Part B: Innovative Data Analysis Approach to Boost Machine Learning for Seakeeping Purposes; Computational Efficiency

The increasing influence of AI across various scientific domains has prompted engineering to embark on new explorations for complex challenges. However, recent studies often overlook the foundational aspects of the maritime field, leading to over-optimistic or oversimplified outputs for real-world application purposes. To address these concerns, in previous works, we highlighted the sensitivity of trained models to noise, computational efficiency, and the necessity for feature engineering/compactness in seakeeping given the stochastic nature of the wave. In response, a novel data analysis framework was introduced with two purposes to augment data for machine learning (ML) models in seakeeping: transferring features from high-fidelity to low-fidelity modelling which resulted in improving feature accuracy for simulation and increasing computational efficiency through feature engineering; the current issue addresses the second objectives. To this end, the experimental data of a spherical model in a wave basin excited by different waves has been employed. The response time series of the model underwent a continuous wavelet transform to extract fine spectral-temporal features. After feature reorganization based on coefficient intensity in a new feature map, additional endogenous features are attached to the arrays to improve feature variability and entropy. Different ML models were trained, where the new data analysis framework substantially reduced training costs while maintaining fair accuracy, with training times slashed from hours to seconds. The significance of the current study extends beyond the maritime context and can be utilised for ML applications in intrinsically stochastic data.