Noise smoothing for structural vibration test signals using an improved wavelet thresholding technique

Area of Science:

• Structural Engineering
• Signal Processing
• Experimental Mechanics

Background:

• Noise in structural vibration tests compromises data reliability and accuracy.
• Traditional denoising methods like hard and soft thresholding have limitations.

Purpose of the Study:

• To present a discrete wavelet transform (DWT) approach for denoising structural vibration signals.
• To introduce an improved sigmoid function-based thresholding scheme to overcome limitations of traditional methods.

Main Methods:

• Utilized discrete wavelet transform (DWT) for signal denoising.
• Investigated wavelet type, decomposition level, thresholding method, and selection rules.
• Developed and applied a sigmoid function-based thresholding scheme.
• Validated the method using benchmark signals and a real-world shaking table experiment.

Main Results:

• The proposed DWT method with sigmoid thresholding effectively denoises signals with varying noise levels.
• Performance was evaluated using signal-to-noise ratio (SNR) and root-mean-square error (RMSE).
• The sigmoid thresholding scheme demonstrated superior denoising performance over traditional hard and soft thresholding.

Conclusions:

• The proposed DWT-based denoising method, particularly with the sigmoid thresholding scheme, offers a robust solution for improving structural vibration test data quality.
• This approach enhances the reliability and accuracy of experimental results in structural dynamics.