AUTHOR=Xiong Haibei , Chen Lin , Yuan Cheng , Kong Qingzhao TITLE=A Novel Piezoceramic-Based Sensing Technology Combined With Visual Domain Networks for Timber Damage Quantification JOURNAL=Frontiers in Materials VOLUME=Volume 8 - 2021 YEAR=2021 URL=https://www.frontiersin.org/journals/materials/articles/10.3389/fmats.2021.688594 DOI=10.3389/fmats.2021.688594 ISSN=2296-8016 ABSTRACT=Crack detection for timber structures has gained increasing attraction as the demand for using wood materials in building constructions continue to grow. A reliable detection method for crack severity, especially for timber beams which are easily to brittle fracture due to cracking, is essential to ensure the structural safety. Inspired by the auto-classification of image, this article proposes a detection method for crack on timber beams using stress wave-based sensing combined with computer vision technique. The framework of the proposed approach consists of three steps: spectrogram visualization by time-frequency analysis, data augmentation by considering environmental resistance of datasets, and the training of deep neural networks for spectrogram images. Numerical simulation is conducted to illustrate the propagation property of stress wave crossing through the timber crack within different depths. Laboratory tests for three timber beam specimens with seven crack depth cases in each are conducted. Both the numerical simulation and experiments utilize the surface-mounted piezoelectric transducer to excite stress wave, and both results show attenuation of the stress wave energy is more intensive as crack depth increasing. The classification results of deep neural network show high accuracy with spectrums produced by different crack depth cases, validating the validity of the investigated approach to identify crack severity. The proposed method has perspective to provide fist determination or real-time detection of crack for timer structures in field applications.