The paper below addresses some of your question. I've seen some additional work on this, but can't recall the authors - will have to search a bit.

http://is.gd/HN8gLN

At GSL we carry out standard grinding energy curves for all minerals giving Particle Size Distributions for a range of work inputs based on a stirred media mill.

The Bond test is not appropriate for fine grinding for several reasons. For one, the ball charge and tumbling motion of a Bond mill is inefficient in this range. Also, screening well at 500 mesh is not easy. Many people overlook the impact of screening on the Bond results but the impact can be significant.

The other issue is that nowadays fine grinding is not performed in tumbling mills for energy efficiency reasons. Depending on the size range a Vertimill (Metso), a stirred media detritor (Metso) or an IsaMill (Xstrata) is typically used. All three use a different motion than a tumbling mill and as a result each has their own laboratory test for scale up. In the case of the stirred media detritor and the IsaMill it is a direct scale up from the laboratory to the plant unit. For the Vertimill a specialized tumbling test is used (Jar Mill test).

I should also point out that if you are thinking of a regrind application then you cannot use the fresh feed (or core for a green-field project) as a test sample. You do need to perform a laboratory rougher flotation (or whatever is the separation step before regrind). The reason for this is that the separation step causes changes the mineralogy and there can be a corresponding change in hardness. As a result, the sample for testing is only available in small quantities, smaller than what Bond requires. The fine grinding tests mentioned above only require a modest amount of material.

I would like to add in range of ultra-fine grinding mills: vibrating ring mill and attrition mill with different types of balls-grinding media as well.

As mentioned above, screening efficiency is problem at 500 mesh. Dry sieving of ultra-fine powder material is very difficult, almost impossible. In this purpose, application of devices such as Coulter Multisizer for determination of PSD is possible, only.

All good answers so far. Many thanks for the responses. I put this out there in response to a client's client requesting a "fine grind" Bond test. I didn't think it was practical but you have all clinched it for me.

There have been several good papers on difficulties of performing Bond test even on coarser (normal) materials. Asa result people may want to consider the potential for errors (bias and excessive variability) even with 'normal' Bond testing and possible implications for mill design.

An argument can be made that the "Bond" exponent (sqrt-root 2) is not really usable down in that size range (See Hukki's 1962 paper). This was part of what Morrell was trying to address with his "Mi" approach (but I have no indication he intended to go to that size range; per everyone's comments about the practicality of doing the test).

Suggest it is best to stand your ground and say "don't go there". Jar mill or signatures plot probably your best alternatives.

A great deal can be said for matching media size to feed; new UFG milling processes take advantage of the efficiencies in this regard. I agree you are best conducting batch tests with mills more suited to this size band using smaller media than Bond Test. [Energy-based tests]

The paper you mentioned is just a marketing paper written objectively by Isamill claiming no one knows anything in the market but then there some basic errors in the way they have calculated OBWi.

Anybody has some papers related; I would like to watch other than Isamill paper

There are some good papers by authors from Draiswerke and Netzsch.

Stehr is one author who comes to mind, however, there are several others with good papers to take a look at. I'm traveling but will look for some of these to let you know.

The first experience I had with this technology was grinding limestone to 0.7 and 1.0 microns & researchers had done some good work around operating conditions, power consumption, etc. for producing these types of products.

I met one of the co-authors of the paper just after it was presented for the first time and he indicated that he was trying to take his name off the paper but it was too late anyway, there are at least five major flaws in this paper. Let's talk about two of them. The idea of assessing a grinding equipment efficiency using Operating BWi is to compare what mill is CURRENTLY doing (OBWi) against CURRENT ore or feed grind-ability characteristic ( or BWi or SE derived from lab tests) on page 11 of the paper they have calculated CURRENT OBWi which is alright so far but have compared it with what mill was sized based on years ago as they say "This was compiled by publicly available data (Vendor installation list)" most of the references in the table on page 11 is Cannington site where the Vertimills were installed in 1996 or 17 years ago. If you base your calculation on installation list then it will be misleading we sized Vertimill for a feed size of 150 um that time and now Cannington is feeding the mills with a feed size of 250 um. They are now virtually two different applications more over at Cannington, the feed comes from seven different ore bodies with different mineralogy and grindability characteristics so which one we are talking about and which one the vendor did the test work 17 years ago. This is the first time in my career I see someone calculate the grinding efficiency in this way and no one virtually questioned this paper.

At Cadia the first Vertmill sizing was based on a Bond Wi of 16 kwh/t ten years ago the ore BWi is now 18.5 kwh/t after ten years so if you turn up at site and compare what the Ball mill or Vertimill or SAG is doing now with what it was sized based on ten years then for sure they are all under-performing 15%.

Depends on what you want to learn. Check out research by Hukki on Bond vis-a-vis other methods. Am not sure if you mean control sieve or result one (bdp) or other?

In case the base for achieving the average WI for each material. I recently seen having a too fine control sieve first one, please look in deep the original paper from Bond, he already gives some suggestion and if not mistaking arrived to conclusion of what control size is worth for which material to test. Giving the can be misleading and giving too optimistic results.

Bond tests were conducted down to 38 microns and there was an increase of about 25% in the BWI vs. 106 microns. I would expect the 25 microns to be even higher. The problem is that how meaningful it can be when it is possible to decrease the energy consumption by more than 35% when using the proper ball size into an industrial mill! I will start some investigation by benchmarking actual grinding circuit. If you have any specific project, I will be glad to help since that we want to develop a fine grinding test.

Conducting a Bond test at a limiting screen of 25 microns implies that you are targeting a grind of - 80% passing 20 microns. Balls mills are very inefficient at that grind due to the coarser ball size used. Explore different fine milling technologies (Metso SMD, Isamill, Deswik) which have their own testwork.

The problem is with the 25 micron rating on the closing of the test circuit, a test as suggested here (Bond).

Do this:

Get the Conventional Bond Index (# 100) - teste de F.Bond;
Occupy a batch ball mill, wet, to get the same P80 obtained at Bond test and see the time used.
With similar sample, search the time required to obtain 25 microns.
Multiply your BWI by (T25 micron / TBond)

We use grind curves over a range of work inputs using stirred media mills to determine the kWh/t to achieve the desired final particle size distribution. This can be done down to the sub-micron sizes if required.