SimScale CAE Forum

What is wrong with my propeller simulation?


I’m quite new to CFD and from tutorials and documentation I found on this site, I attempted to simulate a very simple propeller for a project that I am doing. Here it is below.

In the end, I want to measure the force that the propeller generates.

At first, I attempted to do some MRF simulations, but the force that the propeller generates keeps on increasing as time increases to numbers which seem impossible. Could someone help me to find out if I have made an error which is causing this?


Hi @sdixit,

A couple of possible fixes I can recommend.

Check the position of the rotation center. Do ensure it is at the very center of your propeller assembly.

Try disabling the “Initialize with potential flow” under simulation control and see if that helps stop the increase in lift.

You can reduce the Relaxation factor for field p, Relaxation factor for equation U, Relaxation factor for equation k and Relaxation factor for equation omega by 0.2 each to try to improve simulation stability at the cost of increased simulation time.

Do keep the gradient schemes under numerics and numerical solver to guass linear as leastsquares may cause simulation instability. Generally we should aim to get the simulation running with first order schemes first before moving on to higher order schemes.

Try changing the solver for velocity, Turbulent kinetic energy solver, Specific turbulence dissipation solver to smooth solver if all the above fixes do not work.

Ensure you implement these changes individually as if we do identify the problem, you would want to know which setting is causing the problem.





It is happening most probably due to pressure inlet b.c. There is a high chance of occurence of discrepancies owing to the pressure inlet b.c. as delta_p basically depends upon the geometry and the flow, which is to be calculated.
Recommendaion: Use velocity inlet



Thanks for the help.
I tried out a few different settings according to the suggestion, but I figured it out, the only problem was that the gradient scheme in numerics should be gauss linear instead of leastsquares.