A procedure for the numerical simulation of underwater radiated noise from machineries with real data validation

Underwater radiated noise (URN) from ships is of growing environmental concern, but reliable prediction is limited by scarce full-scale validation data and by a research focus skewed toward propeller noise over machinery-induced contributions. Moreover, the vibro-acoustic tools used in practice differ substantially in their acoustic domain formulation, yet are seldom compared under controlled, validated conditions. This work addresses both gaps through a numerical investigation of machinery-induced URN using two approaches, VA One (hybrid FEM–BEM, with explicit free-surface and seabed boundary conditions) and Actran (FEM with an infinite acoustic domain), applied to a common structural finite element model with machinery excitation applied through a resilient-mount representation. Predictions are validated against BURNSI benchmark measurements (ORCA-class vessel) at two hydrophone locations; using the same structural model and excitation for both solvers isolates the role of the acoustic domain formulation, which is the central novelty of the comparison. VA One achieves mean absolute errors of about 5–7 dB across the two hydrophones. Actran correctly reproduces the dominant spectral peaks at both sensors but underestimates SPL by an average of about 20 dB at one sensor, against about 7 dB at the other, due to its free-field representation, which excludes seabed reflections. VA One and Actran thus prove complementary rather than competing: the former better suited to environmental noise assessment in shallow water, the latter to source characterization under idealized free-field conditions.