Rod mills - not much in favour these days - tend to produce (much) narrower particle distributions (as only the largest material is ever processed in any comminution circuit) but the rods wear and get tangled. In certain applications the grinding balls in a ball mill may give rise to undesired contamination. x95 (ISO uses x for particle size) may be a better control parameter (and would be ideal for sampling minimum mass determination) but we're all used to working with P80's in conjunction with Bond's work index.

The worst thing that you can do is to fall into the trap of focusing only on the p80. Do some research and look at recovery versus primary grind size on operating plants.

At the same p80 you could have the p20 ranging from 10-30um. Or say the % passing 10 um ranging from 10-30% at the same p80. If the recovery of these particles is 60-70 % then there is a significant impact on predictions for overall performance.

Classically rod mills are used in reputable laboratories that understand the need to try to generate a size distribution similar to that of full scale circuit. You are dealing with a multitude of process variables that affect the shape of the size distribution in the real work versus laboratories. The performance of a single stage ball mill in a laboratory in no way emulates that of the full scale plant.

I have seen this to be one of the key parameters which causes us havoc in prediction of performance. Remember that the chemistry effects of flotation become more pronounced at finer particle sizes so if the real world generates more of this material, and the physical equipment cannot recover them well, then you may well have a problem.

You are right, that is the problem I have noted from different metallurgical lab in concentrators, they are using only the P80 as the key parameter and they have different results from lab to lab using the same ore.

Opportunity exists to use rod mills in lab testing to more closely simulate plant grind with ball mills. Ball mills in lab can give too many fines in comparison with the plant. Definitely agree want to be considering the entire PSD & not only P80.

You have two questions here.

Firstly, the use of the P80 as single parameter to assess a batch laboratory grind comes from the Bond equations in which the 80 percent passing sizes for the feed and product are used for grinding power calculations. At times, the P20 is also quoted since the cumulative particle size distribution can be approximated by a linear equation between the 20 and 80 percent passing sizes; the slope giving an indication of the spread of the size distribution.

Secondly, while it is often observed that the particle size distribution obtained from a batch rod mill is narrower than that obtained from a batch ball mill; none directly mimic that obtained in an operating plant when looking at the individual mineral size distributions. The main reason is that in a plant, the final grinding stage is normally operating in closed circuit with a sizing device in which the mineral specific density introduces side effects - the hydrocyclone (the screw classifier in the old days). For gravity separation circuits, screens tend to be used over hydrocyclones for the simple reason that the valuable minerals are usually much heavier than the gangue and the hydrocyclones would lead to over grinding of the values.

One thing to remember is that each sample has a different mineral composition and physical property and hence a suite of samples will yield different P20s when ground to the same P80. This mimics what is observed in a plant especially when comparing low grade feeds with high grade ones.

We need to find some innovative technologies to get very closer particle sizes. In cement mill we use compartment mill. Can we think of similar technology with different grind media sizes, in different compartments? I feel roll crusher can solve our problem to some extent. Reduce ore to 1mm in roll crusher and then do regrinding, with open circuit. In my Research I tried in commercial operation one such problem. Cleaner tails of Zinc circuit was ground in a regrind ball mill with open circuit. Ball mill discharge fed to cyclone. Cyclone under flow fed to rougher, and over flow fed to scavenger cells. To my surprise results were very good. Tailing reduced from 0.4% Zn to o.2% and concentrate grade jumped from 52%Zn to 54%Zn.You too can try such novel innovative flow sheets.

Generally, I have seen that lab ball mills will produce more fines than lab rod mills. Other than look at just P80, I would recommend doing Rosin-Rammler plot and compare your overall distributions. Of course, if you are producing more fines, you will affect the performance of your lab float. The key is to "try" and match the actual plant distribution. You can always alter your ball or rod charge to alter your grind distribution, but that can turn into a research project on its own!

Of course, it is important to the P80 and P50 maybe, this lets you know if your grinding produces a thick or thin configuration is critical for flotation reagent consumption by the separation itself has polymetallic sulphides and hydraulic-mechanical drag.

What kind of laboratory mill used? USBM type? Perhaps the answer is in feed sample size? Follow the advice, the first thing you should do as a metallurgist is to calibrate the standard test. 

How are you testing for PSD? I market the HMK-Test PSA equipment which complies with the ISO and ASTM particle size test equipment standards but is a fraction of the competition's cost. With their AS-2011 unit you will know exactly what your PSD is within 60 seconds and you can run hundreds of tests per day if you like i.e. stop the mill when you have the PSD you want not what comes out.

You can achieve the same k80 using rod or ball mills, the main difference you might find is the generation of finer particles below the selected k80, maybe you can run some tests using the same sample on both mills, so you can evaluate the effect of grinding media on flotation response.

Different grinding media may influence the PSD. It is better to measure it. You can do sieving if the particles are not very fine. Laser diffraction particle sizes are a better choice if you are dealing with very fine particles.

Particles at different size fractions have quite different floatability. Intermediate particles are generally more floatable than either coarse or fine particles. So if your PSD is different, the overall recovery may be different. And since the total particle surface area is also dependent on size distribution, the collector consumption might also be influenced by the PSD.

First of all, thanks for addressed this subject there are some valuable comments here.

From my point of view, indeed PSD is important in lab testing for crushing, grinding, flotation and leaching test, also for Geometallurgy and some advanced techniques for mineral analysis. Rod mill are commonly used in reputable commercial labs because the normal distribution generated. Labs, commercial or not aim for an ideal environment minimizing bias, isolating some variables.

Second, PSD is one of the Key parameters in lab, pilot / demonstration or commercial plant but using it as a unique decision parameter is pretty unwise in any case as Stuart mentioned before - but it does happens, in commercial lab as well as in some operations

Finally, as miners, metallurgist, surveyors, geologists and process engineers we have to understand the role and importance of labs (commercials or base), what is the scope of any test /trial and its limitation, what are our expectations from the test and what the results mean - and how we can escalate them.

Interesting comments! My 'late to the conversation and thus probably already covered above' thoughts are that you can never replicate the plant in a laboratory (no matter which step you are testing), thus trying to exactly replicate the granulometry of the plant will never be possible. It can be argued that the plant itself is incapable of replicating its previous day operation.

Like most metallurgists who have working in a metallurgical laboratory, I have used both rods and balls at various times, and while it is true that they do not produce like size distribution curves, I feel the most important parameter by far is to ensure that your procedure does not change from one test to the next. Again this falls back to my previous comment that the lab cannot replicate the plant. If performing and laboratory campaign, ensuring that everything other than the parameter of interest remains the same is absolutely critical. If you start a project using rods in the laboratory mill, finish with rods in the laboratory mill.

The objective of laboratory testing is almost always looking at responses (such as recovery) to a different reagent, pH, pulp density, etc.That is, you are looking to compare a TEST condition against a STANDARD or REFERENCE in order to give you an indication of how the new reagent, pH, etc. will perform in the plant.

"Thus using balls or rods in your laboratory mill for the most part is not important as long as they remain constant throughout the project (this includes the ball and rod diameters)."

Good comments.

You are correct to state that the most important thing in a laboratory flotation test program is the consistency of conditions and operations with only the variables of interest being manipulated. Mind you, this gets a bit tricky when investigating collector dosages and frother types since these modify the froth characteristics and poor laboratory practices could inhibit the true impact of these conditions on metallurgical performance.

Following this, the laboratory procedure has to mimic the changes in grade-recovery curves (or operating point) in a fashion similar to the concentrator. Namely, if the concentrate tends to have a concentrate grade-recovery slope of -0.5% grade per 1% recovery increment near its designed operating point, then the laboratory procedure should exhibit a similar grade-recovery slope. Similarly for the recovery vs. head relationships being cascaded into the block model and ultimately the financials.

Finally, using balls or rods in the laboratory mill as well as the operating conditions of the flotation cell become an issue when performing confirmatory flotation test work or re-investigating findings from previous programs in a different laboratory. Unfortunately, the impact of these basic test conditions is often poorly understood.

I absolutely agree with this paragraph;

"Finally, using balls or rods in the laboratory mill as well as the operating conditions of the flotation cell become an issue when performing confirmatory flotation test work or re-investigating findings from previous programs in a different laboratory. Unfortunately, the impact of these basic test conditions is often poorly understood."

When looking at lab vs. plant, you should not solely focus on particle size distribution, but that also of the mineral of interest. Lab rod mills tend to produce a particle and mineral P80 of similar size (can vary when high levels of mica or similar), but the plant closed circuit mill with a hydroclone can give finer grind size of the heavier minerals. This ultimately might mean that that the plant grind size can be coarser than the lab grind size.

There has been some research into whether the particle size distribution for the minerals of interest in the laboratory procedure has to match that of the plant.

There are however a number of counter-indication for this:

•The gangue, which is often the hardest material in the feed, has to be also ground to the same P80. Thus, this cannot be used when a design is considered as the resulting power calculation would be inflated over the generally accepted design methodology. This also led to a very fundamental question - what P80 should be used for plant design?

•For froth flotation, entrainment of the gangue would be more severe as well as the genuine flotation of mineral phases with middle range hydrophobicity.

•By performing quantitative mineralogy, as well as size-by-size metallurgical balance, one could assess what might occur in the concentrator by using a modelling approach - and there are quite a few available.

Lab scale studies are only a guide lines for plant design and is based on a few representative samples, which itself vary in plant operation. So it’s necessary to design the plant based on lab/pilot scale studies with necessary consideration for variation both physical and chemical properties of ore and experiences gained from existing plant. There are already technologies available which need to be utilised properly. Ball mills with control of the variables like feed rate, grinding media, retention time with close circuit grinding with a classifying screen instead of cyclone may greatly reduce additional fines generation.

Although this does not really address the topic at hand, I feel that ultimately the discussion about matching the plant size distribution is moot when most laboratory grinds are completed in an inert laboratory grinding mill using inert grinding media at ambient temperature that in no way replicates the true conditions within the plant grinding mills. For me, surface chemistry is as, or more important than ensuring the particle size is accurately achieved.

I have seen on numerous occasions that changes in the grinding environment have a much larger impact on subsequent laboratory flotation performance than differences in grind size. I feel that in general, this is where commercial and onsite metallurgical laboratories have the most to gain in truly replicating the plant conditions.

The expression "inert" for laboratory grinding mills and grinding media has been used to describe many different grinding environments. Are you referring to anything else than a "Magotteaux Mill"?

I concur than surface chemistry is a very important parameter which is often overlooked during flotation test programs - e.g. how many do measure the pulp potential, or even the pH of the pulp after grinding/start of the first conditioning stage?

Although I believe that the Magotteaux Mill is probably the most widely known and developed laboratory mill capable of producing different grinding environments, I do not believe it is the only example out there. But yes you are correct; I was referring to the Magotteaux Mill given my years of experience with it.

This thread has all my favourite metallurgists beating their favourite drums. Despite being an old thread I thought it would be fun to join the bandwagon.

Optimizing a flot circuit is like tuning a radio. Yes using only the P80 is fine. However it is a coarse tuning knob. You can mess about with different machines with different bubble characteristics as well. But that is also coarse tuning. We are all trying to squeeze the last few drops of payable mineral from every last spec of dirt. We need to fine tune. But let's be aware that each ore can be different. I love Mike's mill. But some ore is virtually inert and would not benefit from tight control of mill chemistry. Ore can be mineralogically simple. Such that if your lab P95 and plant P95 were miles apart it would not change the recovery! I've seen some where you could change the P80 dramatically and could never measure a recovery difference.

As you troubleshoot your processes are aware of the tools in your toolkit. And make sure you are aware of the level of scope required to address your problem. Define your problem clearly. Which is to say in a manner that can be measured and disproven? This is important because the old timers processed all the good ore. We will have to apply some Thomas Edison to make money of the remaining dregs. If you can measure well enough because you can only turn the coarse tuning knob then increase the size of the sample or number of tests. For example I can't solve some items in a lab flot test because I don't have enough mass. Sample a small stream on a full circuit and do large scale tests. Then you will have lots of analyses available. Remember they didn't find the Higgs boson with a couple of tests. They needed years of repeats, hundreds of thousands, to statistically show it existed.

Use the right tool for the right job. Start wide and tune in. Spend your money in the right places. The ore will tell you where to look, just like the radio.

Since we are reopening this discussion I would like to throw out some cautionary advice to those doing flotation tests on ores with minerals such as native copper and nuggety gold which are malleable. If doing batch grinding you may find that when you collect the slurry from the mill there are still large pieces of the target mineral as they are malleable and can also potentially amalgamate. It is possible to do flotation tests on these samples and conclude that flotation is not feasible due to poor recoveries. When doing pilot plant tests with closed circuit grinding including a classifier such as screen or hydrocyclone however, all of a sudden the minerals start floating and acceptable recoveries are obtained since the larger particles will not go to the cyclone overflow.

Additionally, since native Cu and Au are significantly heavier than the gangue minerals not only are they going to concentrate in the grinding circuit of any mill with hydrocyclones, (likely requiring some form of bleed stream), but the particles going to the overflow are going to be significantly finer than the p80 of the rest of the stream. It is therefore important to look at the size distribution of the target mineral as well as the size distribution of the gangue minerals; this phenomenon is also seen to a smaller degree with sulphides and other minerals. A final note of caution, it is also very difficult to get pilot plant cyclones operating in a similar manner to full sized cyclones in a plant, this is a significant variable that needs to be accounted for when doing scale up.

My conclusion is that when working with nugget gold or large pieces of copper standard batch grinding followed by flotation kinetic tests can be misleading and send you down the path of gravity recovery when not required. If ground in the actual plant or even pilot plant, flotation may be a suitable mode of recovery particularly when these metals are not the only mineral valuable mineral in the ore body (such as native Cu in the presence of Chalcocite or Chalcopyrite).

Interesting that you thought of reviving the thread - welcome to the band wagon. I particularly like your suggestion of "Define your problem clearly". Well defined problems are so much easier to solve.

Although there is an art component to mineral dressing, mineral separation tests performed to the same conditions (grind size, reagent dosages, and flotation times) by different persons should converge to an approximately similar grade-recovery curve. Otherwise, how can we get investors and/or owners to "believe" the value of executing a project developing a mineral resource or improving an existing operation?

Your point about high density/metallic phases usually poorly recovered in batch flotation tests because the procedure does not fully mimic the full scale grinding circuit (operating in close circuit with hydrocyclones) is unfortunately too often overlooked.

However, if a heavy/malleable mineral phase occurs at a coarse enough size to be amenable to gravity separation, then this processing route, in combination of a flotation circuit, has a number of advantages over attempting to achieve full recovery by flotation alone. Heavy mineral phases tend to float more slowly than other ones due to fragile attachment to the air bubbles resulting in a much higher flotation volume than by combining gravity to flotation.

I agree with you that in all likelihood gravity will be the best recovery method for Gold and possibly native copper dependent on circumstances. When dealing with native copper or the native copper is in the stream with other easily floated minerals such as chalcopyrite then there are a lot of considerations including;

If you already have the flotation cells or are going to construct the mill with them anyway it may be more cost effective to only use flotation.

Native copper can be in concentrations of % rather than g/t and as a result a lot of copper needs to be collected which partially rules out batch concentrators such as Knelsons and Falcons as a lot of units would be required on extremely short cycle times. If using continuous versions of these machines then it may be necessary to do a lot of feed size preparation such as utilizing screens which are costly and space intensive. Multiple stages of gravity concentration including regrind may be necessary to get target con grade particularly if there are heavy gangue minerals such as Barite or there isn't high liberation of the copper.

Other technologies such as Jigs or tables may be suitable but in large throughput plants it might be necessary to have a lot of these or extremely large units which again could be space intensive not to mention requiring a lot of operator input.

Dense media separation may be suitable but is notoriously a large capital outlay.

A lot of spiral classifiers would need to be utilized for a high throughput plant and can be space intensive and operator intensive.

There are multiple other technologies or variations on the above that may be suitable including flash float and hydrofloat technologies for coarse grind flotation.

The nature of the circulating load in the grinding circuit needs to be understood as most of these technologies will only be treating a portion of Cyc UF or other similar stream.

There are a lot of technologies with different strengths weaknesses that should be investigated when looking at native copper in particular and flotation is one that should definitely be investigated properly. It is likely that a combination of technologies will be ideal for the best recovery at the cheapest price.

Particle size distribution has a high impact on your final outcome in terms of metallurgical performance. Commercial labs and labs at site have the tendency to overlook this parameter because they want to get results fast and because they actually haven’t asked themselves this question!! They just do and advance with no questions asked.

I would like to suggest that you determine the p80, p50 and p20 in your lab testing and compare those numbers to what you reach in the actual operation. You can manipulate steel composition in the lab to come up with p80, p50 and p20 that are similar to what you get in the plant.

Remember that the plant PSD is going to have a high variability so you need to understand that before you try to do changes in the lab.

Once you understand that plant variability go back to the lab and come up with steel media distribution that will give you the minimum, maximum and average p80, p50 and p20. Select the same ore and develop tests in duplicate for each one of these values (max, min and average). This will tell you what the influence of the PSD for your ore metallurgical performance.

Different ores will need different PSD to reach optimal performance. I've seen ores reaching optimal performance at 80% liberation but I have also seen ore reaching optimal performance at 50% liberation. So no one can tell you what is best for your ore.

Once you assess all of this then you need to come up with a standard lab procedure that mimics your plant performance

This is another step is skipped in the industry to save money, which is a very dangerous position. You save some thousands of dollars once, while you lose millions to tails per day.

Just to add to your comment, while rod mills are not often used in 3m or larger mills anymore due rods getting tangled, they are still useful in a laboratory mill in that the product being more "even" is closer to a cyclone ball mill product. If you are feeding a gravity circuit after the mill then testing with a laboratory ball mill is OK but there are advantages to using a laboratory rod mill so long as you plot the full grinding curve not just the p80. Shortcuts may enable one to complete a task in the time allotted by your manager but seldom can be relied on.