Signal Denoising Using Stochastic Resonance for Acoustic Emission-Based Structural Health Monitoring

Noise is unavoidable and ever-present in measurements. As a result, signal denoising is a necessity for many scientific and engineering disciplines. In particular, structural health monitoring applications aim to detect weak anomaly responses generated by incipient damages from background noise that contaminates the signals. Among various approaches, stochastic resonance has been widely studied and adopted for denoising and weak signal detection to enhance the reliability of structural heath monitoring applications. On the other hand, many of the advancements have been focused on detecting useful information from the frequency domain generally in a post-processing environment, such as identifying damage-induced frequency changes that become more prominent by utilizing stochastic resonance in bistable systems, rather than recovering the original time domain responses. In this study, a new adaptive signal conditioning strategy is presented for on-line signal denoising and recovery. The core of the new approach is utilizing the stochastic resonance in a bistable dynamic system to average out the noise-induced stochastic transitions. The input amplitude to the bistable system is adaptively adjusted to favorably activate the stochastic resonance based on the noise level of the given signal, which is one of the few quantities that can be assessed from noise contaminated signals in practical situations. Numerical investigations demonstrate the operational principle and confirm the denoising performance of the new method. Experimental verification with denoising and recovering acoustic emission signals by employing a double-well Duffing analog circuit exemplifies the promising potential of implementing the new denoising strategy for enhancing on-line acoustic emission-based structural health monitoring.