A linkedin.com group was recently discussing this by asking:  I’ve tried to estimate AG/SAG Mill power requirements using several different methods:

(1) Using the results of SMC data
(2) Using SAG Power Index (SPI) data
(3) Using the 1989 Barratt method updated on Alex Doll’s Web site (February 16, 2015)
(3) Using Don Burgess’ method (11th Mill Operators Conference 2012)

They seem to give very different predictions for guiding the selection of SAG mill sizes.

I tend to favor the updated Barratt method in terms of the rationale behind the approach, but I don’t have access to a database to really check it out. The Burgess method gives predictions that seem to be over-sensitive to the transfer size. I recently attended a Global Mining Standards and Guidelines Group workshop that focused on the Bond Work and SMC methods; they seem to be the most popular methods, but are they the best?

Many answers went as follow:

1. The spark test is not actually a test but a red light for a consultant. The sparks are caused by the fight between the rock and the machines. It is an indication of high rock competency. Hence competency tests should be developed.
   I agree with your observation, consultants and engineering firms need to develop techniques to highlight the risks in the Owner’s decision making process. I’m almost sure that in some cases described in the papers as failures there is a person that is saying “I told you this was not going to work”. Let’s blame that person for not being able to convince? or is it a harder task to convince to the person who is paying your wages?
2. We certainly prefer using JKSimMet when sufficient DWT and SMC data is available. It’s much easier to test different conditions, ores and flowsheets than other methods.
   My feeling is that there are two reasons for most of the recent mill design screw ups:
   i) poor data. lack of sample representivity, ore types missed or tests screwed up.
   ii) poor analysis. The methods were simply applied wrong. Misinterpretation, poor understanding of the model, poor benchmarking of the results of simulation.
   In summary, the results from the different approaches usually converge if done correctly, except where the conditions are outside the model’s envelope.
3. 1 Rock, 3 faiths http://www.911metallurgist.com/blog/ore-hardness-testing-grinding-measurement
4. I like your analysis and demonstrated knowledge of the subject. Thank you for including our work with the discussion of that done by the acknowledged leaders in our industry. But the use of the word faith reminds me of the old joke about the difference between faith and knowledge.
   Since there are not many direct comparisons between SAGDesign and the other methods, why not compare SAGDesign testing to the plant operations that have been designed and benchmarked using it? Our record in both of these areas proves that it does provide accurate measurement of ore hardness in kWh/t required to grind the ore when proper samples are taken. Total cost and elapsed time to do the design work are also important, especially when the manufacturers are busy. This needs to be part of the discussion. It is interesting that these matters are not usually discussed. We invite any comparisons in all these areas.
5. Faith, OK not really. Knowledge, not really either.
   As I like to say “In God we trust; all others must bring data”, data and methods is more appropriate. We could get all philosophical..
   Faith is defined as complete confidence or trust in a person or thing; or a belief not based on proof.
   Confidence is generally described as a state of being certain either that a hypothesis or prediction is correct or that a chosen course of action is the best or most effective.
   “Unless you have confidence in the ruler’s reliability, if you use a ruler to measure a table you may also be using the table to measure the ruler.”
   We must use the Method that best suits the situation (geographically and commodity dependent).
6. One method missing in your list is the SAGDesign test. I have been through all of them (DWT, SMC, SPI, etc) and here is why you should seriously consider that test:
   * One test on 10 kg sample (more for high SG), not many tests on same sample as proposed above.
   * Gives you both SAG hardness and Bond Ball Mill Work Index (on SAG ground ore, not from crushed cores) – all you need to design circuit & size mills.
   * No hidden formula, no IP – only open technology available i.e. convert the test results (in kWh/t instead of strange or no units) to SAG and Ball Mill pinion energy (in kWh/t). No software, just Excel.
   * No fuzzy factors, no phantom cyclone, no parameter fitting. Just straight through adjustments, sound & documented.Each test has his own strengths and limitations which comes from the experience of using them, make sure collect those from different sources. Malartic was designed on SMC test with 1 DWT if I remember – NI 43-101 available on SEDAR.
7. Alex Doll: I just took a look at your link to the Bond analysis of published survey data. It looks like you have done a lot of good work on that analysis, thanks for sharing.
8. Michel can you please clarify what you intend by the Malartic reference? Obviously (it has been well documented) the SMC test has been used on a very large number of projects and I can speak with familiarity on several that we have been well satisfied with. Based on the users of the SMC test and associated technology in the industry (see the SMC test website) no doubt there are a number of others with similar experiences. I am wondering why you single out one project in particular – is there a more enlightening publication than a 43-101 that speaks to any gaps between design and performance for that case? There are plenty of examples across the industry where things like inadequate sampling for ore characterization, design point selection, flawed engineering or just all round bad decision making and analysis has cost the project ramp up to full production. At the end of the day if the design engineers have a method that they trust and are familiar with then they will be well armed.
9. You did not mention the testwork you and I have patented with Mintek on modelling SAG mills which also provides the SAG specific energy consumption. I guess because we did not test it extensively you prefer to keep it under the radar?
   I think the first requirement to predict accurately the specific energy of SAG mills consist of conducting a test or many separate tests covering the full range of size distribution expected in industrial practice. The testwork to be conducted should also replicate the mode of breakage occurring in a SAG mill. Pilot plants testwork (typically having 1.8 m of diameter) are good in that respect but require a large amount of samples (sometimes not yet available) and are very expensive.
   We have to make therefore assumptions in terms of testwork and use models with parameters calibrated against operating plants or “database”.
   Since none of the method (and there is not one) covers the full range of size distribution typically feed to SAG mills in terms of testwork conducted and not all methods of breakage are reproduced in their testwork, each method is expected to have its own limitations.
   I have reviewed a few reports for greenfield applications where different methods (Morrell, Starkey, Barrat) were used on similar samples to predict the throughput. I have found convergence of results in some cases (within 10 – 15%) and in other cases, the differences where up to 25%.
   I believe that all models should be updated with time as we are getting a better understanding of SAG mills. Database should also be updated as process improvement/optimisation/ control are reducing the SAG specific energy consumption.
10. You bring up a lot of good points – I’ll add my thoughts on the inappropriate extrapolation of 20-30 mm particle breakage to infinite size. It is one of my main complaints about some of the common tests (yes, John, your test too). “We” are throwing away valuable information in the preparation of 20-30 mm samples by assuming it behaves the same as the 20-30 mm material in the apparatus. Capturing the differences in coarse particle breakage, let’s call it “crushing behaviour”, gives bench tests a much clearer window into coarser particle breakage.This is why I’m a particular fan of the Bond series – it has a test in a coarse ~100 mm size range that can conceivably tell you if the material is competent or fractured at coarse sizes. Yes, the Bond crushing test has problems — that could be a whole discussion on its own. My point is that at least the Bond series attempts to characterise the really coarse stuff separately from the “medium” size 20-30 mm stuff. You can have an ore with A×b of 30 and crushing Wi of 5 (Los Bronces, an Andean ore) or another with A×b of 30 and crushing Wi of 30 (Bonddington, an Australian ore) that behave very differently in a SAG mill. You won’t see the difference in a drop-weight test or a SAGDesign test.Final though: All models are wrong, but some models are useful. Both drop weight tests and SAGDesign are useful in the right contexts; same for Bond-based models.
11. Plenty of commentary coming from all directions. However none of the methods have been officially published. So if you are trying to compare one method against another it would be off your interpretation of the equations and your database to validate the model. I’d still vote that experience counts for 80% of the solution and the relationship with your client.
12. The grind-out and locked-cycle tests we developed using the 600 mm torque mill with a 75 mm rock top size certainly merits further work. I still like the idea of a grind-out test with water addition to maintain a constant pulp density inside the mill and with a very restricted grate aperture size to keep specific discharges rate to an absolute minimum. Dynamic simulations of grind-out tests could indicate that it might be possible to back-calculate parameters for a specific discharge rate function and a specific cumulative breakage rate function using this approach. By repeating tests at different volumetric ball loads but the same initial total charge load one could develop a comprehensive SAG model.The big challenge is to reconcile Malcolm Powell’s Grind Curves for a continuously fed mill controlled to maintain a constant total load volume at values above those giving the maximum throughput or power draw. One could, of course handle rocks sizes in a 6ft mill up to about 200 mm without extrapolation. Such tests would probably greatly reduce the amount of ore needed, compared to full blown pilot runs using a continuously fed mill, especially if one wants to check out the effects of ball and total charge loads, and seek a set of optimal operating conditions.
13. With regard to extrapolation, it perhaps worth adding to Paul Morgan’s comments with regard to Mintek’s GrindMill tests. These tests are designed to delineate the energy-based specific breakage rate function in the region of its peak value as well as at fine sizes. Based on this data, together with some assumptions regarding the parameters controlling normal breakage, abnormal breakage, and self-breakage, Paul is able to predict the performance of SAG mills with high ball loads (ROM ball mills) with reasonable levels of confidence. However I’m not familiar with the details of his analysis method.
14. Using the grind-mill test together with Bond tests gives the breakage rate trending with size (once you have a correct breakage function) and will discern these anomalies. You get spectra, instead of single point data. Understanding the breakage function for the rock in question is half of the battle. A, b and ta give an inkling of the breakage function.
15. Something that SMC has addressed better than most is the trade-off between sample availability (size and geomet significance) versus test requirements. The SPI test used to be popular for a reason. It could be done on small samples with small particle sizes. That meant that anyone could go into the core shack and come out with a bunch of samples for free. A JK DWT pretty much needs PQ whole core to do it properly. Most try to get away with half core PQ or whole core HQ, but we try to avoid that for reasons that Alex has mentioned. So, when faced with the prospect of spending months sinking hard to find, PQ drill rods at the cost of millions, it is easy to be tempted to look for inferior alternatives. Pilot plants are even harder to feed!
16. In terms of proprietary small-scale tests, it looks like the SMC and SAGDesign tests are the most popular; both seem to claim about 7 per cent accuracy. Perhaps a significant difference is that Steve’s method predicts only the overall specific energy, whilst John’s method targets, specifically, at the specific energy for the SAG circuit on its own. However, specific energy requirements for ball mill circuits can be predicted from grigrindabilitysts in small torque mills following the approach pioneered by John Herbst, which is well published in the literature. It is perhaps also worth noting that the SAGdesign allows one to back-calculate parameters of breakage and selection parameters, because the test generates both feed and product size distributions (mill power for the Starkey mill can be estimated in the same way it can be estimated for the Bond mill as described many years ago by Levin in the SAIMM journal) and immediately tells you if the Rossin Rammler slopes at fine sizes are preserved (a basic assumption of the validity of the SMC test).
17. Since you did not mention it specifically you may not realize that the SAGDesign test is a second generation comminution test, quite different and more robust than the SPI that you do mention. Rather than take our readers time who do understand the difference, I would recommend that you and I talk personally by email or phone to come to an understanding about what a SAGDesign test provides. I look forward to speaking with you.
    The accuracies you quote need to be better defined to be meaningful. A SAGDesign test is a SAG test at better than 5% and Bond test at the normal variance, about 10%. Together they give better than 7% on most plant benchmark tests. On the other hand an SMC test only relates to coarse breakage.
18. I understand that you have moved on from the SPI, but that does not take from the fact that many mines are still using it successfully for variability testing, even if it has lost popularity for design. We’re disciples of the JK methodology, so I am far from impartial. I remember that we discussed how you would like to see the different schools come together with one common agreed approach. That would make this kind of forum a lot less interesting! Along with many others in this forum, I really admire what you have done. It is great that there are at least two different schools of thought competing to find the truth. It would get destructive if manufacturers start requiring to do proprietary or incompatible tests. We see that with HPGR, where each equipment supplier wants tonnes of samples to do their own tests. It means that you have to choose a supplier during the study stage, which is misaligned with the way projects are usually done. Outotec and FLS certainly seem to be in the SAGDesign camp, whereas Metso and Citic are in the JK camp. It will be interesting to see if this causes more polarization or if we will see some healthy competition leading to some innovation.
19. The task at hand is to provide the truth – which is to give a client a SAG mill that performs as intended. The second challenge after getting the correct kWh/t for the SAG and ball mill on a suite of variable hardness samples is to decide whether the design will be for fixed tonnage or variable tonnage. The effect of variable tonnage on recovery must be unknown because no one seems to be interested in this effect even in copper flotation plants where variable tonnage will cost some recovery loss. My point is that if we design plants for steady tonnage and maximum recovery, there will be a choice of which way to run. And the fixed tonnage plant will be more profitable than the variable plant because of a slightly higher capital investment in comminution equipment. It is not a matter of right vs wrong but a matter of choice to be as profitable as possible. Surely in Chile right now, that is important to recognize for new plants.