Gradient-Based Optimization Framework for Mass Distribution of Floating Wave Energy Converter Hulls

Dataset Overview
This MATLAB suite optimizes the internal mass distribution and inertial properties of floating wave energy converter (WEC) hulls. The gradient-based framework employs Sequential Quadratic Programming (SQP) to tune uncoupled natural frequencies in heave and/or pitch towards resonance with a target sea state whilst strictly maintaining hydrostatic stability constraints. The optimization output provides the design mass distribution and inertial properties under three modes: (i) material- and construction-method-agnostic, for calculating a theoretical mass distribution; (ii) optimized for modular precast construction, for example, for a hull comprising precast ultra-high-performance concrete elements; or (iii) optimized for thin-shell hulls with solid ballast, for example, for a hull fabricated from conventional marine-grade steel.

This dataset accompanies the research paper (linked below):
Kotzamanis, V., Kalliontzis, D. (2026). Multimodal wave energy conversion: Modal coupling effects and power capture, Renewable Energy, 274: 126185. https://doi.org/10.1016/j.renene.2026.126185.

Methodology
The framework executes across a three-stage optimization sequence. Stage 1 (2D Optimization) employs SQP over an iteratively trained 2D surrogate model of the 3D geometry to establish an initial density distribution, enabling rapid multi-start convergence. Stage 2 (3D Optimization) refines this density field across the full 3D parametric hull surface using SQP, initialized by the Stage 1 solution, to yield the unconstrained theoretical mass distribution. Stage 3 (Structural Realization) maps the converged 3D density field onto the selected fabrication archetype, self-correcting mass balance discrepancies between the theoretical and physical designs to output final heave and pitch natural periods, rigid-body mass properties, and structural fabrication schedules.

Included Files
The repository is organized into source code, documentation, and benchmark validation cases. Source code includes MATLAB scripts (.m) for parsing Boundary Element Method (BEM) hydrodynamic outputs, executing the runtime optimization suite, and post-processing or plotting results. Documentation comprises a README detailing repository architecture, a user guide specifying input file formats and execution procedures, and a structured codebase map designed for navigation. The example case provides parametric geometry files (.ms2) for a multimodal WEC (MWEC), associated BEM hydrodynamic databases, pre-computed optimization results, validation plots, and run logs.

Prerequisites & System Dependencies
Execution requires MATLAB R2022b or later with the Optimization Toolbox. Hydrodynamic inputs require frequency-dependent coefficients (added mass, radiation damping, and wave excitation force vectors) calculated via linear potential flow BEM solvers, such as WAMIT, HAMS-MREL, or Capytaine. Native parsers are provided for WAMIT and HAMS-MREL data formats, with specific formatting documented in the technical guide. The suite also supports automated runtime coupling with HAMS-MREL v2.2 on Linux platforms; deployment across other operating systems may require local source recompilation and environment configuration.

Limitations & Assumptions
Hydrodynamic pre-sizing is evaluated strictly under unmoored, free-floating conditions, neglecting active power take-off (PTO) damping and mooring restoring stiffness. The formulation relies on linear potential flow theory, omitting viscous drag, flow separation, and non-linear hydrodynamic forces. Geometry must be parametrically defined to calculate mass properties and drive the density optimization. For automated HAMS-MREL panelization, irregular frequencies are suppressed via a waterplane lid; however, complete suppression may degrade for geometries with waterplane-parallel surfaces at certain drafts, requiring manual inspection of added mass and damping curves to exclude numerical artifacts. The kinematics assume planar 3-DOF motion (surge, heave, pitch) decoupled from lateral modes via geometric symmetry. Single-DOF heave tuning is achieved by setting a zero weighting factor for the pitch natural period in the objective function, whereas adapting the suite to roll-dominated devices requires manually modifying the natural period calculations and hydrodynamic parsing scripts.