Abstract: Protected vehicles remain vulnerable to near-field underbody blasts, which can cause severe floor deformation and transmit high-amplitude transient loads. Under constraints on installation space and structural mass, a protective component must combine load distribution, structural support, and energy dissipation. This study proposes a composite protective component incorporating aluminum-foam-filled U-shaped stiffeners. Longitudinal, transverse, and U-shaped beams form a load-bearing framework that spreads localized blast loads, while aluminum foam inside the U-shaped cavities dissipates energy through progressive crushing. The configuration therefore integrates load-bearing, load-transfer, and energy-absorption functions without requiring a separate continuous sandwich core. A coupled soil-air-explosive-structure finite element model was established to evaluate the blast response. Blast tests were conducted to validate the model using the residual plastic deformation of the base plate; the difference between the experimental and numerical results was approximately 2%. The initial component was also compared with an equal-mass steel plate under identical loading and boundary conditions. Optimal Latin hypercube sampling generated samples across a 17-dimensional design space. Spearman correlation and Sobol global sensitivity analyses identified the dominant variables and reduced the design space to nine variables. A Bayesian additive regression trees (BART) surrogate model was then constructed and benchmarked against a Kriging model. The BART model achieved higher overall prediction accuracy for the nonlinear structural responses considered. A q-expected hypervolume improvement (qEHVI)-based multi-objective Bayesian optimization framework was subsequently developed. The maximum central displacement and maximum kinetic energy of the base plate, as well as the structural mass, were minimized, while structural mass was constrained not to exceed that of the initial design. Starting from 60 high-fidelity samples, ten batch iterations with four candidates per iteration produced a dataset of 100 simulations and an approximation of the Pareto-optimal solution set. A representative compromise solution was selected and re-evaluated using the high-fidelity model. Compared with the initial design, the optimized component reduced the maximum central displacement, residual plastic deformation, and maximum kinetic energy of the base plate by approximately 60%, 15%, and 20%, respectively. These results indicate that the proposed component and optimization framework can improve underbody blast protection under a prescribed mass constraint, providing a reference for the design and multi-objective optimization of vehicle underbody protective structures.