Viscoelastic floating membranes can be used as flexible wave breakers to protect coastal and offshore structures or as flexible wave energy converters. Despite their potential, the role of viscoelastic floating membranes in optimally harvesting or dissipating wave energy remains largely unexplored, particularly regarding how spatially varying material properties influence their performance. To address this gap, we develop an adjoint-based PDE-constrained optimization framework, built on a monolithic finite element formulation of the coupled fluid–structure interaction problem, to investigate and optimize the viscoelastic properties of floating membranes. This methodology enables a systematic optimization of design parameters such as the mass, tension, and damping, which govern the response of the membrane at different wave conditions. In this study we demonstrate that the proposed methodology allows for the optimization of homogeneous and inhomogeneous properties of membranes for different wave excitation frequencies, leading to significant improvements in energy absorption. The framework is implemented in Julia using the Gridap package ecosystem, which enables automatic differentiation of adjoints and avoids the need to derive complex adjoint formulations.

Computer-aided simulations are routinely used to predict a prototype's performance. High-fidelity physics-based simulators might be computationally expensive for design and optimization, spurring the development of cheap deep-learning surrogates. The resulting surrogates often struggle to generalize and predict novel scenarios beyond their training domain. We propose a two-stage methodology addressing the challenge of generalization. It employs physics-based simulators, supplemented with ordinary differential equations integrated into the recurrent architecture, to learn the intrinsic dynamics. The proposed approach captures the inherent causality and generalizes the dynamics irrespective of a data source. The presented numerical experiments encompass five fundamental structural engineering scenarios, including beams on Winkler foundations based on Euler-Bernoulli and Timoshenko theories, beams under moving loads, and catenary-pantograph interactions in railways. The proposed methodology outperforms conventional recurrent methods and remains invariant to data sources, showcasing its efficacy. Numerical experiments highlight its prospects for design optimization, predictive maintenance, and enhancing safety measures.