How did you measure effectiveness? Did you bog the mill after a liner change out because somebody didn't reset the microphone correctly? I've heard of that happening a number of times with load cells. Are you comparing to some planned capacity rate? Is the controller cycling the tonnes and letting the mill internals get smashed up?

Mills are integrators and to control them you need to measure the inventory. But, there are a lot of details to the control scheme other than the sensor that is used to measure the inventory. Could you share anymore information?

While you were typing your comments I was typing mine, see above for a bit more info on how it worked.

The problem we had was that the mill never seemed to fill up enough to carry out efficient grinding. As our target was a certain percentage micron passing P80 we could never meet the target. Once going back to a more traditional way (mill weight and speed) we were able to meet targets.

It seems to be a Hardinge Electric Ear revisited. This device was patented by Hardinge in the thirties to be used in a ball mill with fixed speed and an almost constant volumetric filling level, given the appropriate ball make up. The load remains approximately at the same place all the time, such that with only one sensor you cope all conditions. The only one varying condition is the amount of slurry in the interstices between balls. The rms sound level is inversely proportional to the amount of damping exerted by the slurry. High level, low noise. Low level, high noise.

In a SAG mill, most of these conditions are not valid. Speed changes, such that you may get confused by the noise coming from grinding media impacting mill shell superimposed to the standard load noise. As the global volumetric filling level is also ever changing, your only sensor may easily be fooled by the change of toe position. The actual change of mill filling is now superimposed to the change of relative position of your sensor with respect to load toe. As these factors add up, the feedback to your controller may turn crazy!

That may explain why we could never fill the mill up to achieve a decent grind.

I have seen many sound control inputs over the years and the science seems valid in most cases when the ore body is consistent. Load bearing pads and power draw still usually is the primary data for adjustment, with sound evaluation being another check.

As a standard check we do listen to the mill first hand at the feed and also the discharge end to get some idea of impact levels. Also if the media is targeting the toe of the charge or there is too much ball on liner/lifter direct hits. After experiencing this "listening" over time you can determine these different sounds such as ball on ball, ball on ore (toe or center of charge), and ball on lifter/liner impacts.

Over the years our philosophy has changed, with the one exception of Africa where mills were commonly running at 90% of critical speed to be gentle with incompetent brittle media. The change is to not discuss mill changes in operation, but to design a purpose built media for the equipment and meet the existing operational work practices.

These days a more aggressive, higher impact environment is well within the media design capabilities without fear of breakage.

P80 targets are also more easily achieved using calculated blended sizes of media of matched grades to control the active ball size population within the mill.

Just a quick story - Whilst on a trip to South Africa after a mill tour I returned to the Mill Managers office for a short debrief. His office was directly over the SAG mill, it had to be one of the noisiest offices I have ever been in with obvious vibration emanating from the SAG. I told him he should move - he said NO! "If anything changes in that mill I feel it immediately through my "seat".

Yes, back in my early days in milling you could easily tell if the mill was bogging even if you were working up at the crusher because of the muffled sound emanating from the mill. Back then the Mill was totally steel lined and sounded like a freight train. I place a lot of emphasis on sound when operating.

The sound signal can often be processed to get better control. I've used a 1/3 octave simple band pass filter at 2kHz but I know that others have used more complex approaches. I inherited a good system at Parkes and couldn't find a reason to change it.

In the example you've given how was the mill feed-rate controlled? Overall, it seems that the control system ended up controlling the speed, density and feed-rate to the wrong set-points. I'd call that an optimisation problem (but I might just be being pedantic).

The Control philosophy controlled to a mill load set-point by automatically adjusting the ore feed rate, and the acoustic control controlled the mill speed.

This is all a commissioning exercise and one of the things we wanted to do was - incorporate an additional loop whereby mill speed is automatically adjusted to control mill weight/load. This acoustic control was used when running the Mill in cascade, when in Auto the required speed of the SAG mill was entered by the control room operator as a set-point to the speed controller. And you are correct it was an optimisation issue which we were trying to address.

As you know, there are a number of dependent and independent variables in running a SAG mill, sound is one of the dependent variables that gives information on what is going on.

I don't believe it should be used alone to control a mill.

Good quality well maintained electronic ears work quite well in indicating load LEVEL in the mill (as does load measurement but this is not a true indication of load level because it is prone to ore SG changes). If you are controlling mill speed based on sound you will never achieve steady state because all this does is allow the mill to fill and empty repeatedly. Best result is achieved by aiming for highest possible speed and controlling feed rate to allowable sound levels, in other words if the sound goes up instead of slowing the mill down you increase the feed-rate, for this to work effectively you must remove load as an operating constraint and use it only as an alarm situation. The other advantage of aiming to operate at the highest permissible speed levels is maximum throughput. 

In one mill we used sound to adjust mill speed, without much success. As discussed above key issues were generally the ability to effectively differentiate sound when media contacts media or grinding media hits toe of charge (at high ball / rock charge), or media contacts liner or large ore particle contacts charge. There were some good correlations, but not reliable enough to use mill control. In one 36 ft SAG we used load cells for weight control (to adjust speed), with feed rate control and power control were used in conduction with the speed control. Worked excellent. There was an AMIRA project to develop some instrumentation to analyse sound generated by the operation of mills and use it for mill control. I am not sure how far they went.

I have seen several locations try to run a control strategy with acoustics and many were unsuccessful. They can work, but they have to be set up right and maintained properly. The way they are used in a control strategy is also very important. How are you using the information in your strategy? Is it used to control feed rate, speed, both?

Honestly though, there are more effective methods hitting the market. Products like Mill Scanners have been proven to be much more useful for maintaining tighter control.

Remember tight control is the key to optimizing throughput. Consider recovery as well, it is likely much easier for the controller on the recovery side to react to small steps in tonnage then large swings. You want to keep your system running as tight as possible.

Acoustic control was used to control mill speed.

I might point out for everyone that it one of a few options on Mill control, one that we felt was, as mentioned already by others, not that effective- worked, but not that well.

From all I have read in these comments, the problem with achieving the proper load in the SAG mill is not the 'sound' controller but the 'retention' of the ore in the mill. I suggest that you look at the discharge size of the Pebble Ports. I had a client with a similar problem and easily solved it by this means, successfully achieving the design throughput.

I agree with you totally. Plus there are other additional benefits that must be mentioned. Port discharge should have magnetic separation of the rejected media after screening but before cyclone - this saves cyclone wear/damage and with the newer through hard media designs competent/spherical rejected media can be reloaded into the secondary milling process.

Benefits:

1. More space in the SAG for additional ore.
2. Smaller media that has not enough size/energy to contribute to that primary process.
3. Savings to secondary steel media. (We have seen 20% steel savings).
4. Cyclone maintenance.
5. Throughput improvement.

All good points Mark and very valid. A screen would also work to take out the undersize steel balls which can be used in the Ball Mill - part of the savings you mentioned.

Magnet can be before the screen. This done in SAG applications but I can't see why it wouldn't work. I suppose that the chute that takes the susceptible material might need replacing each shutdown if it’s getting hit by 70mm media that passed through the pebble ports. But when you have to reline the whole mill anyway what is the problem with replacing a small chute as well?

It is interesting that the topic of "poor control" often ends up with the circuit actually needing some new equipment. There is a limit to what control can do!

An interesting concept to use a magnet. Any ideas whether it would be a permanent or electro magnet? Perhaps an electro-magnet would be better since the field could be turned off to discharge the ball load?

There are a few magnets that I have seen, but the best is one that is over the ported ore stream after screening and before any cyclone. These magnets are electro powered with a belt that takes tramp steel off the belt and "flicks" it to a collector bin.

Operational data: http://www.rockproducts.com/technology/material-handling-a-conveying/12222-removing-tramp-metal-using-conveyor-magnets.html#.U584ncJOVqM

The collected reject media these days should remain round and be reloaded into the ball mill as fully competent steel. There should be no media waste at all.

As I explained before, mill sound in a SAG mill may depend either on mill speed or mill filling. We did develop an instrument called "impact meter", which has been marketed by FLSmidth. The device is intended to recognize the specific sound pattern of balls hitting mill liners, in order to avoid such operating condition. We did use three microphones to overcome the effect of changing load level, having always one of them near the toe. We did not develop a specific control product, just made recommendations based on the mentioned findings. You may use the RMS noise level as an estimator of mineral load inside the mill, given that the speed is below the threshold for direct liner hitting and that you have multiple microphones to adapt to variable toe position.

This does make a lot of sense to have tuned sound collection with the important pinpointed sounds of "interest" being flagged. My hearing is really taking a hit from doing this first hand over the years!

We use sound as part of our strategies, but the strategies include bearing pressure, mill power, mill density, recycle pebbles, and screen amps.

Sound when used correctly is a good process variable to help protect your mill liners/lifters, minimize steel depletion, and increase throughput. I'll email you some detailed strategy info for and some info about our Mill Scanner listening device.

Never a waste when new information is offered!

It is hard enough to stay updated with new technologies on a day to day basis - another reason why our industry is slow to respond at times.

Mill Ears - that brings back memories. Love them. But my favourite was the initial work in South Africa with "Instrumented Liner Bolts" (Kloof Gold Mine SA, Mintek) measuring impact and conductivity (indication of viscosity).

In the early days, before wireless signals, we would start the mill with long wires to the liner bolt, record data, and the "test:" would end when the wires snapped.

I haven't read the threads above - they will be more intelligent than mine, but if I was doing a mill survey or research project that I wanted real data from I would like you on the team.