Shape optimization of box-type structures under waves using multi-objective Bayesian Optimization: Reducing wave loads and improving wave attenuation

Box-type structures are widely used as coastal infrastructure - including protective structures, like box-type breakwaters, providing protection to coastal communities, and non-protective structures, such as box-girder coastal bridges. The design of such structures is challenged by increasing intensity of coastal hazards: box-type breakwaters need to provide better wave attenuation performance to enhance coastal resilience while box-girder coastal bridges should be better designed to ensure their safety during wave strikes. Studies have demonstrated the importance of shapes of box-type structures in wave-structure interaction processes, and some shape optimization frameworks have been proposed to automate the process of identifying efficient structural shapes under wave loads. However, existing frameworks are mostly based on simplified wave theory, limiting their ability to capture complex wave-structure interactions. Moreover, most existing frameworks only address single-objective problems. To better address real-world design challenges of box-type structures in the face of wave loads, a novel methodology integrating Smoothed Particle Hydrodynamics (SPH) and Multi-Objective Bayesian Optimization (MOBO) is proposed and used to solve shape optimization problems of box-type coastal infrastructure under wave loads. The methodology is first validated and then applied to both single- and multi-objective shape optimization problems. The optimal design using the MOBO-SPH strategy improves wave attenuation performance by 60% compared to a traditional rectangular-shaped breakwater in the single-objective problem. In the multi-objective optimization problem, a set of optimal solutions are efficiently found that all outperform the traditional rectangular-shaped design in two different design objectives: reducing wave forces experienced by the structure while improving the wave attenuation performance. To reduce computational cost, parallelization scheme is used: in the numerical example, parallelization saves 73.1% of time without sacrificing the quality of optimization. • Multi-objective shape optimization framework proposed for box-type structures under wave loads. SPH simulations used. • Bayesian Optimization used for efficient sampling to reduce computational time and to improve optimization quality. • Pareto Frontier of multi-objective optimization problems efficiently found by the proposed method. • Support high-performance computing platforms and parallelization to reduce computational costs.