Abstract: The bilge structure of a ship's hull is a typical vulnerable area in resisting underwater explosions, and the rapid assessment of its dynamic response and damage is crucial for ship survivability design. To overcome the high computational cost bottleneck of traditional finite element methods, this paper proposes an integrated assessment and prediction framework that combines parametric modeling and machine learning. First, a fluid-structure interaction numerical model based on the Coupled Eulerian-Lagrangian (CEL) method was established and validated through experimental data. Subsequently, parametric modeling of the bilge grillage was conducted, systematically considering six key variables such as curvature radius, stiffener spacing, and plate thickness, resulting in a dataset of 2304 working conditions. Based on this, a two-stage machine learning prediction model was developed: initially, a random forest algorithm was employed to classify deformation and rupture damage modes; then, a conditional generative adversarial network (CGAN) was used to establish an end-to-end mapping relationship from design parameters to full-field stress, strain, and deformation contour maps. The research results indicate that this integrated framework achieves an order-of-magnitude improvement in computational efficiency while maintaining high prediction accuracy, thereby providing an efficient and reliable new approach for the rapid assessment of underwater explosion damage and the optimization of explosion resistance performance in ship hull structures.