Yes, you nailed it, screens are becoming a popular alternative to cyclones as they get more robust and reliable. They sort the density concentration effect out, and can get a steeper cut too. Screen manufacturers have pilot setups to check it out for you. Otherwise jigs and spirals, flash float cells and so on can also be put on (a part of) the recycle.

Having worked at a mill where we used a screen as our primary classifier instead of a cyclone I can tell you that they arent quite as practicle as one would hope. We screened as fine as 150 um but had to continue to open the screen up to prevent blinding and increase throughput. A cyclone adds signficant flexability that is easy to adjust with pump pressure or replacement vortexes and apexes in addition to substantially lower opex due to know expensive screen panels to replace.

Did you try any screen modifications (slope, bed depth, deck profile....).  You raises a very valid point about dense minerals returning to the mill and being ground to slimes. Flotation circuits would prefer screened products rather than the flatter PSD products from cyclones.

Yes, micro screens have much higher CAPEX and OPEX and require "love and care", but if for example 30% increase in grinding circuit throughput is achieved and slimes reduced significantly, compared to cyclones, it will pay off in a economically acceptable time frame. Yes, P80 below 100 um can be achieved with micro screens although the required screening area rises dramatically at these sizes. And finally, final product dilution is concern in many applications.

For primary mill screening, for practical reasons, It is suggested that a screen can be grossly overloaded so that (say) a 2 mm aperture results in a 0.63 mm d50. You will find that even such a classification gives a superior quality of separation than does a high density, low pressure cyclone.
Assuming we can only use a single screen on a mill discharge (distribution is too difficult and takers too much head) then a single 3600 mm wide screen (say 7300 mm long) is about as much area as you can get (even that is difficult to get "spread' over 3600 mm). With help from your local screen manufacturer, a heavily inclined but linear motion screen can achieve a respectable bed depth. I have tried to offer the same solution for itabirite grinding.

A cyclone trying to do this job might consume 600 kW. A screen only (say) 3 x 22 kW. The energy people might make a note!

In ancient history (when my seniors all had less than 10 fingers because of using stamp mills) it was standard practice in the gold & PGM industries to user a gravity circuit in the mill circuit as commented by Jarrod. This could have been corduroys, jigs, strakes (plane or in barrels) plane tables etc etc. Later such things were replaced with centrifugal concentrators on all or a part of the mill discharge or classifier oversize. Also (see Jarrod ,s remarks) circuit float cells - for which there is some specifically designed and with "flashy" names.
Surely we are allowed to learn from the old-timers - take the finished (or nearly finished) product out of the comminution circuit as quickly as you can. This saves energy and much improves the chances of higher recovery

Indeed I think there is a definitive interest of getting finished product out as early as possible - and as Lean management methods explain, recirculation is a type of waste. There are improvements in milling specific energy and downstream recoveries that can be made this way.

I am surprised by the energy savings, I wouldn't have considered they would be so large.

I appreciate the comment about the flash flot, indeed a good tool in this case (downstream process is flotation).

Do you know of any studies or paper describing results in depth? I understand the limitations in CAPEX, OPEX, operability and footprint, but it seems it could be profitable in longer term. As I read also somewhere double cyclone nest could help.

Anyway, I tend to believe that although making comminution machines more efficient is important, part of it is wasted if classification is not as good.

May be replace cyclones with fine screeners in big plants (say 2000-5000 TPH)? In this case will be problems with screen numbers. A screening plant/building similar size to the flotation plant would be required for such high throughput. Instead of grinding/flotation process it would be grinding/classification/flotation process. Big shift in how we look at minerals processing. 

Screening better, then cycloning for closed-circuit mill. Yes, mill capacity is increase, circuiting load is reduce (more for gold ore),

Unless you take the same approach for fine screening. i.e. very inefficient screening in terms of steepness of the curve, but still much better than a cyclone. Very "overloaded" screens hence with an effective aperture much larger than the D50, hence > open area and > t/h/m width also because of "overloading" you still have a dimensional partition (un-affected by particle density) and have a more manageable screening plant. It would still be very many wide screens and a complex distribution system. Screens have fairly high capex but relatively low energy consumption. Opex is mostly screen media wear. Cyclones have high energy consumption AND fairly high wear item consumption from pumps/pipes/cyclones.

Also it depending on the properties of the dense particles, a flash flotation, magnetic or gravity concentrator on the cyclone underflow can remove them.

What is the process stream? Is it within the copper or nickel concentrate regrind circuits at Kevitsa? That said, I am not sure of the details, but if you're looking at cutting @ -100 um+63 um, my experience with Derrick Stack Sizers is quite favourable (though I would not go below 53 um).

1. You need to look at the preferred size distribution for the subsequent separation equipment. If it is say flotation which follows, then using cyclones grinding to the size range to recover the values, it will be possible to leave the lower SG gangue (usually) substantially coarser thus saving circulating load and grinding energy and giving better concentrate grades as the coarse gangue is not recovered! The use of screens would negate these advantages. The cyclone cut-point is set by virtue of the cyclone dimensions (particularly the diameter of the cyclone body and the vortex finder) and the %solids in the overflow to give the required maximum size of values for flotation as the primary way of minimising the production of fine values. See "A comparison of DSM screens and cyclones in grinding circuits treating complex sulphide ores", (Stewart, P.S.B. with Weller, K.R. and Sterns, U.J.), SME Comminution Symposium, Arizona, February (1992).
2. Some fines are advantageous. Fines enhance the viscosity of the slurry in the mill. This means that the slurry will adhere more readily to the balls This results in slurry being incorporated in the bulk of the grinding mass rather than draining to the toe of the mill. Higher overall grinding rates result from the enhanced use of active zones of the grinding media.
3. Further grinding of ore which is already fine enough for liberation and recovery is certainly generally undesirable. Fine gangue is readily entrained and downgrades flotation concentrate. Fine values float more slowly because they have less momentum for bubble attachment. There are several ways of dealing with this. Krebs used to market a device known as a cyclo-wash (invented by Kelsall) which attached to bottom of the cone of a cyclone. Water required to give the desired cyclone overflow % solids is added tangentially through this, backwashing the cyclone underflow, thus removing most fines. Another option is to retreat the cyclone underflow in another separate cyclone again paying due attention to the water balance. The underflow can also be screened but particularly for active screens this is likely to be a high CAPEX and OPEX option. An integrated multi-cyclone system was tested at Broken Hill many years ago. This worked well but required a fair bit of operator attention. Dual cyclones systems are probably enough. See "A practical multiple cyclone arrangement for improved classification", (with Kelsall, D.F, and Restarick, C.J.), Proceedings of First European Conference on Mixing and Centrifugal Separation, Cambridge, E5, 83-94 (1974).

From the references above and the references in their reference lists you will see that these matters have been subject to considerable discussion and serious research for many years!

Cyclones don't actually create slimes, the mill does that. But an improperly set up or selected cyclone for the duty can certainly contribute to creating slimes by circulating undersize back to the mill.

Selecting appropriate cyclone configurations and operating pressures depends on several factors including sg of solids and water (carrier fluid), particle size distribution, solids concentration, volume flow rate and pulp viscosity.
This is why modelling cyclone performance is so important. Any number of of modelling packages are available and I have used JKSimmet with success. There are lots of papers on cyclone as well. 

You can try the AusIMM library for starters and search for Bougainville Copper.
I have not had direct experience with screens closing mill circuits but can imagine that volume flows and high wear rates may be an issue here. You can check out US Steel in the Iron Range for examples of Derrick screens closing mill circuits on magnetite.

There are also some more efficient mass separation devices such as Hydrosizers which use a 'teeter' or fluidized bed technology which significantly reduces 'a' bypass (fines returning to the mill). As states Bougainville experimented with different configurations of Cyclones (Flat bottom and horizontal) back in the late 80's before they were forced to close.

Fines in the <20 micron range usually go where the water goes in just about any wet classifier. For cyclones, using iron ore numbers, keep the feed % solids low, about 40 to 45% solids, and size the apex to keep the underflow % solids high, but not roping, about 75% solids give or take. That's about the best a cyclone can do. You can adjust these % solids by keeping the volume % solids the same for your SpG solids compared to 4.3. This also seems to apply to the new design cyclones made by the companies starting with K and W.

The fine screens made by the company that starts with D are very good compared to cyclones at getting very fine material out of the circulating load, often letting only 5 to 10 % of undersize in the feed report to underflow, compared to 15 to 20% of the fines in the feed reporting to underflow for a correctly operating cyclone. This assumes relatively new decks, good feed solids control, and correct screen operation and loading.

The fine screens used in Minnesota are primarily used to kick out large high silica (low density) particles that report to cyclone overflow. These particles often still have some magnetite attached and will otherwise be kept by the magnetic separators. They are returned to the grinding mill to be further liberated from the magnetite. Doing this allows a lower silica concentrate to be made at the same %-43 micron grind, with fewer ultra fines.

We can definitely use fine screens of Derrick. It will efficiently classify the fine particles in the size range of 20 micron. By the use of these screens circulating load of Mills will come down by 350% to 200% depends on the situation. It not only helps in decreasing the creation of undersize as well as size of the Mills will be optimized. Lot of Power will be saved by selecting suitable size of Mill.

True indeed, fines go with water. However a sharper cut - and as you say a reduction in the ratio of slimes mass in UF/water mass in UF - is always positive. Same applies for anything which is in the floatable range (when downstream process is flotation of course). 

Interesting application indeed.
Anyone would have some link to Bougainville litterature? Would gladly appreciate it!
Anjani, interesting, would per chance this application have been published or some data to share? What is the footprint of those screen, versus the throughput they are fed?

With the use of Derrick screens you will get a circulating load from 80 to 100% Low circulating load will greatly reduce the production of slimes and make energy available for more throughput or reduction of number of mill if there is a large installation with multiple mills.

You have to balance foot print, capital expense of the equipment and the structure to support it, wear rates of consumables, power consumption, other operating costs and maintenance down time and operator ability/training with what you are attempting to achieve to make a value judgement. Both options have their place and in the wrong application, both could be catastrophic. So don't discount either cyclones or screens until you have modeled a pro's/con's scenario and a business case. I was lucky and had the un-biased luxury of marketing both in a previous life.

Separation efficiency of cyclones depend on how the cyclones are operated and controlled. This means first selecting correct gear for the specific application. Then decide on the operating / control points (pressure, no of cyclones). To do the job as efficiently as possible, even a given cyclone circuit has to have the right attachments (apex, vortex finder, inlet size). Then also the right wear parts. A lot operators do not change the wear parts on time. This contributes to inefficient cyclone operation. So even with cyclones, there is a lot operations personnel can do to minimize bypassing of slimes to CUF. Having said that cyclones are inherently less efficient than fine screens (which may be a high capex solution). In my experience in one special case, it was possible to reduce the recirculating load from about +400 % to about 200%, without screens.