CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Scaling up a wave energy converter - free surface flow (

mark_l February 17, 2010 07:42

Scaling up a wave energy converter - free surface flow
3 Attachment(s)
This is my first post on here and I'm fairly new to CFD so hopefully I've covered all the relevant information in this problem.

I am simulating a wave energy converter that uses free surface water waves to capture and compress air. We have carried out extensive experimental work on the device and have been replicating this in CFX to a reasonable degree.

I initially modelled the converter at the same scale as that tested experimentally (about 1m long). However, I am now interested in looking at how the performance of this device scales up. To do this, I began by applying a scale factor (100x to achieve ~full scale) to both the geometry and mesh in ICEM. I scaled up the boundary conditions in pre by their equivalent scale factors. The timestep was kept the same and the total simulation time was increased from 8sec to 80sec. This model would not run as it failed at the first iteration. I have attached the .out file (100_scale.out)

To check that the scaling in ICEM wasn't the problem, I created a new pre simulation and imported the previous 100x scaled mesh but set the import units to cm rather than m, to give the original 1m long model. I ran this with the small scale boundary conditions and it worked fine. (see 100_unscaled.out)

I then began to scale the model down from x100 to a point where it worked. x50 and x30 both failed in the same way as x100, but x10 worked fine and has generated reasonable results (see 10_scale.out)

I have checked the maximum water velocities at the wave generating boundary and this is only about 2.5m/s for the x100 scaled conditions with a 10sec 5m wave. I also tried running the x100 model with smaller waves but to no avail.

A summary of my conditions (these are pretty much the only values that have changed over the 3 simulations)

Small Scale

Wave Period = 1.03 sec Wave Height = 0.050 m Water Depth = 1 m Timestep = 0.01 sec Total Time = 8 sec


Wave Period = 3.257 sec Wave Height = 0.5 m Water Depth = 10 m Timestep = 0.02 sec Total Time = 20 sec


Wave Period = 10.3 sec wave Height = 5.0 m Water Depth = 100 m Timestep = 0.01 sec Total Time = 80 sec

I am cannot see why this shouldn't run. The only logical explaination I can come up with is that CFX doesn't like the mesh size at towards the bottom of the domain. The coarsest regions have roughly a max of 18m between nodes, whereas at the free surface, node spacing is about 0.15m. This order of spacing works fine at the smaller scale though. Would I be better off trying a slightly finer mesh? (trouble is I'm already running 1.2m nodes)

I'd be really grateful if anyone could give me any suggestions as I've been trying to sort this issue for over a week now and people are starting to want some results!

Thanks in advance


ghorrocks February 17, 2010 08:07

I have not looked at it in detail but your assumption that the timestep is constant is not valid. Generally the best idea is to adjust the timestep until you have 3-5 coefficient loops per iteration - although for a multiphase model you might want to increase this to 4-7 or something.

mark_l February 17, 2010 08:14

Thanks Glenn, I'll have a look at this now, but would that necessarily cause problems when scaling, even if it was an adequate assumption previously?

ghorrocks February 17, 2010 17:57

If you are having convergence problems then CFD is a random number generator. It could tell you anything.

All times are GMT -4. The time now is 23:40.