Anomaly Detection via Dynamic Spectral Search Windows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting anomalies in mechanical or electromechanical systems, such as wear or malfunction, face challenges in accurately identifying fundamental, harmonic, and modulation band frequencies from sensor signals due to the complexity of frequency spectra.
Innovation Solution
A method involving a processing device that identifies spectral components by determining lower and upper frequency limits of search windows based on center frequencies and margins of error, allowing for the detection of harmonics and modulation components, and subsequently modifying these parameters to refine the identification of additional components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional frequency spectrum analysis is used to detect anomalies, then the method is simple to implement, but the accuracy of identifying fundamental, harmonic, and modulation band frequencies is limited
Solution Approach 1:
The patent segments the frequency spectrum analysis into multiple targeted search windows, each focused on specific frequency ranges where fundamental, harmonic, and modulation components are expected to appear. This segmentation allows the system to concentrate computational resources on relevant frequency regions rather than analyzing the entire spectrum, thereby improving identification accuracy while managing processing complexity.
Solution Approach 2:
The patent performs preliminary actions by pre-defining search windows based on expected frequency relationships (fundamental frequency, its harmonics, and modulation sidebands). These search windows are established before detailed analysis, allowing the system to efficiently locate and identify spectral components in their expected positions, improving both accuracy and computational efficiency.
2Adaptability or versatility
If the search window size is increased to capture all possible spectral components, then the detection coverage is improved, but the precision of frequency identification deteriorates
Solution Approach 1:
The patent applies local quality by creating search windows with different sizes and positions tailored to specific frequency regions. Each search window is optimized for its local frequency context, allowing high precision in identifying components within each window while maintaining broad coverage through the collection of multiple windows. This local optimization resolves the contradiction between wide coverage and precise identification.
Solution Approach 2:
The patent introduces an additional dimension to the analysis by organizing spectral component identification across multiple search windows arranged in a hierarchical structure. This multi-window approach adds a spatial dimension to the frequency analysis, allowing the system to achieve both broad coverage and high resolution by distributing the analysis across multiple focused regions rather than using a single broad window.
3Reliability
If multiple spectral components are identified to improve anomaly detection accuracy, then the reliability of anomaly identification is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the identification of multiple spectral components into separate, targeted search operations within defined windows. Each search window focuses on identifying specific components (fundamental, harmonics, modulation sidebands) in predetermined frequency regions. This segmentation reduces the computational burden compared to a comprehensive full-spectrum analysis while still capturing all relevant components for reliable anomaly detection.
Solution Approach 2:
The patent performs preliminary action by pre-identifying and flagging potential spectral components within search windows before conducting detailed anomaly analysis. This preliminary identification step organizes the computational work, allowing the system to efficiently process multiple components by having already located and characterized them in the preliminary search phase, thereby reducing the overall computational complexity.
Data Source
Figure 1~3
Figure 4~5
AI summary
The invention relates to a method of detecting an anomaly in a mechanical or electromechanical system, comprising: identifying by a processing device, on the basis of a set of spectral components that is generated on the basis of a signal originating from at least one sensor detecting vibrations and/or electrical fluctuations in the system, at least first and second spectral components, each component representing a harmonic or a modulation associated with a candidate component (Ci): by determining a first search window for identifying the first component; by identifying the first component on the basis of the first search window; by modifying the central frequency and the margin of error of the candidate component on the basis of a central frequency and of a margin of error of the first component; and by determining, on the basis of the modified central frequency (Vi) and of the modified margin of error (Ei), lower and upper frequency limits (Vi±Ei) of a second search window associated with the second component.