An exponential decay is y~e^-x rather than y~1/x^2. If you use an inverse square law to model blast damage you will have infinite bomb damage at zero radius (the whole dividing by zero thing).
For this case let our damage function be D(r)=D0 e^[-c r]: where D is the damage at radius r, D0 is the damage at radius 0 (the max damage of the bomb), c is some coefficient of decay that we have to determine from dropping a bomb and fitting the curve to our observations, and r is the radius.
The first thing that has to be done is to determine c. To do this we will take a 1000 lb bomb and drop it near something that takes a known amount of damage to destroy (town building = 250lbs if I remember right). So we make the drop and find that the maximum distance a 1000 lb bomb can land and take out a building is 100'. We now have all we need to know! We throw some algebra at it to give us c = -1/r Ln[D/D0], plug in the numbers, c = -1/100 Ln[250/1000] = 7.21*10^-3 (1/ft). Then we plug our new found c into our general formula and have a function of damage at a given radius for 1000 lb bombs. If things are modeled consistently in the game this function will hold for all bomb types. So for this example we would have
Damage = Bomb weight e^[(7.21*10^-3) r].
I made a nice plot of the function but I cant figure out how to post and image on here so the function and Idea will have to stand alone.
REMEMBER THIS IS AN EXAMPLE AND SOMEONE NEEDS TO MAKE A MEASUREMENT SO THE CURVE CAN BE PROPERLY FIT. So you just need a known bomb weight destroying an object of known hardness at a known maximum distance and then it can be fit.