基于导向滤波的流固耦合边界数据插值分析

Interpolation Analysis of Fluid-structure Interaction Boundary Data Based on Guided Filtering

考虑到径向基函数插值进行流固耦合边界数据传递不受网格点拓扑关系的约束,尤其适用于CFD数值计算.针对插值数据中异常值会导致插值精度下降问题,采用图像滤波对CFD插值数据异常值进行滤波处理,结合高斯滤波、双边滤波和导向滤波各自特点,将导向滤波与已用于CFD平台的移动最小二乘去噪算法进行对比分析.仿真结果表明:三种滤波方法中,导向滤波的整体效果最为理想,且比移动最小二乘算法更适合于连续物理场数据去噪.将其应用于不同边界点密度情况下的流固耦合边界数据传递计算,显著提高了基于径向基函数插值的边界数据传递精度.

Considering that the radial basis function interpolation for fluid-solid coupling boundary data transmission is not restricted by the topological relationship of grid points,it is especially suitable for CFD numerical calculation.Aiming at the problem that the interpolation accuracy will be reduced due to the outliers in the interpolation data,image filtering was used to filter the outliers in CFD interpolation data.Combining the characteristics of Gaussian filtering,bilateral filtering and guided filtering,the guided filtering was compared with the moving least square denoising algorithm that has been used in CFD platform.The simulation results show that among the three filtering methods,the overall effect of guided filtering is the best,and it is more suitable for continuous physical field data denoising than the moving least square algorithm.Guided filtering is applied to the calculation of fluid-solid coupling boundary data transfer under different boundary point densities,which significantly improves the accuracy of boundary data transfer based on radial basis function interpolation.