SAG mill

SAG mill

An increase of over 10% in mill throughput was achieved by removing the ball scats from a single-stage SAG mill. These scats are non-spherical ball fragments resulting from uneven wear of balls with included porosity. 30t of scats were removed from a total charge load of 70t. Surveying and modelling the mill revealed that the breakage rates had increased dramatically at the coarser end of the size distribution. This work highlights the importance of maintaining a competent ball charge in a mill.


Navachab is an open pit operation and a wide range of ore types are mined. Mill throughput is sensitive to changes in the ore type and a collaborative effort, between mining and process personnel, is required in order to maintain acceptable processing rates. The current mining plan often results in frequent changes to the ore blend as received by the metallurgical plant. Operational tests have shown that the mill is not capable of processing the harder ore blends at an acceptable rate. This is becoming critical as the mine moves into harder ore zones in the pit.

As part of the P9 AMIRA site work for AngloGold, a number of circuit surveys were conducted on the single-stage, run-of-mine (RoM) mill, at Navachab Gold mine in Namibia. The objective of the work was to test, and possibly refine, the JKMRC approach to blending a mixture of ores with very different breakage characteristics into the JK SAG model. A total of seven circuit surveys were conducted. The composition of the feed to the mill was altered for each of the first five surveys in order to measure the response of the circuit to changes in the feed composition.

This is part of the overall research program aimed at better understanding and modelling the high load and speed, and low aspect ratio of South African style SAG mills, Powell et.al, (2001).


The circuit consists of a single-stage, low aspect ratio (diameterlength = 0.5) mill, typical of the majority of SAG mills in the South African gold mining industry. These are locally known as run-of-mine, RoM, mills. The mill is operated at 90% of critical speed and with a volumetric load of 40%. Unlike the majority of RoM mills, it is in closed circuit with a cluster of cyclones, as opposed to a large single cyclone, and a recycle crusher, Table 3. The interior of the mill is illustrated in Figure 1, showing the length and filling of the mill, and low lifters on top of standard AMS grid type liners. The detail shows sections of the discharge grate (18mm round holes extending to near the centre of the grate), and pebble ports (75mm round holes).

The trommel oversize, plus 11mm fraction, is conveyed to the recycle crusher. A belt magnet is used to remove the steel, which was previously dropped into a bypass chute to rejoin the crusher product. A rapid response pneumatic driven bypass chute, activated by a metal detector, is used to protect the crusher from any metal that escapes the belt magnet. The trommel undersize stream is pumped to the cyclone cluster where the underflow rejoins the mill feed and the overflow gravitates to the thickening section.


Whilst conducting the circuit surveys, it was noted that there was an excessive presence of ball scats. The balls are of the cast semi-steel type, typical of those used in the South African gold mining industry. These balls suffer from porosity, the result of core shrinkage during rapid cooling in the mould. The porosity is not central due to uneven cooling of the balls. When the ball is worn to the point where the porosity is exposed, it wears very rapidly into the porous section to form a bowl shaped lump, that soon wears to a plate, Figure 2. These oddly -shaped scats are then competent and wear down at the same general rate as the whole balls.

Evidence of Excess Scats

Mill charge

When inspecting the mill charge during the crash stop at the end of each survey it was observed that the scats represented a substantial portion of the charge surface layer. A few of the many in evidence are highlighted in Figure 3.

Crusher feed belt

It was noted that there was a lot of steel on the crusher feed belt. This was a problem for sampling the belt to obtain the crusher feed size distribution. The steel had to be manually separated from the pebbles at each screen size when the sample was screened. The first sample was screened early on in the test program, and to the surprise of the investigators, it was found that 30% of the crusher recycle stream consisted of steel. This equated to a 10 tph recycle rate of steel. For tests with higher recirculating loads the steel recycle rate increased to 17tph, but remained below 35% of the total pebble recycle stream. Crusher bypass

It was noted that the high recirculating load of steel resulted in the crusher being bypassed rather frequently. This was most undesirable for sampling the performance of the crusher. The bypass periods were timed over a 15 minute interval and it was determined that the crusher was being bypassed for 1/3 of the time. The utilisation of the crusher was thus low and uncrushed pebbles were being returned to the mill.

Crusher reduction ratio

Despite sampling the crusher product only during times that the crusher was being fed, a large percentage of coarse material was observed in the product. The sized data confirmed this, Figure 4, with the reduction ratio, calculated on 80% passing size, being a poor 1.4. This was a consequence of the maintenance foreman setting the closed side setting to 18mm, and the eccentric throw to a wide 34mm, to protect the crusher from the inevitable occasional balls that bypassed the belt magnet. This, along with the low feed rate of 25 to 35 tph, resulted in the crusher not being operated in a choke fed mode, thus producing a coarse product.

Mill grind out

In order to measure the ball load in the mill, the feed to the mill was stopped and it was ground out to a ball charge only. As a result of slurry pooling in the mill, the surface of the charge could unfortunately not be inspected, and the 9.7% ball level was measured relative to the slurry, 398mm below it. However, the evidence of the scats was obvious from the recirculating load of steel that accumulated in the inlet trunnion, Figure 5.


It was concluded in detailed studies of ball size distributions in ball mills conducted by Vermeulen and Howat (1989), Vermeulen (1989), that these scats contribute very little to grinding, but rather draw power for little productivity. Based on the flow rate of scats in the recycle it was estimated that at least 20t of scats could be removed. There were a few competent balls noted at the top end in the recycle stream, but it was decided that it was worth their loss to extract the scats.

Removal of the scats was simple to implement. The magnetic scats removal belt was turned around so that the scats were ejected to the other side of the belt. A chute was placed on this side and fed down to an area accessible to a forklift, where a 44gal. drum was placed. It was planned to add an equivalent tonnage of new balls to those removed, so as to maintain the total ball load in the mill. Stocks of replacement balls were potentially a bit low for this, but it was decided to launch the trial immediately while the investigators were on site.


The rate at which the scats were removed was as expected, but the overall amount extracted exceeded expectations. A total of 25t of scats were removed in the first 8 hours, Table 4. Figure 6 illustrates the rapid drop-off in the rate of extraction, and that scats were still being extracted at a rate of nearly 2tph after 8 hours. A total of 30t were removed in the first day after which only a minor amount was observed in the recycle stream. Only 12t of fresh steel was recharged in the first day, due to concerns about supply. The Operators applied an increased ball addition rate intermittently thereafter, to top up the ball load.

From a 2.5t drum of scats 127.5 kg were classed as competent to borderline competent media, representing 5% of the scats. This low percentage loss of competent media supported the conviction that it was worthwhile removing the scats even though some balls were included in the rejection stream.

Ball size distribution

The ball load in the mill, prior to the removal of the scats, was calculated from the volumetric filling, taken from the full grind-out to be 70t. The average scats size distribution was taken from the 6 test runs, Figure 7.

This shows the measured scats size distribution and the reconstituted full ball size distribution, based upon the measured mass of scats and calculated full ball load, with the 70mm point interpolated. The steel scats particles were typically non-spherical in shape, it was expected therefore that the overall distribution would not be smooth.

The Std. JK size distribution is that taken from ball size distributions measured from dumping and sizing the whole charge of production SAG mills, Morrell (1993). The mill with scats had a considerably finer size distribution, and notably lacks ball in the 40mm to 60mm size range. This is the size at which the porosity is exposed, and they suddenly become smaller and non-spherical. For a normal finer ball size distribution a finer grind and lower throughput would be expected.

The estimated size distribution in the mill at the time of test 7 is based on the approximations that there is almost no media smaller than 55mm, the top size of the scats; and that the mass in the 55mm to 71mm size range remains steady from the pre-test conditions. The mass in the top size is increased by the extra ball addition, 16t. The total mass in the mill is the 40t remaining after scats extraction plus the 16t extra addition. It is assumed that the remaining balls wear at the same steady state as previously. This much coarser size distribution would normally be expected to give a far coarser product size distribution than the finer distribution prior to scats extraction. However, as seen from the performance data, this was not quite the case.


A survey (test 7) was conducted four days after the scats had been removed, to compare the mill performance. The crusher now operated without any bypass, thus allowing a utilisation of , and no steel was found in the pebble belt sample. The results of the test are compared with that of test 1, which had the most similar feed composition. At the time of conducting the comparative survey, the ball charge was not yet up to the original filling. A total of 24t had been added over 4 days, which is only 16t greater than the scheduled amount for that period, leaving the mill 14t low on balls. The ball size distribution would also not have been in equilibrium by then, as it had been depleted at the fine end, and had received a substantial amount of top size balls.

The test conditions are summarised in Table 1. The comparative data is given in full detail in Table 6.

In the post scats test, the reduced ball load was calculated from the difference between extracted scats and replenished new balls. It can be seen from the slurry level that slurry pooling occurred during the post scats test. This pooling introduces inefficiency into the mill and results in a coarsening of the mill product, Morrell and Latchireddi (2000). The slurry pooling and lower ball load account for the lower power draw. Since the operation of the mill had not been retuned by the time that this test was conducted, it is expected that, in the long term, a similar slurry level to that of the other tests would result. The overall effect was an 11% increase in throughput, accompanied by a coarsening of the grind from74f.imto84f.im.

Grinding rates

What was of particular interest, from a fundamental understanding of SAG milling, was the relative grinding rates. These were fitted using the JKSimMet SAG model, Figure 8. The post-scats test has a considerably higher breakage rate at the coarse end, is a bit lower at the 2mm peak rate, and very similar at the fine end. The higher rate at the course end would be expected from the coarsening of the ball charge, but this still occurred with 14t (2% volumetric total) less balls in the mill. The reduction of the breakage rate in the 2mm range would be expected from the removal of the media in the smaller size ranges. What is very different from a standard ball size distribution is that for a very dramatic change in ball size distribution there is a relatively small drop in the production of fine material. The equivalence of the breakage rates in the sub-100f.im range indicates that the fine irregular media in the mill does not contribute significantly to the production of fine product. It is expected that a survey conducted later under equilibrium conditions would show the new true breakage curve.


The performance of the mill over extended periods prior to and after the extraction of the scats, provides perhaps a more realistic comparison of the effect that the scats had on the milling process. Daily production data, supplied by the plant, was used for this, Figure 9. The leap in throughput, immediately after the removal of the scats, is obvious and sustained.

The performance data is summarised in Table 2. An 8% increase in throughput is achieved in the first month after the removal of the scats, accompanied by a 2% reduction in the production of -75j.im material, and a slight reduction in the specific power consumption. This is achieved on an average ore blend that is slightly harder than before. In the longer term the increase in throughput averages 5%, as the feed to the plant increases in hardness. The plant records indicate that the hard ore component was increased from 46% to 58% of the total plant feed. The data show that the plant was able to achieve a 5% increase in throughput, whilst increasing the average ore hardness of the feed. It had been ascertained in the surveys that such an increase in hardness resulted in about a 5% drop in throughput for the same feed size distribution. From this is can be surmised that the sustainable throughput increase is nearer 10% for an equivalent feed. The long-term trends over pre- and post- 6 month periods are shown in Figure 10.

On-going optimisation

The Navachab plant has further options to improve the fineness of grind and throughput. Slightly smaller, 65mm wide rectangular pebble ports, will retain the 5% top size balls that are currently being discharged from the mill, without adversely affecting the pebble discharge rate. This will help counter the increased ball consumption, and give the mill a slightly finer ball size distribution, which will be beneficial to the fineness of grind. The mill may well ball size distribution to develop, and improve the fineness of grind.

Retiming the cyclones and ensuring that they all have equivalent sized spigots will give a sharper product cut. The absence of the recirculating load of steel scats has greatly improved the utilisation of the crusher, as the bypass chute was observed to rarely be in operation. As recirculation of steel is no longer a problem, the crusher can be wound down to a smaller closed side setting, and eccentric throw, to improve its reduction ratio. The very strong function of throughput on feed size makes this plant a prime candidate for a full mine-to-mill optimisation of blast size. The original objective, of better understanding blending optimisation, can add the final touches to a well-tuned milling operation. The bottom line for the plant is that the mill can now process the full throughput for the hardest blend expected from the pit. This has negated the need for an envisaged plant expansion program to deal with the harder ore blend that is being received as the pit expands.


The presence of excess irregular ball scats in a mill is detrimental to its performance, with an 11% increase in throughput being measured in this test work once the scats were removed. If the mill has pebble ports, these can be used to good advantage to extract the steel scats. As very few single stage RoM mills have pebble ports however, this is generally not a practical option. Poor quality cast balls with porosity result in the production of a significant excess mass of steel scats, measured at 30t in this instance which was 43% of the total ball charge. If such balls are being used in a milling operation, it is strongly recommended that the use of higher quality balls be investigated. The economic viability of using more competent media, such as forged balls, should be addressed. That totally depleting the mill charge of minus 55mm media only resulted in a 2% drop in fineness of grind indicates to the investigators that a ball with a soft centre that wears away rapidly would not be detrimental.

This study highlights the usefulness of careful and comprehensive measurement of a circuit in finding areas for improvement and optimisation that tend otherwise to be overlooked.

Categories: Grinding Mill Tags: , , ,

Products Quick Links