greybeard,

First, from what I've read about Silicon dioxide sand, the grains are relative spherical at least much more so that something like garnet. Secondly, and this is pragmatic to use fyffe555's euphemism for not thorough, everyone else doing packing studies seems to assume assumes spherical sand.

The paper you cited from NISS was random packing of spheres of different sizes. The 70% fill rate that they achieve with random packing is only 5% less than the packing of uniform sized identical spheres. They also state that as the size of the spheres added approaches 0 the packing approaches 100%. They got 70% stopping at an arbitrary size some ratio away from the biggest sphere. My gut instinct is that it will scale well to most semi-spherical shapes. Something I saw said the some types of packings actually work better with ellipsoids ?!

Gupta et. al in the paper brunog posted talk about something called binder skin theory and they talk about it accounting for the non spherical nature of the particles as well as some other effects. I have ordered reference 1 from Gupta which is a book of proceedings from the first conference on polymers in concrete, London, 1975 in hopes of it explaining what gupta cited it for.

Lastly, above all of the aggregate theory, we have the Reichhold aggregate formulation for polyester polymer concrete which is undoubtedly at least decent. I have a hunch that a formula based on the NISS paper will work better than one based on either Fuller's Formula or the Reichhold reference formulation. That being said, I'll have to track down brunog and buy him a beer if I'm wrong.

Ultimately, there are probably an uncountably infinite number of good aggregate sizing strategies. Fuller's Formula, a formula based on the NISS paper, and the Reichhold formula are but three. Fuller's formula has 100 years of testing in ordinary concrete. I don't have an implementable formula from the NISS paper yet. Finally, the Reichhold formulation was recommended to help them sell resins so it was obviously thoroughly tested.

One must remember that randomness is not something to fear in engineering. It can be controlled and often used to advantage. The niss paper when extended down to the tiniest particles actually approaches 100% packing. Dense packed spheres of any single size do not and I imagine that Dense packet spheres of any number of uniform sizes also may very well not.
Randomness in particle size should essentially cancel out randomness in particle placement.

For those who are interested in getting machines out the door pronto, I'd suggest considering the formula for the aggregate in old style polymer concrete posted here http://www.cnczone.com/forums/showpo...postcount=1201.

I'd also recommend writing up what you want in terms of percent retained on different sieve sizes rather than just winging it and calling a specialty sand manufacturer such as www.agsco.com to sieve you a batch. If we use custom sand mixes prepared by somebody with the right sieves, we will have a better chance of comparing results and duplicating published formulas.