Fiber-Optic Vibration Detection Using Feature Expansion and Dimensionality Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fiber-optic vibration detection systems face challenges in accurately determining the time and spatial location of vibrations due to high detection error rates, inferior categorization performance, and interference from noise and nonlinear signal conversions, limiting their application in high-reliability fields.
Innovation Solution
A method and system that expand the feature vector of initial fiber-optic signal data, apply dimensionality reduction, and utilize a two-stage classification process involving primary and secondary classification to improve accuracy, incorporating feature expansion, dimensionality reduction, and logical rule-based corrections to enhance vibration detection and specification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional vibration detection technique using distributed fiber-optic sensor is employed, then the system is easy to install and has wide detection range, but the detection error rate is high and the accuracy in determining time and spatial location is poor
Solution Approach 1:
The patent transforms the one-dimensional signal amplitude analysis into multi-dimensional feature space analysis by extracting multiple signal characteristics (time domain, frequency domain, time-frequency domain features) and constructing feature vectors. This dimensional expansion enables more accurate classification and determination of vibration time and spatial location, resolving the contradiction between detection reliability and measurement precision.
Solution Approach 2:
The patent changes the detection parameters from simple amplitude threshold comparison to multi-parameter feature extraction and classification. By analyzing multiple signal parameters simultaneously (including wavelet transform coefficients, spectral features, and temporal characteristics), the system achieves higher accuracy in vibration detection and location determination while maintaining the reliability of distributed fiber-optic sensing.
2Ease of operation
If simple threshold-based classification is used, then the detection process is simple, but the distinction of complicated vibrations to categories is limited and wrong detection rate increases
Solution Approach 1:
The patent segments the vibration detection process into multiple stages: signal acquisition, multi-domain feature extraction, dimensionality reduction, and hierarchical classification. This segmentation allows the system to handle complicated vibration categories systematically while maintaining operational clarity through structured processing steps, resolving the contradiction between simplicity and accuracy.
Solution Approach 2:
The patent introduces dimensionality reduction techniques as an intermediary step between raw signal processing and final classification. This intermediary transformation converts high-dimensional feature vectors into reduced-dimensional representations that preserve essential vibration characteristics, enabling accurate categorization while simplifying the classification process and reducing computational complexity.
3Ease of manufacture
If logical model with predetermined simple vibration equations is applied, then the model is easy to establish, but it can only detect predetermined simple vibrations and involves large errors due to nonlinear conversions
Solution Approach 1:
The patent replaces static predetermined vibration equations with dynamic adaptive feature extraction and classification. The system adaptively extracts features from arbitrary vibration signals and uses machine learning classifiers to identify vibration types, enabling detection of diverse and unknown vibration patterns while maintaining ease of model establishment through data-driven approaches.
Solution Approach 2:
The patent substitutes the mechanical/mathematical logical model with a data-driven signal processing system. Instead of relying on predetermined physical equations that require multiple nonlinear conversions, the system directly processes fiber-optic sensor signals through feature extraction and pattern recognition, eliminating model establishment complexity while expanding vibration type coverage.
4Measurement precision
If feature expansion and dimensionality reduction are applied, then the classification accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies preliminary dimensionality reduction to feature vectors before classification, pre-processing the data to eliminate redundant dimensions and retain only the most discriminative features. This preliminary action reduces the computational burden of subsequent classification operations while preserving the accuracy benefits of feature expansion, effectively managing the trade-off between precision and complexity.
Data Source
AI summary
A method for detecting and specifying a vibration on the basis of a feature of a fiber-optic signal to determine a time and a spatial location of the present invention includes: Step 1 of acquiring a feature-expanded function vector and C-number of vibration categories by expanding a feature of initial data of a vibration signal from a distributed fiber-optic sensor; Step 2 of calculating a dimensionality reduction matrix based on the feature-expanded function vector; Step 3 of acquiring a dimensionality-reduced feature function by operating the dimensionality reduction matrix to the initial data and the feature-expanded function vector; Step 4 of acquiring a primary classification result of the vibration signal by performing a classification with reference to primary classification parameter acquired from a parameter database; and Step 5 of acquiring and outputting a secondary classification result of the vibration signal by performing removal of a wrong detection result and correction of a wrong classification result of the primary classification result.