# Can simpleFoam calculate accurate lift coefficient for vortex shedding of cylinder?

 Register Blogs Members List Search Today's Posts Mark Forums Read

 April 8, 2022, 06:19 Can simpleFoam calculate accurate lift coefficient for vortex shedding of cylinder? #1 Member   Bushra Rasheed Join Date: Dec 2020 Posts: 93 Rep Power: 3 Hi! For flow around cylinder at different reynold numbers, can simpleFoam give accurate lift coefficient ? For Re 200, the lift coefficient calculated by simpleFoam in my case is very low as compared to 0.7-0.8 in literature. I want to know how simpleFoam works for vortex shedding, why is it capturing vortex shedding with time even thoguh it's a steady solver and whether the lift coeffcient calculated through simpleFoam is accurate or not. I can't use icoFoam since it's taking too long. Please guide Thanks

 April 9, 2022, 09:38 #2 Senior Member   Join Date: Sep 2013 Posts: 324 Rep Power: 17 The flow around a cylinder leads to vortex shedding and hence non stationary flow starting from some Reynoldsnumber. Not sure of the exact number but from memory somewhere around 100. Small perturbations in basically any flow field will yields vortices or eddies that change and move in time. If you are solving a flow numerically however those might be damped out by your numerical setting. Think of this like blurring an image. SimpleFoam for example will still generate a solution and even converge to a low residual depending on your discretizatzion. If you use upwind for every scheme you'll generate lots of numerical diffusion. Hence this small time depended oscillations are damped away and you get a steady state solution. In this fashion you can generate a solution for maybe reynolds numbers up to 1000 or even higher still. You will basically average the time depend properties. What you are solving is hence unphysical since it should produce eddies but doesn't, but your pressure and friction forces on the cylinder might still be somewhat accurate. The problem however is that your flow in reality is fluctuating. A LOT. Hence you won't get a low residual, convergence etc on your steady state. Achieving that steady state will become more and more difficult. And simpleFoam won't converge even though it is a steady state solver. Hence you might see oscillations there which will yield those. Hence the only real way to calculate those is by using a transient solver like icoFoam, pimpleFoam etc. You can get drag forces that are somewhat accurate but you will not get a lift force that is accurate from a steady state solution. This is an extremely often encountered problem in CFD. You nearly always want to calculate a steady state because it is a billion times faster to compute, but your convergence will be bullshit. The flow through aircraft engines for example. Or the flow around cars. All of them are not a steady state. And if you want really accurate results you need to compute them in time, resolve all those eddies with LES and average their forces to get an accurate drag force. This will take weeks to months on thousands of cpu cores to achieve. A steady state however can be done within minutes to hours. B_R_Khan likes this.

June 6, 2022, 01:19
#3
Member

Bushra Rasheed
Join Date: Dec 2020
Posts: 93
Rep Power: 3
Quote:
 Originally Posted by Bloerb The flow around a cylinder leads to vortex shedding and hence non stationary flow starting from some Reynoldsnumber. Not sure of the exact number but from memory somewhere around 100. Small perturbations in basically any flow field will yields vortices or eddies that change and move in time. If you are solving a flow numerically however those might be damped out by your numerical setting. Think of this like blurring an image. SimpleFoam for example will still generate a solution and even converge to a low residual depending on your discretizatzion. If you use upwind for every scheme you'll generate lots of numerical diffusion. Hence this small time depended oscillations are damped away and you get a steady state solution. In this fashion you can generate a solution for maybe reynolds numbers up to 1000 or even higher still. You will basically average the time depend properties. What you are solving is hence unphysical since it should produce eddies but doesn't, but your pressure and friction forces on the cylinder might still be somewhat accurate. The problem however is that your flow in reality is fluctuating. A LOT. Hence you won't get a low residual, convergence etc on your steady state. Achieving that steady state will become more and more difficult. And simpleFoam won't converge even though it is a steady state solver. Hence you might see oscillations there which will yield those. Hence the only real way to calculate those is by using a transient solver like icoFoam, pimpleFoam etc. You can get drag forces that are somewhat accurate but you will not get a lift force that is accurate from a steady state solution. This is an extremely often encountered problem in CFD. You nearly always want to calculate a steady state because it is a billion times faster to compute, but your convergence will be bullshit. The flow through aircraft engines for example. Or the flow around cars. All of them are not a steady state. And if you want really accurate results you need to compute them in time, resolve all those eddies with LES and average their forces to get an accurate drag force. This will take weeks to months on thousands of cpu cores to achieve. A steady state however can be done within minutes to hours.
Thanks for this detailed explanation!