Multivariate Time-Series Anomaly Detection Using Recurrence Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex systems generate vast amounts of sensor data, making it difficult to detect anomalous conditions before fault occurrences, as existing methods rely on preset thresholds and are resource-intensive, overlooking inter-relations among sensor data from multiple independent sensors.
Innovation Solution
A system that determines recurrence data, determinism values, and laminarity values from multivariate time series data without using thresholds, generating a determinism-laminarity curve to identify anomalous states, allowing for unsupervised analysis and co-analysis of data from various sensor types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If preset threshold methods are used to detect fault conditions, then fault detection capability is provided, but resource consumption increases and inter-relations among sensor data are overlooked
Solution Approach 1:
The patent transforms the detection approach by changing from fixed threshold parameters to dynamic recurrence quantification parameters (determinism and laminarity values) that adapt to the actual system state, enabling more efficient anomaly detection without exhaustive resource consumption
Solution Approach 2:
The patent introduces a new dimensional approach by plotting determinism versus laminarity values to create recurrence plots, adding a visual and analytical dimension that captures inter-relations among sensor data without requiring traditional threshold-based resource intensive analysis
2Reliability
If traditional threshold-based methods are used, then fault conditions can be detected, but anomalous conditions preceding faults are missed
Solution Approach 1:
The patent enables preliminary detection of anomalous conditions by analyzing recurrence patterns before fault conditions occur, using determinism and laminarity metrics to identify precursor states that traditional threshold methods miss, allowing preventive action before actual faults develop
3Reliability
If multivariate sensor data from multiple sensors is analyzed, then inter-relational patterns can be identified, but data complexity and analysis difficulty increase
Solution Approach 1:
The patent merges multiple sensor data streams into a unified recurrence analysis framework, combining information from multiple independent sensors into determinism and laminarity metrics that reveal inter-relational patterns while simplifying the overall analysis through a common visualization approach
Data Source
AI summary
A method includes obtaining a plurality of data sets, where each data set of the plurality of data sets includes multivariate time series data for a respective sample period of a plurality of sample periods. The method also includes, for each data set of the plurality of data sets, determining recurrence data indicative of recurrent states in the data set and determining, based on the recurrence data, determinism values of a determinism metric and laminarity values of a laminarity metric. The method further includes determining that a particular data set of the plurality of data sets includes data representing an anomalous state based on a determinism-laminarity curve representing the particular data set, where the determinism-laminarity curve is based on the determinism values of the particular data set and the laminarity values of the particular data set. The method also includes generating output data indicating the anomalous state.