High-fidelity numerical modeling of dragonboat

Abstract

Numerical simulation is now a standard tool across industry and, increasingly, in sport, for evaluating and improving equipment and technique. This thesis develops a high-fidelity numerical model of one such application, dragon-boat paddling, a water sport practiced by millions of people worldwide. The stroke combines rapid paddle motion, blade entry and exit, free-surface deformation, and interaction with a translating hull, all of which are difficult to separate experimentally and have not previously been represented in a transient, free-surface computational model based on measured paddle motion. This study developed a numerical environment that converts field measurements into prescribed six-degree-of-freedom paddle kinematics and applies them in ISIS-CFD simulations of a standalone paddle, a bare hull, and a paddle–hull flow-interaction configuration. A representative periodic stroke was reconstructed by phase averaging selected inertial-measurement and paddle-force data, followed by numerical integration of the acceleration signals to obtain the paddle motion. The resulting trajectory was imposed using an arbitrary Lagrangian–Eulerian overset-grid formulation. Adaptive grid refinement concentrated cells near the free surface, blade, hull, and resolved flow structures. The simulations comprised viscous and inviscid paddle cases, multiple AGR-threshold studies for the paddle and oscillating-surge hull, a calm-water hull-resistance sweep from 7 to 13 km/h, and a paddle–hull case with a single paddle alongside the hull. The measured force history was used as a reference for comparison. However, the absolute vertical position of the trajectory relative to the free surface could not be independently determined from the available measurements. A sensitivity study was therefore conducted to assess its influence, after which the vertical position was calibrated based on agreement with the measured forces. Consequently, the assessment should be regarded as calibration-informed rather than as an independent validation. Future experiments should directly measure the free-surface position to reduce this calibration dependence and enable more rigorous independent validation. For the standalone paddle, viscosity changed the peak and root-mean-square blade loads by only 2.74% and 1.55%, respectively, indicating that pressure-dominated unsteady loading controls the integrated response for this prescribed stroke. The finest-grid paddle force had estimated grid-convergence uncertainties of 0.45% for the peak and 0.11% for the root-mean-square value. Hull resistance increased from 49.63 N at 7 km/h to 229.92 N at 13 km/h, while the frictional share decreased from 82.48% to 58.37% as wave-making became more important. Relative to the standalone inviscid paddle, the paddle–hull configuration increased peak and root-mean-square paddle force by 28.50% and 25.22%, respectively. Its propulsive impulse increased from 16.36 to 21.52 N·s and its paddle-in-water ratio from 42.97% to 47.61%. The paddle wake convected downstream into the hull-side flow region, providing qualitative evidence of hydrodynamic interaction. The prescribed blade kinematics were identical between the paddle-only and paddle–hull configurations. The latter exhibited a locally elevated free surface due to the hull-generated wave system, leading to greater effective blade immersion and, consequently, increased paddle force. The paddle motion governed the temporal resolution, with median time steps of 0.192 ms for the standalone paddle and 0.186 ms for the paddle–hull configuration. The paddle–hull simulation was more computationally expensive, primarily because adaptive refinement of the convecting paddle wake resulted in a 2.74-times larger mean cell count over the multiple-stroke simulations. In the present simulations, both paddle kinematics and hull surge motion were prescribed, providing a consistent framework for resolving the paddle-induced flow and its interaction with the hull. A separate appendix reports a proof-of-concept structural coupling between the CFD solver and the geometrically exact beam model, BeamDyn. The coupling is benchmarked against closed-form solutions and verified through an eleven-case kinematic and added-mass campaign. The study demonstrates its load- and motion-transfer capabilities without applying CFD-generated loads to a flexible paddle. The resulting framework provides a reproducible basis for comparing athlete techniques and establishes a foundation for future extensions toward independent validation, dynamically coupled hull surge driven by computed paddle-induced forces, and simulations involving multiple paddles.