[rjm1982]

Hello, I am simulating a pitching and heaving airfoil at Re=500000 using Spalart Allmaras (I have read other similar studies that have recommended this turbulence model for this case). I am using smoothing and remeshing to achieve the rotational and translating motion, with a UDF to acheive the required motion.

My problem is that the lift coefficient time series contains very pronounced spikes. Does anyone have any idea what could be causing this? I have tried a few things to fix it (reducing the body force under-relaxation factor, reducing time step, improving mesh resolution, trying different solvers (coupled and SIMPLE)) but none of them worked and I am now a bit lost for ideas.

My suspicion is that it is related to the spring smoothing system I have applied for the dynamic mesh smoothing, or the remeshing. I am using a triangular mesh (it is a 2D case) and I have tried diffusion smoothing as well but this also gave a strange time series with lots of wiggles. Although I have read that diffusion should be better for cases with lots of rotation, I have also read that it is not well suited to trianglar meshes, and indeed the quality of the mesh becomes very poor quite quickly using this approach. Therefore I would like to stick with spring smoothing, which preserves mesh quality and gives a good lift force except for the spikes.

If anyone has any experience of similar spiky time series I would be very grateful for any advice you could offer.

Richard

[rjm1982]

Just an update on this: my airfoil is located within a circular sub-domain that has radius 2.5c, where c is the chord length of the airfoil. I have noticed that when I make the sub-domain extremely large (radius 8c), the spikes in the lift force time series disappear. Therefore I suppose the spikes occur when a particular remeshing operation occurs in the wake of the airfoil, somehow causing a temporary disturbance in the pressure field and producing a spike in the lift force.

Making the roatating domain larger effectively moves the remeshing operation away from more critical zones, and the lift force calculation is therefore not affected.

Unfortunately this solution is not useful for me as, due to the geometry of my problem, my rotating domain really has to be no larger than about 2.5 chord lengths. If anyone has any thoughts on how to solve this I would be very grateful to hear them.

[rjm1982]

Another update on this problem in case anyone is interested:

The spikes turned out to be related to parallelisation in Fluent. I had been running in parallel on 4 CPUs but when I changed to series the spikes miraculously disappeared! In fact I ran another longer simulation in parallel and after quite a few oscillation cycles the spikes stopped appearing. So, one way or another, I think I will be able to live with the spikes for now.

[Emanuel Camacho]

Hi. Have you tried adding compiler directives to your UDF?

Compiler derivatives: http://www.afs.enea.it/project/neptu...df/node219.htm

[coffee_break]

I am having same issue with my descending airfoil near ground(see figure below). Spikes in Cl accumulates in time, they get bigger and then they fade away as shown in figure below. Airfoil descends with constant velocity, only layering and smoothing done on mesh. During these spikes occur convergence is not achieved for each time step.
Problem definition: https://drive.google.com/file/d/16uX...ew?usp=sharing
Cl spikes: https://drive.google.com/file/d/1GdK...ew?usp=sharing