I've worked on a lot of projects where we used the cut whole-diameter PQ core for the Bond low-energy impact crushing test. One instance where the core was crushed instead of loaded into the test resulted in quite different values (lower results than whole-diameter) which we interpreted as the crushing pre-stressing the rock creating fractures. A paper in the SAG 2011 conference demonstrates how we recommend doing the test. http://is.gd/0CcREF

The Bond crushing work index test procedure is not terribly well described, and different laboratories have different procedures that can result in different results (by up to -50% +100%). I am not aware of the calibration of the test being checked in over 50 years, so it is impossible to say which procedure is "correct". It is a noisy test and the industry just has to live with it.

Since a rock must be in "cubical shape" before conducting the Bond crushability work Index, it is better to cut it instead of crushing.

It is also important to make sure that the laboratory conducting the Bond crushability Work index test is using the standard Bond equipment and calculations. As reported by C. Bailey et al (2009) in their paper what can go wrong in comminution circuit design, a modified design of the machine can generate lower values of CWI for competent rocks.

I have seen round robin results of Bond crushability work index conducted between different laboratories on - 30 rocks of the same material each laboratory and the average results were within 30%.

CWi is a noisy test, and I think that Bailey et al mistook that noise for errors in laboratory procedure and equipment. Their analysis is based on comparison of CWi with the drop weight index which is a fundamentally flawed analysis. I can duplicate the discrepancy they reported on the same machine with the same ores by just rotating a sample by 90° when inserting it into the apparatus.

The 30% difference is within the -50% +100% variation I mentioned above. It is troubling, but it is not significant. That is why I would like to see a better test for generating work index determinations in +100 mm sizes, but I don't see anything on the horizon that looks useful.

For Crushing Work Index you need twenty rock specimens sized at -76+51 mm. The suggested core types are PQ core or large lump ore. Approximately 15 kg of sample is needed to conduct CWI test. This is the requirement for SGS, Amtec (Now ALS) and Amdel (Now Bureau Veritas) laboratories.

Recently I was dealing with a crusher vendor and they were also reluctant to rely on the crusher work index results we had from the client side. I suppose all the sides involved (the labs, equipment suppliers and the consultants) need to agree on one procedure or one test work to make things easier.

The situation is actually worse than that. The formula that underlay’s the crushing work index hasn't been calibrated since Fred Bond did it in the 1950's (probably based on data collected in the 1940's). Modern crushing machines, liner designs and the compression forces involved look almost nothing like the machines that Bond calibrated his formula against.

Moreover, Hukki (SME 1962) estimated that the size range where Bond's -½ exponent was valid is in the size range of rod and ball milling (not crushing); and Millard observed in MetPlant 2002 that the AMCT results don't match Bond's exponent, either. The good news is that SAG sizing formulae, like the ones in http://is.gd/8PSK9A (calibrated to Andean ores), aren't that sensitive to CWi. The effect of all this uncertainty is minimal when I'm designing a SAG mill for the Andes.

There is compression of a bed particles in crushers since crushers are typically choke feed in comparison to impact used for the Bond Crushability testwork. The Bond Crushability testwork do not also use the particle size distribution of the product to relate accurately the energy consumption to the size distribution obtained. There is therefore room for improvement/recalibration of the testwork.

From the SMC test, S. Morrel derives a Mic parameter (which is correlated to the DWi) which can also be used to determine crushers power requirement. In a near future, as database is build up around the world, we will have a comparison between both methods.

There are two obvious flaws with the Mic method (and a third that is less obvious).

First is that the Mic test is based on SMC drop weight tests conducted on particles about 25-30 mm nominal dimension where compression failure of rock matrix is observed? The breakage that happens in crushers (for my purposes) happens in the 50-1000 mm size range where rock failure due to existing fractures dominates. There is no relationship between a rock matrix and its fractures in the Andean ores I tend to work with (see Doll & Barratt, Procemin 2009). The Mic is measuring the wrong the breakage mechanism for my purposes.

Second is related to the first, the relationship between coarse, fracture-based breakage and medium scale, matrix-based breakage is calibrated through a proprietary SMC database. So the validity of the Mic is dictated by how "similar" your ore is to the SMC database. Andean and Australian ores are very dissimilar (fractured Mesozoic to Quaternary mountains v. competent Proterozoic shield), so I doubt it works for me; South African and Australian can be similar (Quaternary age shield rock, both competent), so it may work in ZA.

The third, non-obvious, issue with SMC method is the exponent on the size parameter. If you predicted that the exponent of coarser rocks should tend to zero, based on his interpretation of the Kick model. The SMC exponent tends to negative infinity at coarse sizes. I can accept that either situation might be true, but do not have an opinion at this time.

As more data are generated, correction factors are added to original equations to improve the prediction between actual power consumption and predicted.

Rowland did a good job with correction factors to improve the Ball mills power prediction using the Bond Ball Work Index test results. As you have mentioned, the calibration of the Bond crushability test is overdue. Are you working on it? As it stand the Bond method is also not perfect but crushers are sometimes "forgiving" although there are many installations were the crushing capacity was under designed.

The original equation published by S. morel in 2007 to predict power requirement for crushers was revised in 2009 to include a "coarse particle" ore hardness parameter SC to improve the prediction. I expect more correction factors to be introduced as he is increasing the database. It is nevertheless a good starting point especially when the Bond crushability test cannot be conducted due to sample requirement (size and/or amount).

I have some data of CWI and Mic on the same ore from different regions and plant data. It helps to understand the limitations of different methods.

I see a clear regional bias between the South American and Australian labs with the North American labs falling somewhere in the middle. These trends could be due to lab procedures, but my gut says this is actually the breakage character of the rocks in each region.

We start building some CWi data in our lab of rocks from Indonesia. Do you aware of any the research on crushability that I can contribute by sharing the data?

The Mic approach of Morrell is not “flawed”. It is an empirically based technique which has been calibrated with data ranging from primary crushers down to tertiary crushers and is the only technique that is currently available that has data to back up its validity. If indeed there is a structural component that influences rock strength in the 50-1000mm size range (and this is pure speculation not based on any data that are currently available), then blasting most likely will activate those structural weaknesses as the blast shock wave propagates through the in-situ rock, leaving the feed to the primary crusher relieved of these weaknesses. The “issue” that the SMC Test uses a nominal 25-30mm particle then becomes irrelevant.

Hukki’s work is purely theoretical and the fact that he postulated at coarser sizes that the size exponent should tend to zero and Morrell’s does not is irrelevant. Our objective as designers is to get the best answer possible and that means using techniques that can be demonstrated to work, which is what Morrell’s technique provides. In fact Morrell’s exponent equals -1 at 700mm and only tends to – infinity with infinitely large rocks – a situation which is not worth contemplating. A further point to note is that rocks do not fail due to compression in most comminution machine but fail under tension. They may well be loaded under compressive forces but failure is due to tensile stressing.

The Bond crushing work index test is an inherently poor test and has been shown time and time again in Round Robin studies to have poor precision (repeatability). Angove and Dunne’s paper on the subject gives results that show factors of over 2 in terms of reported work index values from the same sample from different labs. Coupled with the fact that there is no published data on how well the test actually predicts real crusher performance would suggest we should not be using the test in the first place. We certainly as an industry should not “live with it”.

Cut whole-diameter core cannot be used for the Bond crushing work index test. Bond’s procedure is well described by Bond as are the detailed specs of the machine that should be used (Bond, 1946, Crushing Tests by Pressure and Impact, Mining Technology). He specified quite clearly that rock specimens must be selected such that they pass a 3”square sieve and are retained on a 2”square sieve.

As PQ core has a nominal diameter of 85mm it is not possible to satisfy this criterion. You will therefore have to use crushed material, even though in the act of crushing the core you may affect its apparent strength.

Dr Cameron Brigg’s 1997 PhD thesis from the JKMRC “A Fundamental Model of a Cone Crusher” is particularly enlightening on this subject. He measured the strength of rocks after various degrees of crushing using a Hopkinson pressure bar. There was no uniform pattern. With some rocks the apparent strength increased after successive stages of crushing, whilst in others it did not and in one case the apparent strength reduced.

Arguably we should be crushing the core in the first instance as we are interested in the rock strength so as to estimate the specific energy of crushers. With the exception of primary crushers, all crushers have feed that has been previously crushed and hence crushing the PQ core mimics to some extent what the feed will experience in real life.

A further point worth mentioning is that the Bond crushing test is not valid on rock sizes other than 2-3”, though it is clear that often tests are done with other size fractions (usually smaller particles). The reason is that Bond’s laboratory equation is of the form WI=2.59C/sg, where C is the input energy per unit thickness, rather than per unit volume. As the units of WI are meant to be kWh/t, the fact that C has the units of energy per unit thickness creates an imbalance in the units across the WI equation and hence a bias in the equation for all rock sizes that are outside of Bond’s specified range.

A lot of work in this area was done about 10-15years ago by Oreway Mineral Consultants who did a large number of Bond crushing work index tests on a wide range of rock sizes using a wide range of different rock types. Their results showed convincingly that if smaller rock sizes are used the apparent Bond crushing work index tends to decrease and if larger sizes are used the apparent work index increases, in line with the mathematical error in the Wi equation and contrary to physics and most other rock strength tests which show that hardness decreases as rock size increases. 

We did crush the PQ instead of cut. Some samples only have less than 10 pieces of larger than 2 inches.

I've said my bit on Bond Wi for crushing so will focus on the exponent in Mi formula and the about Hukki v. Morrell disagreeing.

Where Hukki focused his work was in fine grinding through to rod/ball milling. The fine grinding exponents determined by Hukki largely match values we now see in 'signature plots' for Isamills and Vertimills, though the exponents vary from material to material. As a gross generalisation, Isamill signature plots have exponents up to -2 (the calibration standard they use has exponents between -1.5 and -1.9 in round-robin tests), Vertimill signature plots tend to have exponents of -1, and tend to be used for coarser regrinding than Isamills. Hukki observed, as Bond did, that the exponent in the rod/ball mill range was -0.5.

So what happens in the crushing range? If the characteristic is part of a continuum where the finest regrinding has exponent -2, moderate regrinding is -1 and "conventional" rod/ball grinding is -0.5, then one would expect the exponent not to be infinity, but to tend towards 0. This is how Kick's model was interpreted by Hukki: (size^0) means size doesn't matter.

Worth keeping in mind that "all models are wrong, but some models are useful"! Just because Hukki's conjecture, Bond's third theory and Morrell's formulae are all wrong doesn't mean that any and all can't be useful.

The shortfalls of the Bond crushability work index test have been identified for years and attempts to correct them or have better predictive testwork or correlation should be encouraged in our industry.

We all agree that something must be done to improve the prediction of power requirement for crushers.

Among the reasons we are still using the Bond crushability test today is:

•The extensive database from different operations available today. Values of new Bond crushability Work Index generated can be benched mark to different operations.

•Less interest by researchers in the past to improve the accuracy due to the lower specific energy consumption for crushers. This is currently changing. For instance the number papers on crusher/crushing circuit at the last comminution 2014 conference was - 10.

•The crusher selection (typically in primary crushing application) depends not only on power requirement but also on the top size of the ore.

•The crusher specifications available (installed power). For instance, if my calculated power requirement is 650 kW or 10% more, I will likely end up with the same crusher.

Crushing vs. cutting method for drill cores! Starting with drill cores, it is difficult to achieve the required size range specified by Bond which is -76 mm +50 mm by crushing the ore. You will end up with a considerable amount of rejects. In order to minimise sample utilisation since drill cores are expensive, I think it is preferable to cut or break the core to achieve the size required.

The advance media competency test (AMCT) designed by OMC conduct Bond crushability test on 5 different size fraction on "survivor" rocks after tumbling test in a -1.8 m diameter mill. The coarser rocks surviving the tumbling environment are generally more competent than the average feed. We have therefore to be careful in interpreting their results (before or after the tumbling test?). 

There is a lot to be said on the empirical derivation of energy required for crushing at different size ranges! It also depends upon the degree of repetitive actions for breakage; the Bond Impact test is one action/impact. Multiple actions (at low energy) could better describe cone crushing rather than gyratory or jaw crushing where application of force on size is more important?