Efficient computation of cross-spectral densities in the stochastic modelling of waves and wave loads

A new method is presented for efficient calculation of auto- and cross-spectral densities in the stochastic modelling of ocean waves and wave loads. As part of the short-term response analyses, the method may contribute to more efficient long-term response prediction. Specifically the cross-spectral densities of the first order wave excitation forces are considered, but the method is straightforwardly generalized to other cross-spectral densities, e.g. for wave elevation, wave kinematics or second order loads. The method can be used with any choice of directional spreading function, but special attention is given to the commonly used cos-2s type directional distribution. In addition to the development of the new method, the traditional method using the trapezoidal rule for numerical quadrature is improved by developing an adaptive way of choosing the number of integration points. The accuracy of the adaptive method and the new method is investigated, revealing rapid convergence for both methods. However, the new method appears more robust as it avoids so-called spurious hat errors. When applied to two different pontoon type floating bridges the adaptive method and the new method both achieve a great improvement in computational effort compared to the traditional trapezoidal rule method. When the dimensions of the floating bridge increase, i.e. the number of pontoons and their relative distances increase, the new method is superior with respect to computation time.