Data-driven uncertainty-aware seakeeping prediction of the Delft 372 catamaran using ensemble Hankel dynamic mode decomposition

2025-11-06

AI Summary

Accurately predicting motion responses—particularly heave and pitch—of high-speed catamarans (e.g., Delft 372 model) in irregular waves remains challenging due to strong nonlinearities, memory effects, and inherent uncertainties. Method: We propose a data-driven, equation-free linear reduced-order modeling framework: Hankel Dynamic Mode Decomposition with control inputs (HDMDc), which explicitly embeds time-lagged states and external wave excitations to capture nonlinearity and temporal memory. To enhance robustness, we further develop a dual-path ensemble strategy—FHDMDc—integrating Bayesian and frequentist uncertainty quantification. Contribution/Results: Validated against high-fidelity URANS simulations, FHDMDc achieves significant improvements in probabilistic prediction accuracy of motion response probability density functions while maintaining low computational cost. The uncertainty estimates are reliable and physically consistent. This work establishes an interpretable, generalizable modeling paradigm for rapid seakeeping assessment and intelligent decision-making in high-performance marine vehicles.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

In this study, we present and validate an ensemble-based Hankel Dynamic Mode Decomposition with control (HDMDc) for uncertainty-aware seakeeping predictions of a high-speed catamaran, namely the Delft 372 model. Experimental measurements (time histories) of wave elevation at the longitudinal center of gravity, heave, pitch, notional flight-deck velocity, notional bridge acceleration, and total resistance were collected from irregular wave basin tests on a 1:33.3 scale replica of the Delft 372 model under sea state 5 conditions at Fr = 0.425, and organized into training, validation, and test sets. The HDMDc algorithm constructs an equation-free linear reduced-order model of the seakeeping vessel by augmenting states and inputs with their time-lagged copies to capture nonlinear and memory effects. Two ensembling strategies, namely Bayesian HDMDc (BHDMDc), which samples hyperparameters considered stochastic variables with prior distribution to produce posterior mean forecasts with confidence intervals, and Frequentist HDMDc (FHDMDc), which aggregates multiple model obtained over data subsets, are compared in providing seakeeping prediction and uncertainty quantification. The FHDMDc approach is found to improve the accuracy of the predictions compared to the deterministic counterpart, also providing robust uncertainty estimation; whereas the application of BHDMDc to the present test case is not found beneficial in comparison to the deterministic model. FHDMDc-derived probability density functions for the motions closely match both experimental data and URANS results, demonstrating reliable and computationally efficient seakeeping prediction for design and operational support.

Problem

• Predicts seakeeping performance of catamaran with uncertainty quantification
• Compares Bayesian and Frequentist ensemble methods for motion forecasting
• Develops data-driven reduced-order model using dynamic mode decomposition

Innovation

• Ensemble Hankel DMD with control for seakeeping
• Bayesian and Frequentist methods for uncertainty quantification
• Equation-free linear reduced-order model from time-lagged data