Using a Machine Learning Method to Predict the Penetration Depth of a Gravity Corer

Abstract

The study of penetration depth of gravity piston samplers has an essential impact on sampling efficiency and instrument safety. This study focuses on predicting penetration depth based on the characteristic parameters of the sampled seafloor sediments and the sampler parameters. Although numerous studies of gravity corer penetration depth have been carried out, most have been based on the energy conservation equation, which considers a varying number of influencing factors. Furthermore, most research has focused on the same research idea of finding analytical solutions. The present study proposes a new approach to predicting gravity corer penetration depth based on a machine learning method that uses real sampling data from the sea and experimental data from a gravity sampling physical model for training and testing. Experimental results indicate that the machine learning model can accurately predict gravity corer penetration depth. Moreover, predictions were made for the same penetration conditions using the machine learning model and three other analytical solution models. Results show that the prediction accuracy of machine learning outperforms that of the analytical prediction model under various statistical rubrics. This study demonstrates the capacity of the proposed machine learning model and provides civil engineers with an effective tool to predict the penetration depth of gravity corers.