Little implementation detail


I see everywhere in every LBM code the derivation of the flow speed basically implemented like this :
u = sum(c_i*f_i) / sum(f_i)
I’m using LBM for model where particles concentrate on certain locations and desert others, so I face the problem of sum(f_i) ~= 0 for certain spatial locations, and by consequent speeds that can be of the order of unity (both numerator and denominator being near to zero), while it should be (near to) zero. So up to now I directly visualize the momentum flux, but I was wandering if there is a clever way to tackle this problem?