Denoising Acoustic Emission Signal using Empirical Mode Decomposition Method

Abstract

Acoustic emission (AE) plays an important role in Structural Health Monitoring (SHM) applications by providing the early-stage damage assessment of composite materials.However, the collection of AE signals is challenging due to complex noise arising from the mechanical equipment, temperature, vibration, friction as well as external and internal environments of the structure.In order to overcome this challenge, even though many denoising methods have been introduced to acquire the denoised AE signals, there is still a lack of effectiveness in denoising without degrading the originality of the AE signals.Therefore, this paper adopts an efficient denoising method named Empirical Mode Decomposition (EMD) to remove most of the noises of the acquired AE signals by keeping its original properties.The adopted method is initially utilised on synthetic datasets which are randomly generated inducing various levels of Gaussian white noise.The obtained results are then compared to the original properties of the randomly generated clean dataset to evaluate the effectiveness of the EMD method.Experiments have been carried out to acquire the AE signals added with friction and vibration noises and then the EMD method is applied to them to eliminate the noises.The performance of the EMD method has been evaluated based on different performance metrics.Results show that the EMD method effectively removes most of the noises without disrupting original properties of the AE signals.