Laboratory Testing & General Mineral Processing Engineering

Laboratory Testing & General Mineral Processing Engineering

  • To participate in the 911Metallurgist Forums, be sure to JOINLOGIN
  • Use Add New Topic to ask a New Question/Discussion about Mineral Processing or Laboratory Work.
  • OR Select a Topic that Interests you.
  • Use Add Reply = to Reply/Participate in a Topic/Discussion (most frequent).
    Using Add Reply allows you to Attach Images or PDF files and provide a more complete input.
  • Use Add Comment = to comment on someone else’s Reply in an already active Topic/Discussion.

Sample Splitting Equipment (24 replies)

U
Unterstarm
8 years ago
Unterstarm 8 years ago

What is best practice when splitting a -80mesh fraction of stream sediment collected for gold analysis? Small riffle splitter, rotary splitting device (RSD) or other?

The -80mesh fractions weigh up to several hundred grams and need splitting for a 50g analysis method. Splitting by an RSD is more expensive but does it produce a better (unbiased) split?

(unknown)
8 years ago
(unknown) 8 years ago

It is strongly recommended to use a Rotary Sample Divider. However, you must also perform quality control of RSD. People forget that this equipment should be checked and are not infallible; they can also produce bias if not held periodic maintenance.

(unknown)
8 years ago
(unknown) 8 years ago

What checklist that we should inspect and instrument check to make sure that our RSD is perfect and will not produces bias?

O
OberstGruppen
8 years ago
OberstGruppen 8 years ago

Rotary splitter/spinning riffler is the only way to get the sample-to-sample variation less than 1%. See the Table in Terry Allen Particle Size Measurement Chapter 1. Chapman and Hall (any edition from the 1st to the 5th - the page number only changes).

Helena Russell
8 years ago
Helena Russell 8 years ago

The best practices in Chile for sample preparation to chemical analysis say us (regarding RSD):

Vibratory feeder (or belt conveyor. Depend of the model): check the rate of feed weekly. The operator can't vary the rate of feed daily. This must be standard for avoiding the dust production and the segregation of the particles. Rotating sample collector carousel: check the rpm weekly. It must be standard according the supplier.

Removable stainless steel containers (buckets): Put 10Kg of ore into the hopper (weighted in a precision scale, 0,001g sensitivity), particle size 90% -8# Tyler. Place two 5% containers (marked N°1 and 2) and two 10% containers (marked N°3 and 4), facing each other, in the carousel. Run the RSD. Finally, measure the weight of the obtained samples. We can accept 5% of variation from the expected weight. For example, the sample in the 5% container must be 500g ± 25g (5%). (Note: Usually we split 1Kg to pulverize 95% -150# Tyler and to perform chemical analysis). To evaluate the performance of the RSD, we perform a screening test to the containers, using at least 5 meshes. Each fraction obtained must be weighed and analyzed for Total Copper or other element.

(unknown)
8 years ago
(unknown) 8 years ago

Rotary splitter/divider is the only way and yes, checks the speeds. The only 'universal' method is a slow speed with constant feed rate and also, a pre-riffle check on reduced agglomeration. It may be a good idea to check all sorts of parameters that may be sample dependent, such as the SG and actual size distribution, as nominal -80 can be practically useless without such benchmarking. Attempts to speed up the process will create bias which lack of periodic replicate checks will miss and then all effort is wasted.

U
Unterstarm
8 years ago
Unterstarm 8 years ago

Thanks for the comments guys. RSD seems definitely the way to go. I did see a table that quantified the split-to-split variance in the 'Rocks to Results' booklet available from SGS and from what I recall (I'm on the road at present so don't have the booklet to hand) it also put the split-to-split variance at 1% by RSD compared to 5% by riffle splitter.

Next step is to check with the lab in question if they check their RSD on a regular basis as mentioned above.

Amar
8 years ago
Amar 8 years ago

A rule-of-thumb for the vibrating feeder to the RSD is that the feed speed for the material fed to the RSD should be such that the sample split should involve at least 50 cuts. Less cuts means failed representative quality.

(unknown)
8 years ago
(unknown) 8 years ago

With only a few hundreds of grams, I think you should mill the - 80 mesh fraction to minus 200 mesh prior to riffle splitting or rotary sample device. Most labs have mill bowls capable of milling up to 1 kg.

A sampling device can be thought of as a gambling device - the roulette wheel and most RSD samplers have some similarities. As an example, if the RSD has ten trays, you (the gambler) get to bet on one receiver tray. The reject (the house) bets on all the remaining nine receiver trays.

To help give all of the receiver trays or riffles in a jones splitter an even chance, mixing the sample, prior to loading any sampling device will help distribute particles more evenly and counteract segregation within the sample.

You can also improve your chances by taking more cuts, in the case of most RSD's this involves slowing the feed rate, however this may be impractical due to higher sampling charges or slower sample turnaround times.

But if you do not have enough gold particles to create something approaching constant feed stream - again using rsd example above, your single receiver tray will more easily avoid the gold than the nine reject receiver trays, as the reject trays stay in the stream nine times longer.

If you can mill the sample prior to using a sampling device, you may be able to break up rock particles containing multiple gold particles. Keep in mind that if this same particle remains unbroken, the sampling device can only see ONE gold particle. Milling can also create more gold particles by cutting, scratching, smearing and abrading larger gold particles.

Either way - there is a very good chance that milling the sample will result in more particles being available - this will make it less likely that any of the receiver trays (including the one you bet on ) in the sampling device can avoid your gold particles.

(unknown)
8 years ago
(unknown) 8 years ago

A summary if I may: A splitter only works well in producing a representative small aliquot if there are enough particles of a given size/composition to be shared among all the containers and that they are delivered over a sufficiently long time period that they don't all end up in one or two pans. Rotary riffling is generally considered to be the best way to go, but the delivery rate needs to be reasonably slow and the rotation speed of the pans relatively high.

Some blending can help, but don't be surprised if there's some segregation again due to the vibratory motion of the feed hopper.

Agglomeration of milled material (pulp) can still be a problem: this "snowball/hail stone" effect can cause gold assay to be inconsistent (at least not as good as hoped for) even when carefully riffled. It might be that a somewhat coarser milling might be better - still a statistically sufficient number of particles, but less agglomeration.
Quantachrome makes lab scale rotary rifflers: 120ml and 2L hoppers.

O
Obergruppenfuhrer
8 years ago
Obergruppenfuhrer 8 years ago

RSDs are better than riffles (there are other studies in addition to the one you noted - see the website below for one). However, you did say up to several hundred grams, which is small for most RSD splitters. The practicality of using RSDs for small samples needs to be considered as most RSDs are designed for larger samples and particle sizes. The following website has an example of a smaller RSD that might be suitable (in addition to the Quantachrome website), if your lab has one and they can demonstrate proper usage. http://is.gd/UKhiCi

Practicality may mean considering mini-riffle splitters, providing they are fed properly and have an adequate number of riffles. Labs that I review use mini-riffle splitters for pulp samples much larger than several hundred grams if the client wants an improvement over fractional scooping and an adequate RSD is not available due to cost and limited client usage requirements.

(unknown)
8 years ago
(unknown) 8 years ago

I like the analogy here a lot: "A sampling device can be thought of as a gambling device - the roulette wheel" and will borrow this phraseology in the future (with correct attribution, of course! Similar to the phrases Francis Pitard uses: "Are you taking a sample, a specimen, or are you gambling?" "Are you using a scoop like your grocer? etc).

80# is 180 microns and 200# is 75 microns in ASTM E11. Given a gangue background of say quartz at 2.65 g/cm3 then from a PSD (note: not chemistry/mineralogy: the latter is vital from an extraction perspective) perspective roughly each 8.1g is homogeneous (1% standard error) at an x99 point of 180 microns and at 75 microns (1% SE) then the sample requirement is 0.6 g approximately (substantially reduced -- in proportion to the cube of the size). The real problem here is the nugget effect and imagine only one small (< 75 um) piece of gold in the sample - then this nugget can only sit in one of the riffle trays - homogeneity is impossible in such a scenario. Milling may only flatten this piece.

However, the above is no excuse for not trying. Without an attempt at a scientific division we can only calculate the ore content of a mine by total gold out (grams probably)/total ore mined (probably tonnes) over say a year.

U
Unterstarm
8 years ago
Unterstarm 8 years ago

Thanks for the recent replies. The intention was to try and make representative 50g splits from ca. 200g of fine stream sediment. BUT then all 4 splits would be analysed to ensure all gold particles in the original sample are accounted for.

I must admit that in the scenario above I would go with a correctly used mini-riffle splitter for simplicity. I've always received advice not to pulverise (mill) sediment or a soil sample due to smearing/flattening - it is assumed that the gold grains in this particle fraction have already been released from the host.

K
Kumar Choudhry
8 years ago
Kumar Choudhry 8 years ago

When using Gy’s (1956) “General Preferred Sample Mass Nonogram” to plot a sample preparation and mass reduction scheme for drill samples – does the maximum ore particle diameter on the X axis refer to the sample diameter (i.e. RC chips, core crushed to 10mm etc.) or the largest (i.e. top size) of the ore mineral of interest? (e.g. largest graphite flake size, largest gold particle etc)
Does anyone have excel sheet that produces plots on a Nonogram?
When applying Gy’s sampling theory formula to industrial minerals does anyone how is the Fundamental Sampling Error (relative variance) calculated? Also are there published numbers for α and k for different types of minerals and deposits?Has anyone got an excel spreadsheet that will calculate sample mass using Gy’s formula?

(unknown)
8 years ago
(unknown) 8 years ago

Thanks for the question. I'm sure there'll be other responses but here is my response:

The size refers to the point in the particle size distribution that would need to be specified for the calculation to be effected..... It is usual to refer to a P95 (although this term has no international standing - the official ISO terminology is x95 although many others use D95 or d95). It's the top end of the distribution that's crucial in sampling and thus this represents 95% passing or 5% retained on a screen, for example. However you can still effect the calculations on other top points (x90 or x99, for example)

I would contact Francis Pitard to see if he has an appropriate sheet. I've always drawn these by hand (for the only one or 2 times I've actually done this)

FSE (it's identical to the term 'standard error' that math people use) would normally be pre-specified in order to calculate a minimum mass required to meet this required level of precision. The calculation can be reversed, though, if you know the mass of sample you take, the x95 (or other point - this provides a multiplication factor, aL) point, and the density (assumed or measured) of the particles. There's a webinar showing how easy this is:

December 14th, 2010 Sampling for particle size analysis - estimation of standard error

http://is.gd/baQkBj
http://is.gd/Rud8BX

Yes, I have spreadsheets that do this. But this is giving you the fish. The webinar will teach you how to fish and you will easily be able to construct your own....

In more detail (values for constants) then the books I'd recommend would be Francis Pitard's "Pierre Gy's Sampling Theory and Sampling Practice" and the chapter by Whateley in "Introduction to Mineral Exploration Charles J. Moon , Michael K. G. Whateley , Anthony M. Evans, Wiley (2005)

(unknown)
8 years ago
(unknown) 8 years ago

Good points. I took a look at the sampling tests - thanks for putting that up, very interesting. I will also take some more time on your brainshark/malvern/brunton story - again, thanks for making the effort.

I agree that milling (ring and puck or puck) will flatten gold as the rings/ pucks roll and hammer the material in the bowl during the milling process. However, flatter particles should be a little easier break up as they would present a larger and thinner surface to the abrasive cutting materials - which in this case, are the rock particles making up most of the sample material.

Tests I have seen, clearly show that gold can be cut or abraded in the ring/puck mills. These same tests did not produce a "homogeneous" sample, but did indicate that a few gold fire assay beads could be broken into many smaller pieces by this type of mill.

I guess, in your case, detecting the presence of gold is the goal. If he can put more gold particles into his sampling device, he improves his chances.

I just saw your reply. Was your objective to look into the practicality of taking bigger stream sediment samples which would result in you having to cope with a larger minus 80 mesh fraction ?

These are good points made regarding rsd's. Also slow and potentially expensive if you are dealing with large samples and wish to keep a high cut rate - multiple rsd's may be required.

One doubt I have, regarding the larger rsd's, is the achievable cut rate on the larger samples. Another is the now common practice of mounting the crusher directly to an rsd.

An rsd combined with a crusher effectively puts the operator in the position of having to match the rsd throughput with the throughput of the crusher. On the plus side, the cuts taken will be roughly proportional to the size of the sample as the crusher throughput will control the rsd feed rate.

On the negative side - crusher throughput, effectively now controls the feed rate of the rsd, subsequently limiting the number of cuts which can be taken.

Add to this that the crushing process may be faster than it once was - some commercial labs now crush in the range of 70 to 80% minus 2 mm where as a few years back many crushed to 90 or 95%. Less crushing is faster crushing, which in turn allows less time for the rsd taking a high number of cuts.

It is worth pointing out that many labs offer fine crush as an extra cost option - so if they use a combined crusher/rsd set up - in this case, with finer crushing, the rsd will also take more cuts due to the slower crushing process.

However, combined crusher / rsd systems also suffer from a much bigger problem - which is that there is very little mixing of the crushed material prior to the rsd process. Sample material falls directly from the crusher jaws, into the rsd feeder trough or belt. Crushers do not mix samples - it is a flow through process - first in, first out.

If anyone has trouble visualising the problem, try dropping a lump of quartz into the crusher - followed a few seconds later by a lump of dark coloured rock, then look at the rsd feeder trough, it will be apparent by the distinct colour separation of the material - that there is little if any, mixing prior to the feed entering the rsd.

If for example a 1 meter section of drill core had a 100 mm section containing most of the gold particles - assuming the scenario above, the rsd may only have a few seconds to sample the section of material containing the gold.

Rsd's should provide an improvement over Jones riffles - but it is interesting to ponder the limits and unintended consequences in operation.

Good comment re rsd - slow feed and a high rotation speed would be ideal - assuming more cuts taken, is good thing. My guess is that taking more cuts is a good thing as insufficient mixing and or potential segregation may leave the gold particles or other particles of interest in relatively small pockets or zones within crushed material.

If you can take more cuts during the period that the zones are falling from the feeder - you increase your chances of getting close to a more representative split.

With existing large lab rsd's (or at least any of the types I have seen) there is little possibility of increasing rotation to achieve more cuts - rotating tray types become dust pumps or batting devices as the rotation increases. The Rocklabs or Glen Creston type with a rotating spout distributing the feed, again seem to be limited to relatively low rotation speeds - not sure if the speed limitation is due to the increasing force of the throw at the end of the spout, possibly allowing some particles two or more shots at the exiting the sampling slot or bounce back of particles hitting the edge of the sampling slot.

I believe Rocklabsrsd can sample at around 50 rpm - so may be a good option for large crushed samples. If anyone knows of alternative large sample / crushed material capable rsd's with a higher cut rate, please let me know.

Progradexrsd (or a scaled down lab version) could provide an improvement over existing lab rsd's - it appears he has a practical solution to any dust pumping and potential sample loss. Although a variation of the rotating spout type,the Progradexrsd (or at least the model shown in the website animation) also appears to eliminate the potential problem of any particles having two or more shots at exiting a sampling slot.

It looks like the Progradex machine has no exposed edges (trays or edges of sampling slots) to cause bounce back of particles which would bias either the reject or the sample. Containing the fines and eliminating trays or sampling slots would help in achieving higher rotation.

Meantime, if the option of increased rotation as a method to increase the cut rate, is limited - the viable option will be a slower feed rate to increase the cut rate.

Running two rsd's per crusher would increase the potential cut rate by 100%. The initial outlay may give you a heart attack, however unlike some sample preparation equipment, most rsd's last a long time so cost per sample should be very low - over say 10 years.

In regard to segregation in an rsd using a vibratoryfeeder, I have no good info on this potential problem. I have observed some unusual flow patterns at times in the feeder trough - appear to be related to sample to sample, material variations such as particle size and density. You can also observe what is effectively a variable speed feed effect during the period when the hopper has emptied but material is still present in the feeder trough - feed rate can increased significantly. It follows that cut rate will decrease during this period.

Belt feeders should be relatively vibration free, so segregation, the variable rate feed effect and the subsequent cut rate changes, should be minimised.

I have no experience with belt fed rsd's. I would be interested to hear any comments on the pro's and cons if you or others have some insights.

K
Kumar Choudhry
8 years ago
Kumar Choudhry 8 years ago

Reading the posted comments - are we saying that the classic Riffle Splitters (Jones etc.) that are routinely employed on Reverse Circulation drilling rigs for mass reducing a nominal sample (typically 1 metre in precious metals and 2 metres in bulk commodities) down to approximately 3kg for submission to a commercial laboratory for pulverising to ~P80-85 -75µm prior to chemical analysis - are intrinsically biased?

I know that some gold mines that use RC grade control have gone to a cone splitter which has demonstrated reduced sampling error (variability).

Does anyone have published material (or other) showing that Riffle Splitting produces systematic bias?

(unknown)
8 years ago
(unknown) 8 years ago

The popularity of the 'static' cone splitter is interesting - as far as I have been able to fine they are based on the idea that an inverted cone placed in the sample stream (say a silo inflow or outflow) with the goal of reducing segregation in the stream by spreading out the flow over a wider area. However, if you think about this all you are doing is spreading the stream out and if there is segregation in the steam you still risk a sampling bias as the collection chute only takes part of the stream at the same locations. As such, the splitter is incorrect in terms of the sampling rule that all parts of the stream should have equal chance of ending up in the sample. It is always worthwhile looking at the potential for bias in these devices by collecting a duplicate from the second port, which is often available. Really need some sort of rotating sampler on the base of the cone.

Gruppen
8 years ago
Gruppen 8 years ago

Does anyone have excel sheet that produces plots on a Nomogram? I have one and I have another for Pierre Gy's rule.

(unknown)
8 years ago
(unknown) 8 years ago

With better than 90% of a heterogeneous sample going to reject much of the time in the mining game - expect systematic bias, with any sampling device, a riffle splitter being no exception.

Riffle splitter's (jones type) can range from good to very bad - take a look at the progradex website for example of the very bad (stacked or tiered riffle type).

A good riffle splitter should have many channels (riffles chutes) - the problem is that feed size determines the width of the channels, so with R.C. samples you can end up with a pretty large and unwieldy device. Large samples and an unwieldy riffle splitter make it tough for the operators to feed the device. Eratic or uneven feeding will increase the bias.

In regard to published material on riffle splitters: A critical evaluation of powder sampling procedures by A.A. Khan and T. Allen is on the web.

http://is.gd/bKtPZr

There is also a similar study done by Terence Allen ( unfortunately, I can't find it ) where he used samples containing from 1 to 5% mix of copper sulfate with sand - one of his conclusions was the efficiency of the sampling devices was improved as the percentage went up. I guess it could be inferred from this that efficiency would be reduced when dealing in the ppm range.

Keep in mind that the a.a. khan, t. allen studies were done using a riffle splitter with only 10 channel or chutes. If they had used a riffle splitter with 32 or 64 chutes, I expect the standard deviation of the riffle splitter would have been reduced.

The Khan/Allen study shows the rotary sampling device to be better than a not very good riffle splitter - however they do make some very good observations. In general, they make a very good case for rotary sampling devices.

The observations on the static cone sampling devices make a lot of sense. I would add that the flow pattern of material exiting the cyclone can look like a spiral, with more material exiting one side of the outlet. Getting a 360 degree spread of material on top and around the cone must be close to impossible.

(unknown)
8 years ago
(unknown) 8 years ago

The Allen/Khan material from University of Bradford days are well-referenced and explained in Chapter 1 of all the 5 editions of Terry Allen's "Particle Size Measurement". If I recall correctly all the data comes from a 1968 reference that is Khan's M.Sc thesis.

In terms of calculators then there's one called SMC-1 given away on the attached CD in BastiaanGeelhoed's book "Approaches in Material Sampling", IOS Press 2010. Copies available from http://is.gd/XnTLEX ($67 was the cheapest, so it's not that inexpensive).

Victor Bergman
8 years ago
Victor Bergman 8 years ago

This subject has certainly stirred a lot of interest and many very helpful comments have been made. I am, however, troubled with the general preference here for use of rules and “expert” opinion over the process of actually measuring the levels of precision of rotary dividers or riffles operating with the various choices available. It is worth the time expense of doing such experiments in order to assure the optimum choices of equipment and numbers of increments or divisions of the riffle are made for the application. It is my belief that we should be taking this approach with all sampling equipment, sample preparation equipment and analytical processes. If you do this and do it properly you will find substantial opportunities for optimizing these measurement processes while achieving your target levels of precision and bias.

(unknown)
8 years ago
(unknown) 8 years ago

Good comments. Yes, the sample-to-sample reproducibility is easily tested in practice and practical values tend to be higher than those that the literature. I agree - that's the best way,

In an early 1993 application note (well-dated now; I've learnt a lot since then) I did do a bunch of practical work on a spinning riffler and included these results: http://is.gd/bLXmyx

(unknown)
8 years ago
(unknown) 8 years ago

Your comment is good in regard to testing and working towards getting the best from sampling devices. In regard to your concerns on rules and "expert" opinion, my thinking on this, is that I would rather hear it, than not.

The rules, in regard to operating/sampling rsd's and Jones riffle splitters are of interest to me as I sometimes fabricate equipment or adapt equipment if I can think of a way to improve sampling.

Earlier comment flags up possible problems in maintaining fixed feed rate and rotation rates on rsd's - a lock out for feed and rotation settings may help, however drive belt (rotation) slip may be a factor on some rsd's. A time counter/stop watch on the feeder and a rotation counter may also help.

A rotation counter would also provide cuts per kg count - dividing the rotation count result by the sample weight would do the trick. Cuts per kg number would demonstrate some control over the process to any interested parties.

With most rsd's you have a choice of taking tens of cuts, or thousands of cuts per sample - or anywhere in between. Having seen both extremes in operation, it struck me that you have a potentially excellent sampling device and a potentiallyduff sampling device - the same machine does both.

Similarly, with Jones/riffles (but without the potential extremes, good and bad) - one lab may use a 12 chute riffle and another may use a 24 chute riffle.

I fabricated two vibratory 64 chute Jones/riffles a while back, one with 6 mm chutes for 35 kg crushed 95% minus 1 mm material and another with 3 mm for large 3 kg milled minus 75 micron material.

Although as pointed out, there may be some applications when Jones/riffles may be the only practical option - the potential for significant improvement may be limited.

Do you have any thoughts on why your values were higher than those?

I enjoyed your application note, I do not pretend to understand much of it, but I always learn something from the stuff you put on here. 

(unknown)
8 years ago
(unknown) 8 years ago

Thank you for your kind comments. To answer your question in relation to figures, I suspect the reason for his 'better' rsd's was simply that he was working with a model two-component system (that also involved no 'dispersion'). The numbers I show in the link I posted were obtained experimentally (tedious - 25 samples from our spinning riffler; 10 repeat - wet - measurements of each complete sub-sample) on a commercial purified terephthalic acid of pretty broad distribution. This PTA had fines as well as material up to 1 mm or so.

Please join and login to participate and leave a comment.