As far as I am aware the Minnovex Group were the first services group to look at estimating changes in milling performance for different oretypes. I believe it is their framework that was the reference used in the thesis abstract by Keeny (Williams and Richardson - both these authors being SGS).

The geomet. approach is emerging, and possibly in the next 10 years we will see many new groups. What I specifically liked about the Minnovex (now SGS) approach was the integration of metallurgical modelling with geostatistics.  I would certainly appreciate comment about what other groups are now adopting the approach of variability studies using a metallurgical performance basis - not just a testwork basis.

We used textural and geological data in Lead-Zinc Ores at Mount Isa to determine ore processing performance.

It was reported in:

Bojcevski, D, Vink, L, Johnson, N W, Landmark, V, Mackenzie, J, Young, M F, 1998, Metallurgical Characterization of George Fisher Ore Textures and Implications for Ore Processing. AusIMM Mine to Mill Conference, 11-14 October, Brisbane, Australia pp 29-41

The AusIMM Mine to Mill Conference in 1998 was a good snap shot of Geometallurgy at the time (before the Geometallurgy term was being used) and the directions that were being used and investigated.

Also further progress was reported in the 2003 AusIMM Mill Operators Conference on how it was integrated into the geological block model.

A large amount of work has been done in the milling circuit at LKAB for different oretypes, and reported in the current AMIRA P9O project by Prof. Malcolm Powell. As you are a sponsor you should be able to download this data from the AMIRA web site.

The type of geomet I was referring to, was about building metallurgical performance models into the geological block models to improve cut-off grade determination and maximize the NSR (net smelter return) of the ore being mined over a short term (weekly) or long term period.(annually/LOM).

The methods that we used were detailed in

Young M. F., Johnston M. and Simpson J., 2003, Metallurgical Performance Models to improve Cut-Off Grade Determination at Mount Isa Mines. AusIMM Eight Mill Operators Conference, Townsville, Australia, 22-23 July, pp177-184

The on-line methods you seem to be referring to are similar to those investigated in the AMIRA projects P483 & P483A - Optimization of Mine Fragmentation for Downstream Processing 1996-2002

We used the outcomes of this project at the Mount Isa Copper Concentrator, which used the on-line Dynamic SAG Mill model, developed by Walter Valery at the JKMRC, and on-line feed belt particle size measurement, allowing us to change the feed quality/particle size to maximize the production of the grinding circuit.

Similar methods were also successfully used at KGCM in Western Australia during the project, and also at other sponsors operations.

Walter Valery now runs the Process Technology and Innovation Group at Metso Minerals in Brisbane, and his group implements these tools in plants to optimize the operations.

I applied conditional simulation and other geostatistical methods to the ore texture data on the George Fisher deposit at Mt Isa in the mid 1990s. David Bojevski was the metallurgist at the time, Laurence Vink the Mineralogist, and Dale Simms the geologist. I believe that this was the first time microscopic ore textures (as an estimate of metallurgical performance) were incorporated into block models for use in mine planning. As then and for the vast majority of geometallurgical data it was critical to deal with the non-linearity of the data. Traditional geostatisitical methods applied directly to non-linear data such as recovery, hardness and mill throughput have caused significant problems when the resultant block models have been used to assist mill design and undertake financial evaluation.

Interesting comment about geostatistical applying for non linear data. This is a recurrent topic in copper industry. In my experience if the effects of this sinergic effects, positive or negative, has a non-relevant impact, or if the plant performance are in a "linear zone" of the non linear function, geostatistics interpolators it could be applied with excellent results. Mainly if this geometallurgical parameters like energy consumption or metal recovery will be scaled to the industrial behaviour using adjusting factors or simulators(CEET, JKSiMMet or other similar tool).

Related to geometallurgical estimating models using geostatistical tools, my first experience was starting 2000's with a Work Index model using ordinary kriging. This work was published in IV International Mineral Processing Worshop in 2006 with the name "Modelo Gometalúrgico de Work Index en División Andina de Codelco Chile". Recently I have been involved in Axb estimation modelling using conditional simulation to be used in JKSimmMet simulators. In my opinion is a more refined way to estimate this kind of variables with the plus of the determination of uncertainly of estimation.

I would suggest a correct measure of hardness for geostatistical modelling should have units: Energy per unit tonne (to grind particles to a particular size).

With respect to A*B I think you will find the units are tonnes per unit energy.

Hence 1/(A*B) is more additive.

As an aside, being at JKMRC for 17 years I always found it difficult to accept the disingenuous descriptor 'A*B' to represent tonnage per unit energy.

All physicists are familiar with the concept of using an appropriate symbol:

i.e.
g for gravity
N for Newton
J for Joules

But in comminution they use 'A*B' for Hardness?

I am of the strong view that the total lack of effort in providing appropriate descriptors contributes significantly to confusion in the mining industry- and I could write a treatise on this.

I totally agree that with regarding the use of 'recovery' in geostatistics. It should be linked to a model. The word 'recovery' generally has implied meanings not disclosed to the reader/audience.

For example, I once listed to a seminar on 'increasing recovery in flotation'. After half an hour 'I said but if you wanted to increase recovery just send everything to the con'. The reaction I received to my remark was not complimentary - but only because the speaker would not clearly say what their constraints we eat the outset. However I maintain my argument that 'recovery' actually implies 'margin' i.e. value returned from processing. So why not simply use 'margin' that has direct economic meaning rather than 'recovery' which could mean the results of any of: laboratory test, pilot plant test, or plant simulation.

If margin has units 'profit per unit tonne' then it is additive. Terribly simple to work with.

A summary presentation with some thoughts from Scandinavia - "Lamberg, P., Geometallurgy - what, why, and how?, 8th Fennoscandian Expl. and Mining, FEM 2011, Levi, Finland." Provides some thoughts on an approach to geometallurgy. It would have been interesting to hear the talk.