Did you check the flow rate in your boundary at the end of the simulation to make sure it matches?
I'm not exactly sure what could cause an error like this, Flow Sim does obey the basic physical laws of mass and heat transfer including mass conservation. Do you have a large density gradient in the pipe? Is your flow compressible?
Yes, I checked it and the speed is still 20 m/s before and after the simulation. I'm working with air and i also remarked that the speed is 20 m/s at the point of the 9 arrows that are represented to show the boundary condition, but it differs in the other points of the section. Is there a way to change the "sampling", I mean to change the number (here 9) of the arrows where is set the speed in order to lower this error ?
Here is a picture to explain my problem (Sorry, it's a french version) :
I think I see the problem.
For the boundary condition you set the velocity on the surface to be exactly 20 m/s at every point, however you can see that the flow is still developing inside the pipe entrance.
You have a few options as I see it:
1. Extend the pipe length to be at least 10D long, and make sure you set the option in your boundary condition for "fully developed flow"
2. Instead of setting an outlet velocity, set a mass/volume outlet flow rate and allow the fluid to take its own profile.
In general when dealing with CFD models it's a wise idea to let any internal flows have enough length to fully develop before leaving the domain (even if it means extending your model beyond what you would build practically). Otherwise you end up with unnatural models such as this, and typically you will have vortices crossing your boundaries as well.