Abstract
System modeling tools for wave energy converters (WECs), such as WEC-Sim, require hydrodynamic coefficients to predict the behavior of the system in response to waves. These coefficients can be found through experiments, high-fidelity simulations such as computational fluid dynamics (CFD) or smoothed particle hydrodynamics (SPH), and most commonly, boundary element methods (BEM). BEM is a computationally efficient way to calculate these values, using panel methods that only require spatial discretization on the bounding surface. BEM is based on linear potential flow theory, which has assumptions (e.g., linear, incompressible, inviscid, irrotational flow with small amplitude motions) that speed up simulation time, but may decrease the accuracy of the hydrodynamic estimates in some conditions. A number of commercial and open-source BEM packages (e.g., WAMIT, Ansys Aqwa, Capytaine, and Nemoh) have been developed, with WAMIT and Aqwa considered industry standard. Despite the inherent limitations, results from these BEM codes are widely used in time-domain modeling with little regard for software variability. In particular, few studies have investigated the variation in performance between BEM software packages for different WEC shapes. In this work, we aim to characterize the hydrodynamic coefficients predicted by BEM for WEC floats across the design space, and assess the impact of geometric variation by comparing the performance of three BEM programs (Capytaine, Aqwa, and WAMIT) against experimental results. In doing so, we will determine benefits and shortfalls for each code for their use in system modeling of WECs.
Performance for each software package is defined by sensitivity to mesh variations and variation from experimental results. To control for mesh structure, the same meshes are used in each program. A mesh sensitivity study using both the classical grid convergence index and least-squares grid convergence index developed by Eça et al. is performed for each program to evaluate the sensitivity to mesh variations, estimate the uncertainty from the mesh, and determine the best mesh for experimental comparison. With a sufficiently refined mesh and an estimation of uncertainty for each program, we can define hydrodynamic parameters for each geometry from each BEM program. Validity of BEM for each geometry is determined by the coherence of hydrodynamic parameters between programs and comparison to experimentally determined values. Specifically, the added mass and radiation damping coefficients from BEM are validated with experimental data from forced oscillation tests in heave. These initial tests use the Sandia “WaveBot” geometry to benchmark accuracy of our methods against prior results. We show that all three BEM programs can accurately calculate the hydrodynamic coefficients for the WaveBot geometry with low variance. Initial results indicate that WAMIT is more dependent on mesh structure than Aqwa, but more resilient to other numerical changes, such as solver type and mesh refinement. Subsequent tests involve geometries of comparable size to the benchmark case that sample more of the WEC design space to identify variations in hydrodynamic response and model-measurement accuracy. Overall, this study offers insight into when BEM provides sufficient accuracy for WEC modeling and reveals important sensitivities across codes.