Hi everyone,

I'm currently using FDS 3 from NIST to analyse fire behaviour in buildings. I have no experience with other cfd-packages and I am somewhat of a novice concerning cfd.

With FDS it's possible to stretch or compress the geometry-grid at numerous points. This way I can compress the grid at places I have wall-sections and other tight spots and stretch it on long walls and places where there's nothing. This gives me accurate geometry while keeping the grid relatively small and thus giving me faster calculation times on my pc.

My question is if these grid-alterations affect the end results of the calculation. It will undoubtly lower the resolution of the end-result, but will it affect the outcome at a given gridpoint? In other words: at one given gridpoint, does it make a difference in outcome whether there are a lot of gridpoints or just a few near it? If so, is the difference worth the extra calculation time?

Thanks, Danny

"In other words: at one given gridpoint, does it make a difference in outcome whether there are a lot of gridpoints or just a few near it?"

Definitely. You might want to read about resolution issues in some of the modern texts - the younger folks on this forum will have recommendations. Not only does the distance between the points impact the solution at a particular node, but the spacing ratio (or mesh growth ratio) also makes a difference.

"If so, is the difference worth the extra calculation time?"

This is harder to answer. You're getting into economics big time here. The cost of your time needs to be weighed against the impact of the accuracy of your solution - is a 50% error acceptable? How about 10%? How many solutions will you need for different sets of initial and boundary conditions?

I would be inclined to run several meshes varying the total number of nodes as well as the spacing density and compare the solutions very carefully. You'll not only get a hint at the answer to your questions, but you'll learn a lot about the code you're using. If this is too expensive, you should be very modest about the validity of your solution.