Abstract
This study compares three methods for estimating quadratic transfer functions (QTFs) of low-frequency hydrodynamic loads on floating structures from experiments: Cross-Bi-Spectral (CBS) analysis, Regression-Based (RB) estimation, and Harmonic Probing of Nonlinear Auto-Regressive eXogenous models (NARX-HP). All methods are applied to experimental data from model tests of the INO WINDMOOR 12 MW semi-submersible, subjected to various sea states. The QTFs are derived to capture the nonlinear, low-frequency wave forces critical to mooring design and floater response. Each method’s performance is benchmarked against measured surge displacements from four validation experiments realized from a JONSWAP sea state. Results show that the RB method achieves the lowest normalized root-mean-square error, followed by NARX-HP and CBS. While all methods effectively capture frequency content, RB and NARX-HP achieved the best extreme response predictions. The study discusses each method’s theoretical basis, computational demands, and practical applicability, highlighting efficiency and accuracy for offshore engineering applications.