Evolution mechanisms and model-based prediction of directional pressure distribution in underwater explosion shock waves from cylindrical charges

The geometric shape of a charge critically influences the spatial distribution of underwater explosion shock wave loads. While recent studies have investigated the basic characteristics of non-spherical charges, quantitative boundaries for shape effects and efficient omnidirectional prediction methods remain challenges in engineering applications. This study systematically investigates the spatiotemporal evolution of shock waves from cylindrical charges using integrated experiments and numerical simulations. Distinct from conventional axial-radial analysis, this work quantitatively characterizes the directional evolution of shock waves, revealing that the high-pressure zone concentrates within the radial sector (typically 60°–152°), while explicit "Enhancement" and "Diminishment" zones are mapped across the full spectrum. A key contribution of this research is the establishment of a critical distance criterion ( L c ), formulated as L c / r c = 1.96 + 12.83 λ , which delineates the operational domain where charge shape effects must be considered. Furthermore, to address the limitation of existing empirical formulas, a novel shape factor is proposed and embedded into a deep neural network (DNN). This physics-informed data-driven model achieves high-precision prediction (error < 10%) of full-field shock wave pressure for charges with varying length-to-diameter ratios (1–10). This work provides theoretical insights into directional evolution and offers a practical, rapid assessment tool for the blast-resistant design of marine structures. • Reveals the spatiotemporal evolution mechanisms of shock waves from cylindrical charges. • Maps the “Enhancement” and “Diminishment” zones across the full 0°–180° spectrum. • Establishes a quantitative critical distance criterion ( L c ) to delineate shape effect boundaries. • Develops a physics-informed deep learning model for rapid, omnidirectional pressure prediction. • Achieves high-precision prediction (error < 10%) for varying aspect ratios and standoff distances.