Computational fluid dynamics on 3D point set surfaces

Hassan Bouchiba,Simon Santoso,Jean-Emmanuel Deschaud,Luisa Rocha-Da-Silva,François Goulette,Thierry Coupez

doi:10.1016/j.jcpx.2020.100069

Abstract

Computational fluid dynamics (CFD) in many cases requires designing 3D models manually, which is a tedious task that requires specific skills. In this paper, we present a novel method for performing CFD directly on scanned 3D point clouds. The proposed method builds an anisotropic volumetric tetrahedral mesh adapted around a point-sampled surface, without an explicit surface reconstruction step. The surface is represented by a new extended implicit moving least squares (EIMLS) scalar representation that extends the definition of the function to the entire computational domain, which makes it possible for use in immersed boundary flow simulations. The workflow we present allows us to compute flows around point-sampled geometries automatically. It also gives a better control of the precision around the surface with a limited number of computational nodes, which is a critical issue in CFD.

Highlights

1.1 General frameworkIn many cases, numerical simulation of physical phenomena such as external aerodynamics, acoustics wave propagation, or heat transfer, requires a 3D model of the object
We introduced a new method for simulating flows around 3D point clouds acquired from real-world 3D scans
We demonstrated the ability of the proposed method to capture complex phenomena on various detailed 2D and 3D datasets

Summary

Introduction

1.1 General frameworkIn many cases, numerical simulation of physical phenomena such as external aerodynamics, acoustics wave propagation, or heat transfer, requires a 3D model of the object. For urban environments mobile mapping [Goulette et al, 2006] enables fast and accurate acquisition with vehicles driven at regular speed in traffic These techniques are an interesting alternative to manual modeling as they produce accurate high-resolution models. 3D scanning techniques never directly yield a watertight and manifold surface mesh of the scanned object or scene Instead, they generate a noisy and unevenly spaced set of points sampled on the object’s surface, which is called a point cloud. They generate a noisy and unevenly spaced set of points sampled on the object’s surface, which is called a point cloud It can be very dense, it does not hold any topological or connectivity information which makes it impossible to use as is, in typical numerical simulation frameworks. These models are usually massive as today’s scanners produce up to one million points per second

Objectives

Methods

Results

Conclusion