Advanced classification of sand particle size distribution with acoustic sensing and convolutional neural networks

In addressing the critical role of sand particle size in hydraulic property prediction and applications like groundwater management, this study introduces a novel method combining ultrasonic acoustic sensing and convolutional neural network (CNN) for sand particle size distribution (PSD) classification. Recognizing the limitations of traditional particle size measurements in capturing temporal and spatial changes, our framework proposes a continuous, non-destructive monitoring solution. Laboratory experiments to construct a diverse echo signal dataset reflecting median sand sizes yielded a CNN model with high accuracy for four sand types. This approach promises low-cost, efficient monitoring of sand deposits, showcasing potential for broader geotechnical applications.