Seasonality Compensation for Out-of-Phase Time-Series Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional seasonality-characterization techniques fail to effectively detect anomalies in time-series sensor signals from critical assets due to multiple out-of-phase seasonality modes and dynamic lead/lag relationships, which are not accounted for in their assumptions of in-phase modes and fixed numbers of seasonality components.
Innovation Solution
A system that identifies and filters out seasonality modes by determining frequencies and phase angles using cross power spectral density (CPSD) and applies an inferential model, such as Multivariate State Estimation Technique (MSET), to seasonality-compensated time-series signals for anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional seasonality-characterization techniques are used, then the technique is simple to implement, but it performs poorly when multiple out-of-phase seasonality modes exist
Solution Approach 1:
The patent segments the complex seasonality characterization problem into distinct frequency components using spectral analysis. By decomposing the time-series signal into its frequency spectrum, the system identifies and processes each seasonality mode separately, allowing accurate handling of multiple out-of-phase modes without requiring a single complex model.
Solution Approach 2:
The patent introduces an intermediary phase adjustment mechanism that operates between signal decomposition and anomaly detection. This intermediary component calculates phase differences between seasonality modes and applies corrective transformations, enabling accurate anomaly detection in the presence of out-of-phase modes without complicating the overall system architecture.
2Measurement precision
If traditional seasonality-characterization techniques assume in-phase modes, then the assumption simplifies analysis, but it fails when modes are out of phase with each other
Solution Approach 1:
The patent changes the measurement parameter from time-domain phase relationships to frequency-domain spectral characteristics. By transforming the signal via Fourier analysis, the system measures frequency magnitudes and phases separately, enabling precise identification of out-of-phase modes through spectral peaks rather than attempting to measure time-aligned phase relationships directly.
Solution Approach 2:
The patent transitions from analyzing seasonality modes in the time dimension to analyzing them in the frequency dimension. This dimensional transformation allows simultaneous identification of multiple modes with different phases by examining spectral peaks at different frequencies, converting a difficult time-domain phase alignment problem into a simpler frequency-domain magnitude detection problem.
3Adaptability or versatility
If the number of seasonality modes is not known a priori, then the system must adapt dynamically, but traditional techniques cannot handle this scenario
Solution Approach 1:
The patent implements a dynamic seasonality characterization approach where the number and parameters of seasonality modes are determined adaptively from the data itself. Using spectral analysis, the system automatically identifies the number of significant spectral peaks, which corresponds to the number of seasonality modes present in the signal, eliminating the need for predetermined mode counts while maintaining detection accuracy.
4Adaptability or versatility
If lead and lag times among seasonality modes change dynamically, then the system must be flexible, but traditional techniques assume fixed relationships
Solution Approach 1:
The patent handles dynamic lead and lag times by transforming the problem from the time domain to the frequency domain. In the spectral representation, phase relationships between modes are captured as phase angles at different frequencies, which can vary dynamically without requiring time-domain alignment. This allows the system to adapt to changing lead/lag relationships while maintaining reliable anomaly detection.
Data Source
AI summary
The disclosed embodiments provide a system that performs seasonality-compensated prognostic-surveillance operations for an asset. During operation, the system obtains time-series sensor signals gathered from sensors in the asset during operation of the asset. Next, the system identifies seasonality modes in the time-series sensor signals. The system then determines frequencies and phase angles for the identified seasonality modes. Next, the system uses the determined frequencies and phase angles to filter out the seasonality modes from the time-series sensor signals to produce seasonality-compensated time-series sensor signals. The system then applies an inferential model to the seasonality-compensated time-series sensor signals to detect incipient anomalies that arise during operation of the asset. Finally, when an incipient anomaly is detected, the system generates a notification regarding the anomaly.