Experimental investigation and prediction models for vibration-flow coupled scour around pile-bucket foundations

Abstract

The pile-bucket foundation, which integrates the benefits of monopile and bucket designs, offers enhanced bearing capacity and deformation control, making it suitable for large-capacity offshore wind turbines. However, its scour behavior under coupled vibration-flow remains poorly understood, posing challenges for design and stability. This study systematically investigates the effects of vibration amplitude, frequency, and flow velocity on scour depth around pile-bucket foundations under vibration-flow coupling. The results suggest that vibration amplitude may influence scour depth by promoting soil migration and compaction. Vibration frequency accelerates initial scour but stabilizes it in the later stage, with its impact depending on flow velocity. Flow velocity exhibits the strongest correlation with scour depth, and a vibration intensity factor was introduced to quantify the combined effects. Then, a physics-based exponential empirical formula was proposed, achieving prediction errors within 20%. Furthermore, four machine learning models were developed, among which the data-driven multilayer perceptron model achieved the highest prediction accuracy (R 2 = 0.9426) within the investigated experimental domain. The combination of empirical formulas with explicit physical interpretation and high-precision machine learning models provides a complementary framework for predicting scour depth around pile-bucket foundations under coupled vibration-flow conditions.