A data-driven methodology for wave time-series measurement on floating structures

This paper presents a data-driven model that utilizes artificial neural networks (ANNs) to estimate incident-wave elevation based on the motion and air-gap responses of floating structures. Accurate wave measurements ensure safe and optimized operation of offshore structures. These measurements can be obtained using both in-situ and remote sensing methods, each with its own challenges and limitations. Given these limitations, we proposed a data-driven approach that exploits the wave-induced motion responses of floating structures. Wave information could be extracted from the air-gap responses using the proposed method. An ANN model was trained using the scaled model test data and fine-tuned using Bayesian optimization. The proposed method demonstrated superior robustness in reconstructing the wave-surface elevation around a floating platform, outperforming the motion compensation algorithm and other data-driven models. The wave elevation calculated using the proposed method in both the time and frequency domains effectively minimized low-frequency errors owing to platform motion and those stemming from the wave-structure interaction. The statistical distribution characteristics also demonstrated the effectiveness of the method.