Facility Situation Imaging for Multisensor Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current smart factory systems face challenges in accurately analyzing multivariate time-series data from sensors due to limitations in finding optimal variables and analyzing partial correlations between multiple variables, which hinders effective detection of abnormal situations.
Innovation Solution
A system that converts multivariate time-series data into images to process local connectivity and determines abnormal situations by learning relationships between the images, using a pre-learned situation determination model to generate anomaly scores and detect sensor locations associated with abnormalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If statistical methodology and neural network models are used to analyze sensing data, then automation level is improved, but the ability to find optimal variables and analyze partial correlations deteriorates
Solution Approach 1:
The patent transforms multivariate time-series data into image data, adding a spatial dimension to the analysis. This dimensionality change allows the system to preserve local connectivity information and partial correlations between variables while maintaining automation, as images can represent multiple variables and their relationships in a two-dimensional space where local patterns are visually preserved
Solution Approach 2:
The patent introduces image data as an intermediary representation between the raw sensing data and the analysis model. This intermediary form allows complex multivariate relationships to be captured visually, enabling the neural network to analyze partial correlations more effectively while maintaining automation through the image-based processing pipeline
2Measurement precision
If multiple sensors are installed to measure sensing data, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple sensing data streams from different sensors into a single image representation. By encoding multivariate time-series data from multiple sensors into image pixels and spatial patterns, the system consolidates complex multi-sensor information into a unified visual format, reducing processing complexity while preserving measurement precision from all sensors
3Productivity
If multivariate time-series data is processed directly, then productivity is maintained, but the ability to detect abnormal situations deteriorates
Solution Approach 1:
The patent transforms time-series data into image data, adding spatial dimensions that make abnormal patterns more detectable. This transformation allows the system to maintain processing speed while improving detection reliability, as visual patterns in images are more easily distinguished than raw numerical time-series data, enabling faster and more accurate anomaly identification
Data Source
AI summary
Embodiments relate to a method and system for determining a situation of a facility by imaging a sensing data of the facility including receiving sensing data through a plurality of sensors at a query time, generating a situation image at the query time, showing the situation of the facility at the query time based on the sensing data, and determining if an abnormal situation occurred at the query time by applying the situation image to a pre-learned situation determination model.