Integrated Framework for Real-time Sea State Estimation of Stationary Marine Units Using Wave Buoy Analogy

Abstract

Understanding the impact of environmental factors, particularly seaway, on marine units is critical for developing efficient control and decision support systems. Despite extensive research on the wave buoy analogy (WBA), real-time sea state estimation (SSE) has remained challenging due to the need for adequate data to construct a consistent probability density function of waves. The current study builds on previous work, aiming to propose an AI framework to reduce the estimation time lag between exciting waves and respective estimation by transforming temporal/spectral features into a manipulated scalogram form for subsequent data analysis. To achieve this, an adaptive ship response predictor and deep learning approach have been incorporated to classify seaway data while minimizing network complexity through feature engineering. Specifically, a Convolution Neural Network (CNN) and deep transfer learning have been employed for training purposes. The system's performance was evaluated using data obtained from an experimental test on a semi-submersible platform, and the results demonstrate the promising functionality of the proposed approach for a fully automated system. Furthermore, the deficits with the feature transformation of the existing data-driven SSE models have been discussed. This study provides a foundation for improving online sea state estimation and promoting the seaway acquisition for stationary marine units.