Good morning Laureano,
It really depends on the goal of your model. Is it meant to compare against some standard, or are you trying to match experiment?
1. For matching to experiment, yes, a 3D model would work well, but there is definitely cost associated with this. Running a transient 3D will take a lot more computational time than your simple 2D model. Is it worth it? This is an engineering judgement you have to make.
2. If you are trying to show that this configuration meets some standard, then your 2D model will be useful. Why? Because it is conservative. This means if your 2D model meets the criteria then the real 3D case will too.
3. Depending on exactly the goals of the model, there may be a way to re-scale the input heat flux. I don't have a scaling factor in my head, nor do I have my copy of Incropera & DeWitt handy. But there may be something in there that could be useful.
Good work so far!
You are navigating away from the AIS Discovery experience