Sensor Signal Denoising with Adaptive SSA Window Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current noise removal methods for industrial signals, particularly in laser welding processes, face challenges such as imperfect orthogonality in Empirical Mode Decomposition (EMD) and complexity in selecting filtering parameters, leading to suboptimal noise separation and energy leakage.
Innovation Solution
A modified Singular Spectrum Analysis (SSA) method is introduced, involving iterative single value decomposition with adaptive window size selection and energy tracking to achieve orthogonal intrinsic mode functions, ensuring effective noise removal and defect identification in industrial processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Empirical Mode Decomposition (EMD) is used for noise removal, then the signal can be decomposed into intrinsic mode functions, but the orthogonality of IMFs is imperfect leading to energy leakage
Solution Approach 1:
The patent applies Gram-Schmidt orthogonalization to transform the non-orthogonal intrinsic mode functions (IMFs) obtained from EMD into orthogonal intrinsic mode functions (OIMFs). This parameter transformation changes the mathematical properties of the decomposition components, ensuring perfect orthogonality and eliminating energy leakage while preserving the noise removal capability.
2Measurement precision
If traditional Singular Spectrum_analysis is used, then noise removal can be performed, but the computational complexity increases and parameter selection becomes difficult
Solution Approach 1:
The patent segments the signal processing task into two distinct stages: first applying EMD to decompose the signal into IMFs, then applying Gram-Schmidt orthogonalization only to the resulting IMFs. This segmentation avoids the computational complexity of traditional Singular Spectrum Analysis while achieving orthogonal decomposition and effective noise separation.
3Measurement precision
If iterative SSA with adaptive window size is used, then noise removal effectiveness is improved, but the processing time and computational load increase
Solution Approach 1:
The patent applies Gram-Schmidt orthogonalization selectively to the intrinsic mode functions obtained from EMD, rather than performing exhaustive iterative processing on the entire signal. This partial action achieves the necessary orthogonality for effective noise removal without the excessive computational time and processing load of iterative SSA with adaptive window sizing.
Data Source
AI summary
A method for performing a noise removal operation includes decomposing an acquired signal considered as one dimensional series. A trajectory matrix is constructed, transforming the trajectory matrix in a form to which single value decomposition is applicable. A single value decomposition is done on the transformed matrix computing eigenvalues and eigenvectors of the matrix. A one dimensional series is reconstructed, corresponding to the denoised signal. After the single value decomposition operation is provided, a single value decomposition is applied sequentially starting from a given window value. For each iteration, the root mean square value is calculated between a current and previous eigenvalue, calculating a minimum and its position of said root mean square value. The iterations are halted if the minimum is lower than a determined threshold value, otherwise increasing the window value and returning to the operation of decomposition of the acquired signal.