional representation where patterns and anomalies can be detected without being constrained by temporal alignment. This dimensionality transformation allows reliable anomaly detection even when snapshots are not perfectly aligned in time, as the encoded representations capture essential patterns independent of exact timing.
Data Source
AI summary
Techniques for managing machine operations using encoded multi-scale time series data are provided. In one technique, operational data is received from a sensor coupled to an industrial device. For each portion in a first set of portions of the operational data (where each portion corresponds to a first time scale), first aggregated data is generated based on time series data from that portion and a first encoding is generated based on the first aggregated data. For each portion in a second set of portions of the operational data (where each portion of the second set corresponds to a second time scale that is different than the first time scale), second aggregated data is generated based on time series data from that portion and a second encoding is generated based on the second aggregated data. The operational data is classified to determine a condition of the industrial device during the time interval based on the first and second encodings.