Optimization of submerged floating tunnel cross section based on parametric Bézier curves and hybrid backpropagation - genetic algorithm

Abstract The cross-section geometry of a submerged floating tunnel (SFT) has a large effect on hydrodynamic characteristics, structural behavior and service level, making the tunnel cross section the primary factor in optimizing efficiency. Minimizing the mean drag and the dynamic variability in the lift of the SFT cross section under bi-directional (i.e., tidal) flow has a dramatic impact on the reduction of structural displacements and mooring loads. Based on a parametric Bezier curve dynamically comprising the leading-edge radius, tunnel height and width to define the SFT geometry, a sensitivity analysis of the Bezier curve parameters for a fixed aspect ratio with prototype dimensions under uniform flow conditions was conducted by applying Computational Fluid Dynamics (CFD), and the pressure distribution around the SFT cross-section surface was analyzed. A theoretical method comprising the Karman vortex street parameters was employed to verify the CFD simulation results. In order to determine the SFT cross section with optimal hydrodynamic properties, the mean drag and Root Mean Square (RMS) lift coefficients were selected as optimization objectives, and four Bezier curve parameters were the input variables, in a neural network and genetic algorithm optimization process (a hybrid BP-GA structure), which is less likely to become trapped in local minima. The results show the optimal tunnel cross section has a mean drag and a RMS lift coefficient reduced by 0.9% and 6.3%, respectively, compared to the original CFD dataset.