Machine-learning optimisation of piggy-backed plate anchors using LDFE simulations

Piggy-backed plate anchor systems have emerged as an effective solution for mooring floating offshore structures in deep-water environments. However, their embedment behaviour involves complex soil–structure interaction and strong coupling between the front and rear anchors, making design optimisation computationally and experimentally demanding. This study proposes a machine-learning-assisted optimisation framework to enhance the design efficiency and performance of piggy-backed plate anchors. A database is first established using large-deformation finite element simulations based on the coupled Eulerian–Lagrangian method. Gaussian process regression surrogate models are then developed to capture the non-linear mapping between key design parameters and the resulting embedment depths. Sensitivity analysis reveals that the front anchor’s depth is primarily governed by its own shank angle, while the rear anchor’s depth is highly sensitive to anchor spacing due to soil disturbance effects. Multi-objective optimisation using the differential evolution algorithm reveals a clear trade-off between the two anchors, forming a Pareto front. The results demonstrate that a larger rear shank angle is beneficial, while optimal spacing varies significantly depending on whether the design priority is the front anchor or the total system depth. The proposed framework provides an efficient and reliable tool for designing offshore anchoring systems.