Marine coral clay, a critical fine-grained component in reclaimed island foundations for ocean engineering, is typically mixed with marine coral sand to form composite foundation soil that governs offshore infrastructure stability. To address accurate strength prediction of this soil reinforced by 3D-printed bionic honeycomb polymer grid (BHPG), this study develops a CNN-LSTM model, uses SHapley Additive exPlanations (SHAP) to quantify input parameter importance, and validates it with 1200 triaxial shear tests. Results confirm high accuracy and identify reinforcement type, layers, and confining pressure as key factors, while a derived empirical formula enables rapid engineering use. A user-friendly graphical user interface (GUI) is also developed for ocean engineering practitioners to get real-time strength predictions. This work reduces test costs, advances deep learning-marine engineering integration, and supports BHPG application in reclaimed islands and offshore platforms. • Established a CNN-LSTM deep learning model for MCCM strength prediction. • Integrated spatial and temporal feature learning to enhance prediction accuracy. • Applied SHAP analysis to quantify input parameter importance in the model. • Derived an interpretable empirical formula from model results for engineering use. • Developed a GUI enabling real-time deep learning prediction for marine projects.
Xu et al. (Thu,) studied this question.