Geometallurgy should start no later than the economic study phase well before the mine is built. (There have been a number of recent plant failures due to not understanding variability of oretypes and due to met testwork on non-representative metallurgical bulk samples). If you have a preliminary economic assessment that is positive and a decision has been made to advance to the next stage, a plan should be implemented to build a geometallurgical model. This model typically focusses on developing a 3D/ geological understanding of the controls on the variation in acid generating properties, variation in comminution characteristics and variation in mineralogy that affects process characteristics. Geomet data gathering and generation of geomet models can be expensive and time consuming thus I would recommend only starting the geomet program after a positive PEA. Ideally you should have a decent geomet model in place before selecting metallurgical samples for bankable feasibility study.

The geomet data is essentially de-risking the project and removing assumptions which is important especially now that capital is so difficult to come by.  Geomet work however can and has been used to optimize economics on existing operations. In these cases, the need is driven by poor recoveries, or lower than anticipated throughput.

The provision of geomet data and the timing of major capital spending decisions is still out of sync on many projects. Owners and managers are still gaining an appreciation of the risks involved in selecting non-representative met samples and not understanding oretype variation over the life of mine. Although the numbers are declining, the industry still has a few practitioners who are of the belief that a few bulk samples collected from surface representing the mean grade of a deposit is all you need to design a plant with.

On a positive note. I do see more and more clients taking the time to de-risk their projects by embarking on infill drill programs which provide sample for large scale variability geomet testwork. I have found that linking this effort to improving the resource classification, can make it an easier sell to the owners who have a hard time with the added delay.

One of the more common problems I see with those that rely on geological models is that these models tend to suffer from serious quality of data issues. Often core is logged by numerous individuals and companies over many phases of exploration with significantly different results. Because its based on visual estimates, I find that consistency in logging lithologies, structure, mineralogy, hardness, percent of sulphides, etc. over the life of the exploration project is of such poor quality that the data adds limited value. Its far better to replace visual estimates with hard lab data.

In preparing for resource estimation, I try to team up with a metallurgist and typically pull 30-40 drill holes per deposit from a range of depths, strike extent, phases of exploration, and variety of geology and often find that the quality of logging, geomet understanding, oretype classification and met sample selection is not where it should be in relation to the level of economic study.

We still need to do a lot more to educate study managers, mining engineers, exploration geologists, resource estimators and metallurgists in the value of geometallurgy.

One area that I see a lot of potential for extracting further value from deposits in is creating resource estimation constraint domains which places more weight on geometallurgical characteristics.

Populating block models with far less populated geometallurgical data is another challenge that needs further research.

Good discussion,

My only additional comment is to start a bit earlier than when PEA is done. Based on my experience in geometetallurgical studies (going back 24 years now) if a company thinks they have enough "critical mass" to conduct a PEA it is also time to initiate a geometallurgical study.

Good discussion. I agree with virtually all comments made here. A couple of my own to add. Geometallurgy begins with a common understanding of the orebody being held by the geologists, miners and metallurgists. Late in the exploration phase there is often still an assumption by the geologists involved in the project that their view of ore types is going to survive through to operations. This may or may not be true or appropriate. In the worst cases the view does survive into the operations phase when it is entirely inappropriate.

It is essential that this assumption be tested - this is the first job of the process engineer with a gemetallurgical bent.

The first test samples from a deposit can be a small number (2 or 3) of typical mineralized samples or composites that simply confirm the deposit type and likely treatment route (big picture only, leach, flotation physical separation, etc). It may also be good to do some confirmatory QEMSCAN to make sure the minerals you think are there are actually there. I have uncovered some surprises for the geologists by doing this, even removed entire "ore types" from the resource. The main outcome from this work is to scope the process flowsheet and consequently scope the tests that must be performed. If there are a number of possible flowsheets then tests must be done that support all options and unit operations.

The next set of test samples is the one that tests the geological assumptions. Select the two or three most common geological "ore" types or domains and test three to 5 examples of each. You need to do about 10 to 15 tests covering the major process issues. For example, if it is a large, moderately hard, copper sulphide ore then SAG/ball/float is highly likely to be the backbone of the final flowsheet and should have been evident from the scoping tests. The tests on the 10 to 15 samples (must be core, no RC) should be SMC, BWI, Ai, rougher flotation. You will need 10 to 30 kg of core per sample and each sample should be a single run of a geological domain. Assuming two major domains and 10 tests you should have 5 examples of each domain. Compare the competence (SMC), grindability (BWI), abrasiveness (Ai) and rougher recovery of 5 samples from each domain. If domain 1 is distinct from domain 2 then you have the advantage of the geological domains having metallurgical meaning. If the properties of the domains overlap strongly or are indistinguishable then you need to start investigating to find an alternative method of domaining the resource.

I could go on and describe the process all the way through to DFS & implementation but that would dilute my point. Which is: would you rather be re-domaining an orebody at scoping and PFS level to something metallurgically meaningful than during a DFS or worse, after operation commences! The latter approach is a recipe for failure as alluded to in some comments above.

Thank you all for the contributions and discussion here!

The common theme that I see emerging is that geomet should be started relative early in the process - say at PEA (depending on deposit type etc). I suspect that these comments reflect the nature of the audience in this group - i.e. those with a firm grasp of geomet who are championing best practice in their work.

I wonder how often such work is really started this early in the wider industry? As a generalised comment I see the industry starting to recognise the value of such work in de-risking earlier decision making - but that this still largely remains lip service in many instances; with early stage operations still willing to proceed and hope that things work out as they think they will - then ramping up geomet-type work later, or if there is a problem. Is that a fair comment?

Is there really a disparagy between industry recognising such best practice, and industry actually implementing such best practice? And if so - why? What can we do as geometallurgy champions to raise awareness and understanding? Is anyone aware of published case studies that really highlight the value of undertaking geometallurgy earlier in the mining lifecycle? Very hard to quantify of course.

Great discussion. A few more points that may help.

Again depending on the type of deposit, there is a very coarse form of "geomet" that can be organized prior to PEA during exploration core logging that adds little cost. Its rarely done as many pure exploration geologists don't tend to have a grasp of what data will be needed later on. The benefit of this early stage is to help us design a more informative geomet program in later stages of study.

Basic scratch/hammer tests can be included during logging on the entire drill core dataset to help frame broad comminution characteristics and the rough proportions of hard and soft material types. Particle size fraction descriptions on some type of deposits is also useful as clays and friable oretypes can have a big impact on processing. Having sample prep equipment in the exploration assay labs that provide us with some hardness data would be a big help but is not yet available.

As mentioned by others, small populations of MLA, Qemscan, microprobe work, or simple mineralogical studies all help at the early stages. Equally, percent and type of sulphides can be logged for each assay interval however one needs a quality control program to ensure all geologists log and estimate percentages the same way. Geologists need to be especially careful in defining oxide/sulphide boundaries. An oxidized surface of a sulfide grain logged by the field geo as oxide ore behaves vastly different in a process plant from a fully oxidized sulphide grain. Geologists need to break open oxidized sulphide grains, before classifying oxide/sulphide domain boundaries. Although core logging can help with geomet understanding, all visual estimates should eventually be replaced with hard test data.

I would argue that there is no standard number of MLA, Qemscan, Mineralogical samples one should collect at the pre-or post PEA stage. Its completely dependent on how the deposit was mineralized. Were there multiple mineralizing events over the life of the deposit, was there one or more remobilization events of ore minerals, were there one or more metamorphic, tectonic or intrusive events which affected hardness, grain size, ore mineralogy, how pervasive or localized were these events, was the deposit penetrated by groundwater, what was the depth of surface weathering etc? Which structures, lith contacts were affected by groundwater, The range in potential complexity of ore deposits is vast. Some deposits are relatively simple where one can get away with a half dozen geomet samples to other far more mineralogically complex deposits where the entire deposit has been populated >1000 geomet samples. The type of required geomet sample testwork also depends on the type of deposit.

In practice, I see that the major mining companies tend to run best practice geomet programs. The geomet sampling programs often blanket the entire depth and strike extent of deposits. In these programs there tends to be extensive geomet datasets with sufficient data to populate a block model with throughput rates, recoveries, product penalties, at a sufficient degree of confidence to optimize plant throughput and product quality throughout the life of mine plan.

Having been part of a number of failed plant design post mortems, Its still the mid tier western mining companies and a few foreign major companies, that are trying to fast track their projects, and constrain the study period and costs that suffer the most from insufficient geomet sampling and testwork, and resulting failed plant design.

Junior companies still largely struggle with the collection of geomet data as their focus is often to option out their project. Its very rare to see a junior implement a full geomet program and take it to EPCM however it does happen.

There were a number of good papers referenced at the March 2014 PDAC presentations on geomet that you should be able to get a hold of through their website.

In the 10 years I have worked with AMEC in Perth (formerly Minproc) we have dealt with the full spectrum of clients. The worst situations are when we are presented with a client wanting to put a DFS on TSX and they have ~5 composite samples representing "the orebody". The "best case I have of this was where the 5 samples in question had BWI values all in the range 11.2 to 11.9. It turned out that each was made up of 30 to 50 scattered (spatially and lithologically) core intervals. All it said was if they had 10 shovels in ore at all times there would never be a significant variation in grindability. Of course they were going to have a single shovel in ore and they had no idea what variability in grindability to design for and no idea of the implications of mining one "ore type" over another.

Obviously we have arrived on the scene too late in this sort of instance. When we have arrived on the scene early (scoping study, PEA) then we immediately start the conversation between geologists and process engineers. It begins a bit like this

• AMEC: I would like your full drill database so I can begin to understand the project

Client: but you are a process engineer - what do you need a drill database for

• AMEC: This is where my samples are going to come from.

Client: but we have already selected your two met sample holes......

About 30 minutes later we have discussed the actual sampling requirements, how we will use the database, what is actually needed for conducting the preliminary tests and what those tests will tell us.

About 5 weeks later we have initial met results and we start to formulate, with the geologists and mining engineers, what the next set of samples will look like.

We make sure the next selections are from within any early pit (years 1-5, the time-frame in which you need to be making an operating profit and paying back the majority of initial capital) being contemplated by the miners. It is unlikely that the early pit will change substantially for the remainder of the project, unless, of course, a new and richer deposit is discovered nearby in the meantime! Provided the drill database is notated to indicate what is inside the 5 year pit it is a simple matter to select the samples. It is also recommended that the 1 year pit be indicated in the drill database as this represents commissioning ore - how many times have projects been surprised by the ore they startup on - just the time when the financial plan needs to have strong revenue flow.

In AMEC's experience, clients that are educated of the risks of not understanding their only asset, and are prepared to engage fully across the disciplines, gain a valuable early geometallurgical understanding of the orebody and have a desire to continue it into production.

Great discussion and excellent comments by all.

From my perspective I have been intricately involved in kimberlite based (diamond processing) geometallurgical studies since 1996 when developing this thinking initially with De Beers and ever since as an independent consultant. A few papers are available which highlight the generic approach normally followed in diamond geometallurgical studies. Such papers have been presented at various SAIMM forums in the past.

In principle there are a few key building blocks for diamond mining geomet info generation, which in the case of Greenfield deposits basically starts at the early stage of Advanced Exploration (in the case of diamond mining). The application of the concept of geomet principles during bulk sample treatment, which really goes hand in hand at the PEA (pre economic assessment) phase coupled with some initial drill core analyses such as mentioned by Dave Dean (e.g. SMC testing) can go a long way of de-risking and highlighting technology opportunities at such early stages of the mine development life cycle. As such the use of a well structured and metallurgically excellently monitored bulk sample plant (BSP) treatment campaign is crucial to enable early stage bang-for-buck geomet data/info generation. The design and use of a basic onsite geometallurgical lab situated at the BSP coupled with offsite specialized test facilities is key to generate the required flowsheet design parameters at an early stage. Underlying the whole approach is a well designed and applied geometallurgical sub-sampling philosophy.

The use of the BSP by the very nature of diamond mine development can be perceived as being an added (and different) geomet approach when compared to other mineral deposit development projects where samples are more readily ‘assayable’ from drill core only than in the case of diamond winning. In addition, and similar to other mineral geomet practices, and as stipulated in earlier discussions, mineral/ore type specific metallurgical testing of geologically well interpreted drill core remains absolutely fundamental ahead of more capital intense pilot scale testwork.

The geomet plan which ideally rides alongside the mine development project stages is thus to get maximum bang-for-buck in terms of developing a geometallurgical campaign from the early concept stages with peak geomet expenditure (from a design viewpoint) during the pre-feasibility stage. The latter studies will also encompass pilot stage and equipment specific testwork per major geological ore type in order to de-risk the design and to establish a accurate bankable design.

Another important aspect of geometallurgy is to be reminded that this is a dynamic ongoing approach even after a mine has successfully been developed and has entered into the realm of being Brownfield. Each resource extension programme (for example core drilling for further geological delineation of the resource) is a fantastic opportunity to mitigate and plan for future metallurgical treatment risk due to significant, or as often found even subtle changes in ore mineralogy. Constant striving for metallurgical treatment optimization and shareholder value enhancement must be a fundamental philosophy for geomet practitioners and mine developers (project client) alike!!

Considering my experience of more than 30 years spanning a variety of complex zinc sulfides , I have seen the various stages of evolution of geomet modelling ,which like Al has said would be more of the chicken following the egg type , but currently with the number of tools available for working with very small size of samples a pretty appropriate geomettlurgical model can be created suitable for flow sheet development , mine planning in the initial stages itself .

For ex I have found that Point load tester can give a good idea of the work index , so at the time of exploration logging use of PLT can give good idea of the variation in workindex along the strike and depth . JKFI test can give index of floatability of diffrent types of ore and a mettlurgical model of the depost can be made .

It is an ongoing process and the model neeeds to be continously updated .

I agree all commentary so far and suggest that there are two key times at which a Geomet program adds value. The first is from prefeas through to final plant design and construction and the second is through ongoing operations...basically over the whole life of a project.

The first is targeted at understanding the nature of the beast to get the process and design right while the second is to make the most of the design, to provide a basis for dealing with with ongoing variability, to aid performance troubleshooting, to support business plan design and support the inevitable process and plant optimisation.

Sample selection for metallurgical characterisation and testing should start when initial concept drilling is completed, cores are assayed and logged, initial modelling completed and it looks like there may be something worth pursuing. The drill database should include assays, lithology, mineralisation, alteration, petrography, ICP scan assay data etc and geo-tech data such as PLI, RQD, RMR. The PLI has already been mentioned. RQD in particular reflects in situ rock size and has a bearing on blasted rock size distribution that in turn drives crusher and mill throughput.

We might have a potential resource that has tonnes and grade in it that looks like it might be economic but can we convert this latent value and process into a valuable saleable product? What might the process plant flow sheet look like? Are there any red flags that might need particular attention? As Dean points out, met test sample selection should always be done by mets/process engineers in conjunction with geos so that a range of samples are selected for testing to cover major and minor ore types/lithologies, relevant grade ranges and reflect mineralogical and alteration differences that might impact plant design, met performance and long term production.

Met characterisation (hardness etc) and testing results (recovery and grade etc) should be incorporated into the geological model once a preferred flow sheet has been nominated so these can be mapped and will generally be post prefeasibility drill and test. More samples will be added as definition drilling continues and the process and plant design is refined. Post construction and start of operations, infill and ongoing drilling and met testing should continue with continual update of the database and models.

It would be frustrating, time consuming and costly to complete met testing on samples selected in isolation for met characterisation and testing only to find once work has been completed that the samples selected represent for example interesting geological anomalies but don't properly reflect the nature of the orebody. The results have limited use for developing a conceptual flow sheet and associated process economics to inform ongoing exploration decision making. Worse still would be building a plant based on flawed sample selection. As a minimum, the work must be done again so far better to get it right first time.

From an ongoing operations point of view, geomet might be as simple as linking daily ore delivery information and mill information and performance to build an understanding of how day to day operations and met performance is influenced by ore characteristics. This provides a simple tool to troubleshoot plant performance variance and identify what is driven by orebody or operational factors so strategy can be developed to deal with these. This might extend to advanced sampling, testing and modeling.

Newmont's Batu Hijau provides a real example of the path followed to get to a point of sustained applied Geomet in practice. For more on this, take a look at the papers we submitted for SAG 2006 (Batu Hijau – Controlled Mine Blasting and Blending To Optimise Process Production at Batu Hijau) and SAG 2011 (Applied Geometallurgical Characterisation for Life of Mine Throughput Prediction at Batu Hijau) I think these are available via OneMine.

This is a great discussion. I concur with all of the earlier comments. Geometallurgy should be introduced as early as possible, ideally at the pre-conceptual stage of a project. An understanding of ore-body variability is imperative to the success of a geometallurgical programme. Unfortunately, too many operations are using geometallurgy in a reactive capacity, rather than in a predictive one.

I did read the paper entitled: “Batu Hijau Model for Throughput Forecast, Mining and Milling Optimization and Expansion Studies” a couple of weeks and found it to be very useful in providing a practical approach to the use of comminution data in the generation of a geometallurgical throughput model. I have recommended it to some of my colleagues. Thank you for the additional references.

I concur with your statement on reactivity. Unfortunately there is also a lot of misguided geomet out there as well. Sometimes it is done to satisfy the nebulous "we have to do geomet on the orebody" without actually having a goal (a project value adding proposition) in mind. This leads to pointless geometallurgy which is stopped as soon as someone comes along as asks the value question. Another problem is the "do everything" geometallurgy. This is the expensive geomet program, probably with some sort of economic justification, but it turns out to have a life of its own and results in one of two outcomes. 1. the value cant be found because there is too much untargeted measurement and not enough time to analyse it or 2. before the value is realized the value question is raised and the program is seen to be lacking justification.

Sometimes the reactive solution is useful because you are actually reacting to a problem and someone has proposed a way to address the problem through a specific geomet exercise. Rather than a "we have to do geomet" approach without a goal, the issue becomes "we have problem 'X' and it can be addressed by a geomet program that measures these specific things that make 'X' happen". This leads to a value being placed on curing the 'X' problem and a set of targeted measures, whose cost to obtain can be weighed up against the losses due to 'X'.

As we are dealing with nature and the inherent variability of orebodies, we are often in the position where we have to be reactive because the issue suddenly arises. We hit a new ore type that is problematic and it is so serious that we have to be able to predict how often this is going to happen in the future.

If we have a structure geomet approach we can be reactive in the correct way. If we have a serious look at the process issues during the design phase and we have a knowledge of geomet then we can also be proactive without it costing the earth up front.

Thank you for your comments. I absolutely agree with you. Working without a goal is never advisable. One may strike it lucky periodically, but one may, also spend an inordinate amount of time chasing one's tail. I too believe that in more mature operations, learning from the findings of the "reactive" approach could be of great value. However, if faced with a less mature operation, the introduction of a geometallurgical programme (i.e. the generation of a spatial model for relevant metallurgical parameters to determine orebody variability) would prove invaluable in risk reduction. It should be anticipated that as an operation matures there are going to be times when "reaction" will be necessary. These new findings/learnings from these ad hoc troubleshooting exercises should then be incorporated into the existing model to alleviate the necessity for revisiting problems (work duplication), i.e. yet again moving towards a predictive approach, rather than a reactive one.

Thank you again for your comments. I have found this particular conversation most thought provoking.

Thanks all for a very interesting discussion. Ore bodies are unique. There might be similarities and that helps us to start off with what we have learnt from elsewhere. And, its important to start off..like Michael says. Risk can be mitigated but we won't know the full story until all has passed through the plant- the best way to sample is to mine (and process)! That is very costly sampling. Please hear my story:

I have worked on a highly variable polymetallic igneous (and partly metamorphosed, of cause) deposit. The inherent geological variability is metre scale and is directly linked to metallurgical variability in complicated relationships. Two plants were processing this ore in parallel. The older of the two plants had seen >20years of metallurgical engineering evolution and was still counting! The newer of the two plants was supposed to have incorporated all the lessons learnt earlier but alas, there were still more lessons to be learnt! If a 3rd plant had to be added on, what would it look like?

'Geometallurgy is about listening to the geology in order to get the most value along all aspects of the resources value chain.' Once a deposit is known, brown fields exploration should be used to maximise ore body variability understanding. It is correct to do that because it is cheapest to do that at this stage. Rigidity in process plant designs is a weakness. The older plant mentioned above turned up in my experience to be the 'real geomet evolved plant'. A lot of the added-on funtionalities remained on-off optionals. That dynamism in my view is where real geomet adds serious value.

A lot of modern systems include online monitoring technologies which if strategically positioned, can actually monitor geological variability early enough for the process plant to react, if it can react. And everything else downstream from there.

"Geometallurgy" to mining is like "gourmet" to the restaurateur. If making delicious, unique, well balanced plates is what makes a gourmet, then every lunch room and sandwich shop would likely give you a 20 min explanation of how they qualify. Good mining projects have always included geologists, miners and metallurgists as a team. They have always looked at ore types variability testing and tried to make better projections of metallurgical performance.
In a PEA, which used to be called a scoping study, there is usually only time, sample and budget for fatal flaw analysis testing. Major composites are tested and little more. The composites are put together with geometallurgical guidance.
As a project progresses, the metallurgy is usually battling to keep up with the resource measurement. By the time you are in a PFS, some serious ore domaining and variability testing should be going on. Some would call this geometallurgy.
By the time the a feasibility study is finished, all major domains should be identified, quantified and have sufficient testwork done on them to support a final project.
It's risky in an old school way to do too little testing. 5 out of 10 times I think you are probably fine.
Worse is sucking in time, resources, drilling, samples and budget to make a thousand data points, of which you only know how to use 58.
Let's stop pretending that geometallurgy is a new science. Sure we have bigger deposits, faster computers and better testing tools, but we're still just calling for a pizza on our $800 mobile phones and we're still just working out how to make mud cheaper with our geomet models.