Digital Twin in Wind Assistant Propulsion for marittime transport
Please login to view abstract download link
In the present work, we investigate high dimensional parametrized fluid dynamics problem raising from optimization in real-time of wings of a Wind Assistant Propulsion for TransOceanic Cargo. To achieve such challenge, we adopt a surrogate model "Gaussian Process" , which uses a low-rank methods (e.g. samplets, pivoted-cholesky) for reducing the computational complexity. The trade-off of exploration of input design space, accuracy of surrogate model and variation with respect to the Pareto front of optimizer is achieved by using the Bayesian optimization coupled with constrained optimization. The results are validated on airfoils profile simulated with Xfoils, where the input parameters are defined by control points of airfoils, and on fluid dynamics of industrial configuration of Wind Assistant Propulstion TransOceanic Cargo. The project received funding from H2020 EuroHPC, simOcean.
