AeroDoodle is a wind tunnel simulation. In the simulation window, fluid flows from left to right. In areas of high velocity the colour of the simulation window is red. In areas of low velocity, the colour is blue.
When you draw an object in the simulation window, the fluid is not able to penetrate it. Additionally, any fluid in contact with the obstacle must have a velocity of zero, this is known as the no-slip condition.
As the fluid flows, it exerts forces upon the obstacle. These forces are commonly known as lift and drag. Lift acts in a direction perpendicular to the freestream flow and the drag acts in the direction of the freestream flow.
Simply knowing the value of these forces is not particularly useful when comparing different geometries. This is because the forces are proportional to the size of the object. A much larger object will generate much larger forces, however it may not generate them as efficiently. The size of an object is accounted for if coefficients of lift and drag are compared instead
Lift coefficient = Lift/qA
Drag Coefficient = Drag/qA
Where q is the dynamic pressure of the oncoming flow (q=0.5xdensityxvelocity^2) and A is a representative area of the geometry under investigation.
The flow of fluid is described by the Navier Stokes equations, which describe the relationship between velocity, pressure and density. Solutions to this set of equations exist for simple cases such as,
However no general solution exists which describes fluid flow over arbitrarily shaped objects.
Computational fluid dynamics, or CFD, allows us to study such flows by using numerical techniques to approximate solutions the the Navier Stokes equations. These numerical techniques typically involve subdividing space into a grid of small continuum elements. For each element an approximation of density, pressure, and velocity is acheived through a series of arithmetic operations.
For an engineer today, a number of CFD techniques are available. The most mature of these are the Finite Volume and Finite Element methods. More recently developed techniques include Smoothed Particle Hydrodynamics and the Lattice Boltzmann Method. Each of these methods has their relative benefits, AeroDoodle uses the Lattice Boltzmann method because it is a naturarally parallel algorithm.
For every second of a AeroDoodle simulation, approximately 300 million equations are evaluated. Simulations of the Bloodhound SSC far exceed this.
The term high performance computing, or HPC, refers to a field of computer science which seeks to evaluate these equation in the most efficient way possible, leading to reductions in the time required for a simulation to complete. For CFD simulations the most effective way to do this is to compute the solution for each continuum element in parallel.
The AeroDoodle simulation is divided into a grid of continuum elements which is 1024 elements wide, and 128 elements high. For each frame of the simulation, these elements are distributed accross the many computational cores of your computers graphics chip. The graphics chip is used in this way because it typically has many more cores than a CPU.