Advancing hull monitoring through physics-informed machine learning : towards a real-time approach

This paper introduces a novel approach to virtual hull monitoring for ships, focusing on timedomain simulations and integrating Machine Learning (ML). Conventional frequency domain methods in hull monitoring lack phase information of the waves, limiting their ability to address dynamic effects and nonlinearities like steep waves. To overcome these constraints, this study combines hydrodynamic simulations in the frequency domain with Finite Element (FE) simulations, producing quasi time-series data for a comprehensive understanding of structural responses. However, real-time application of these simulations is impractical. To address this, a Physics-Informed ML concept is employed, training an ML model with simulation data to predict the ship’s structural response in quasi real-time. The first findings demonstrate the feasibility of adapting this methodology to the time domain, offering a cost-effective solution for holistic ship structure monitoring. This innovative approach leverages the advantages of time-domain simulations and ML efficiency, addressing current limitations and paving the way for real-time hull monitoring.

DDC Class

620: Engineering

690: Building, Construction

Options

Advancing hull monitoring through physics-informed machine learning : towards a real-time approach