Most projects I have been involved with will look at extremes in the mine plan and evaluate the differences in plant performance using a simulation program such as METSIM. A view then has to be taken with respect to frequency of variability excursions that will take the plant outside of its design envelope and whether it is appropriate to design the plant to handle such excursions or to live with them and try to plan ahead to blend the feed when these events are approaching. Also focussing on the first 5-6 years of feed during which period the project's financials are most sensitive to production or recovery hiccups is important - i.e. don't get carried away with designing for variability issues that will be encountered at Year 10, you have to be profitable for Years 1 to 9 first !

I have been looking at some interesting projects lately where surface sampling is easy and as a result a lot of the metallurgical testwork has been conducted on large grab samples or bulk samples. Whilst this allows for larger scale testwork, it misses the issue of variability. I personally like domaining the ore body into lithologies and conducting testwork by lithology at the PFS stage, as well as some mine blends if I have enough sample. I like to look at the variability within the DFS test program with as many samples as I can get hold of or afford, both for comminution and processing properties. Then assembling some mine blends of the lithologies that are representative of Year 1, Years 2-3 and then the rest of the mine to deal with Boyd's discussion about getting the first few years correct. Whilst blending is costly, making space for a decent size ROM pad in the study phase is useful because you will always dig up some weird stuff you did not know was coming that you will want to blend into the major plant feed stream.

I think that the concept of variability is relatively well understood with respect to communition. The OPEX can often be more accurately determined than the CAPEX and with the tonnes as denominator, any bottleneck in the communition circuit will be an issue. Also, power for the grinding mills, at ~40%-50% of the total power requirement of a concentrator is highly visible.

Variability testing for communition will always be easier than for metal extraction or mineral separation - sample aging has minimal impact on the properties measured.

One challenge for variability testing for metal extraction or mineral separation is that the flowsheet must be relatively well defined - and by that time a vast majority of the exploration sample are too old to be useable for this purpose and very few clients are open-minded about "drilling the deposit again" come the time to do the feasibility study.

Moving the flowsheet development test work to a much earlier stage - i.e. one the size and grades of the deposit is approximately defined - could be one way to approach this. Another way might be to carry out inexpensive characterization testing of samples as exploration programs generate samples.

Once in the late feasibility/execution phase of a project, one really needs to look at the payback period and have high confidence in the ore grades, the throughput capability of the metallurgical plant, and the metal yields. And, by the way, this exercise needs to continue throughout the life of the mine as part of the production planning process.

Although prior to the engineering phase most plant designs are based on extensive laboratory and pilot plant testwork to determine the most economic and rugged process for the variability ore samples submitted to experiment, and accompanying decisions are made regarding the zones to be mined for the production period envisaged, there is perhaps cause to continue through the life of the production a geometallurgical mapping program. Such analysis on zones still to be treated, using the key (sensitive) parameters of comminution and mineral processing (and extractive metallurgical) operations which most significantly affect performance, could provide advance warning of risk and need for further evaluation of changeable operating parameters.

Variability needs to be defined as a function of time, short term, long term, mid term all have differing response times and can be dealt with to some extent. Designing a process plant to cope with all possible situations is impractical and unnecessary, modern designs integrate geology, mining and processing into the design criteria to cover most reasonable variations. If variation becomes unreasonable then the mining people need to be spoken too. I also believe that designing plants to reject low grade material at the earliest opportunity will help to smooth variation. After all when you consider that most ores only contain a few percent paydirt it translates into process plants putting a lot of effort into producing a lot of rubbish!

I agree. The time over which the variations occur is critical to the questions that need to be asked. The DFS program can be used to identify which parts of the circuit are sensitive to variation. Solutions to variation can then be catered to the high risk items of that specific orebody. The sample selection presents the risk of garbage in garbage out. If you blend a sample with fast flotation kinetics with one that is slow you do not get a happy medium. You only lose the opportunity to figure out what conditions are required to process both types with the same equipment. The mining guys and the process guys need to work together. Predictable feed pays with stable optimized conditions achieving high recoveries. Instead of 'how do I design for variable ore?' Perhaps we should ask 'how do I get samples that can forecast the required conditions or performance?' And 'how do I plan the feed to group sections requiring similar conditions?' Both sides then carry some responsibility. Let the process guys know what's coming, and try to line it up. In this fashion the short term variability could be minimized at the cost of detailed metallurgy. Not depth but breadth.

We have faced this issue many times as grinding media suppliers - but very rarely suggest equipment changes. Feed variations are the usual culprit with F80 and WI changes from the original plan. In almost every case the situation can be rectified with a change in media size, media grade or a multiple size preblended media solution. This plan works very well, even in case of crusher breakdowns when SAG is the first stage of milling.

With processing plants having higher WI ore coming up the belt without notice a quick fix can be the addition of a different media that is held on site for such an occurrence.

We have some toll milling customers that keep a wider range of media on hand as the next milling campaign may require a different grinding media solution which can be adjusted easily in just a very short time to meet the ore characteristics.

These days media is available in singularly crafted designs to deal with such feed variables as and when they occur.

An example of this has been our long experience in a single stage mill that has a primary function to produce a 56uM product with the best throughput at any one time. The initial plan was a blended, matched graded ball addition of two sizes. WI reduced as did the F80 leaving an adjustment to a single size/grade ball. Then the customer decided to increase the ball charge which then required a different grade of ball to be used as ball on ball frequency was greatly increased - a small trade off to eliminate spalling with a small increase in steel consumption.

We are always open to collect customer data and check (at no cost) if the media additions can be improved to better suit the milling environment or ore variations.