Dr. Guo's Article "A Deep Learning P-Wave Arrival Picker for Laboratory Acoustic Emissions: Model Training and Its Performance" Published

Dr. Guo, Dr. Vanorio, and Dr. Ding are excited to share their recent publication, “A Deep-Learning P-Wave Arrival Picker for Laboratory Acoustic Emissions: Model Training and Its Performance,” in Rock Mechanics and Rock Engineering.

"The acoustic emission (AE) technique has been widely used in laboratory deformation and mechanical tests on geomaterials such as rocks, concrete, and asphalt" Dr. Guo says. "Accurate P-wave arrival picking is essential for advanced AE data analysis, including localizing AE events, constructing P-wave velocity models, and determining the focal mechanisms of AE events.

"In this study, we introduce a deep-learning model, AE-PNet, designed to efficiently pick P-wave arrivals of large AE datasets in laboratory settings—an area currently underserved in rock mechanics research. We investigated the optimal training dataset size for AE-PNet, which is crucial for improving its reliability and addressing the limitations of open AE datasets constrained by the time-intensive nature of manual picking. We demonstrated that AE-PNet trained by the minimum number (i.e. 1500) of manually picked waveforms significantly outperforms Akaike Information Criterion (AIC), a classical picking algorithm, showing its great potential to pick P-wave arrivals accurately and efficiently.

"To further support research on deep-learning-based P-wave arrival pickers for AE applications, we have publicly made approximately 50,000 manually picked (labeled) AE waveforms available to the community."