Development of the non-contact acoustic array approach to answer key questions for semi-autogenous grinding, autogenous grinding and ball mills with a view to diagnostics, control and understanding

### The challenge of tumbling mills

Comminution using energy intensive tumbling machines are the mainstay of mineral processing plants, especially in the hard rock sectors that include the recovery of gold, platinum and copper. The types of mills include the semi-autogenous (SAG)/autogenous (AG) mills and the subsequent ball mills to downsize the feed ores so that the separation of valuable from gangue can be achieved. Information concerning these unit operations has conventionally been obtained by conducting surveys of the feed and product distributions, paying attention to sampling so that the results are unbiased, and ensuring that steady state operation has been achieved by taking note of all the measurements recorded by a distributed control system (DCS) such as power, speed, throughput, etc. Over many years, careful and concerted studies have been conducted in this way to enable a picture of the inner workings of the tumbling mills. In recent years the use of positron emission particle tracking (PEPT) and the discrete element method (DEM) have revealed further insight into how these mills might operate. However, these latter approaches are limited by the level of insight, assumptions and the scale of the laboratory experimental work. PEPT studies are often done on mills that are much smaller than industrial mills. The resultant models have been successfully used to evaluate grinding optimisation, mill modifications and new applications of mills based on pragmatic criteria and rock characterisation, within their uncertainties (Napier-Munn et al, 1996).

The approach described above is limited to steady state, bulk analysis of the unit operation and does not necessarily provide dynamic information that may be used for a more comprehensive understanding of the process flow sheet and hence could be used for advanced automated control. This has not stopped attempts of control of grinding mills, but efforts have been limited to ensuring that that mills do not have periods of instability whilst maintaining throughput – until the next major disturbance comes along. Recognising this situation has encouraged exploration of real-time measurement of the dynamic processes that can occur inside a mill. These systems have included instrumented strain gauge bolts, piezo measurements on the side of the mill, analysis of the power signal of the mill, etc. Non-contact acoustic measurement (Pax, 2011) provides a unique opportunity to investigate the internal dynamics of a tumbling mill, since it is the particle interactions that give rise to the ‘noise’ being generated.

### Acoustic measurement using 12 stationary microphones

Non-contact acoustic measurement using a microphone array determines the behaviour of the particles inside a tumbling mill by analysing the dynamics of the signals generated by the diversity of physical interactions that these particles undergo. Since the microphones are stationary, they listen to the emitted sounds in the same frame of reference as the charge of the mill. Any component of the measurement signal generated by the rotating mill is eliminated. The microphones used are very directional so that background sounds are attenuated and irrelevant when compared to the sounds emanating from inside the mill. Acoustics is generated by the dynamics of the steel balls and rocks inside the mill. The particle dynamics inside the mill has been shown to be influenced significantly by the feed particle size distribution, feed rate as well as the water addition rate, mill physical operating conditions and the composition of charge inside the mill. In essence, the machine acoustic measurement approach can be likened to listening to music and discerning the sounds from the individual instruments. The fast (80 ms) measurement has provided new insight of internal processes of a mill. The acoustic determinations demonstrate the sensitivity and value of using microphones for real-time measurement.

At MMG Century mine, a twelve microphone array was installed around an 11 m diameter SAG mill at a longitudinally central position, avoiding a central circumferential flange (Figure 1; Pax et al, 2014) . The signals from each microphone were routed through an amplifier and digitised before being fed, using an optical link, to a computer executing the SmartNoise software. The software controls the data acquisition, analysis and communication to the Century DCS. From the acoustic data and validation using the process data, information was obtained concerning the influence of feed water on slurry transport and ore grinding as well as mill stability (using the water hold-up signal) (Pax and Cornish, 2016a); the effect of feed size variation (Pax and Cornish, 2016a); the differing zones of impacting around the circumference of the mill and the existence of top space collisions inside the mill (Pax et al, 2014; Pax et al, 2016b). The latter being an energy loss mechanism heretofore unconsidered and not demonstrated in DEM and ball trajectory calculations. These collisions are due to ‘billiard ball’ type collisions between different particles leaving the lifter system at different times with different starting velocities.

*Figure 1. Photograph of the SAG mill at MMG Century mine with the gantry for twelve microphones in place. The installed gantry is also shown separately for clarity and consisted* *of two components with microphones installed every 30° starting at the top location.*

### Anatomy of the system

The mill acoustic array (Pax, 2011) was conceived and implemented as a solution to allow the determination of in-mill processes with time scales significantly shorter than the residence time of the particles and slurry inside the mill. If the multitude of interactions could be discerned then it would be possible to determine cause and effect relationships, and provide a robust methodology to determine the detail of events such as steel ball impacting and mill charge dynamics optimised for mill throughput and grind size. Also of importance is the determination of the mill conditions at different locations, both around the mill circumference and longitudinally; this provides a means to differentiate the operating zones in the mill but also, when the data is combined from the different microphones, can provide a self-consistent holistic assessment of the operation of the mill. The data and logic used at each location of a microphone in an array has to be consistent with the data obtained at the other locations with the same logic, ie the fundamental physics used to configure and interpret the signals is the same no matter where the measurements are taken. This is the value of using an array of microphones and is an important cross-validation procedure to ensure that the signals have been configured correctly. Further validation is derived from the process measurements, other sensors concurrently installed and testing protocols.

An important first step for an acoustic measurement system is the development of a set of basis signals that describe the ‘noise’ signal. The sounds emitted from a grinding mill are not random and can be quantified by parameters that quantify noise level, impacting, a steel signal and a rock signal. These can then be used to further describe the many-particle tumbling mill in terms of steel ball impacts, rock impacts, toe and shoulder angles for both rocks and steel (Figure 2). A water hold-up signal (Figure 3) is also possible since it influences the basis signals. The multiple signals from each microphone in combination and in isolation can provide a complete description of a tumbling mill.

### A holistic acoustic signal for mill stability and load

These signals can be used to describe the entire operation of the mill, acoustically, without recourse to other measurements. For example, it is normally required that the toe angles and shoulder angles are at particular angular locations, that the impacting of steel balls is an acceptable minimum at the 4 o’clock position, that a ball to rock ratio of the charge is within certain bounds, the water hold-up inside the mill is reasonable – reflecting the slurry transport inside the mill, minimisation of top space collisions and a certain amount of dynamics exhibited in the underneath microphones. When this situation occurs, then the mill will operate with an industry acceptable load inside the mill, the breakage rates will have been optimised and a desirable throughput is achieved at the desired grind size distributions. A combined signal that reflects these criteria is the ‘holistic load signal’ and tested successfully at MMG (Pax and Cornish, 2015b). Since the measurements are fast, a rapid analysis of mill operation is possible, much faster than conventional surveys and their analysis.

Furthermore, the holistic signal allows the detection of overloading and underloading of the mill, and events leading to the potential instability of the mill with timely preventative action possible. All the above acoustic information is available for control and/or real-time diagnostics. Whole circuit analysis is also achieved by considering the mill together with the acoustic array as a sensor and can be used for real-time mine-to-mill, feed characterisation and disturbance rejection, and the determination of mill and circuit change evaluation studies.

### Direct observation of mill stability and achievement – increased throughput

Using a fortuitous discontinuity in the lifter configuration at one location inside the mill, a regular motion (once a revolution) of the charge in the mill was observed directly using the steel shoulder angle (Pax and Cornish, 2015a). The regularity of the signal disappeared when the ore feed rate was increased significantly and the mill operated unstably, indicating that additional dynamics had been introduced into the mill. However, stable operation was again achieved by increasing the feed water addition rate significantly (from 20 per cent to 43 per cent). The water hold-up signal was also used to monitor the mill and the acceptable band of operation was noted. Additional available detail included the onset of effective grinding, since it influences the water hold-up via the slurry viscosity. The importance of the variability in the water addition rate was established and can now be determined using acoustics.

### Mill liner and lifter protection

A signal that uses the unique steel signal (validated, robust and identical for a wide variety of mill types and duties), impacting information coupled with time domain data provides very specific information about steel balls hitting the mill shell, at the exclusion of collisional events occurring elsewhere. This mill shell steel hits (MSSH) signal is used successfully to minimise the liner/lifter wear and damage consistent with desired mill throughput and discharge grind size distribution, by using the information to adjust the mill speed automatically in a control scheme (Figure 3). Although mill speed will directly influence the steel ball trajectories inside the mill, not all particle trajectories are due to the conventional shoulder-to-toe type and are not necessarily influenced as directly by mill speed changes.

### Acoustic surveys

Although a permanent installation of a microphone array is most useful, hand-held equipment running SmartNoise and a roving microphone allows an acoustic survey to be conducted of a grinding mill to determine its ball to rock ratio and the magnitude and frequency of impacting events along the length of the mill. Information about the location of the high impacting zones where a potential of high liner/lifter wear rates or even cracking could occur can then be obtained. The location of where the feed ore is impacting the mill charge inside a mill is easily determined, as is the extent of steel ball impacting at the discharge end of the mill. Some aspects of unusual dynamics inside a mill can also be determined.

### Conclusion

The success of arrayed acoustics using SmartNoise, and at MMG Century in particular, was manifold. Key outcomes included significantly increased throughput by more than 20 per cent, reduced steel ball impacting on the mill shell to reduce mill liner wear, stable mill operation with early detection of causes of events and the development and use of an all-acoustic holistic load indicator used to rapidly determine the status of mill operation.

Non-contact acoustics, implemented using SmartNoise, have been shown to successfully determine in-mill processes in real-time and can be used for the rapid diagnosis of issues that may occur both inside a mill and within a grinding circuit. It has also been successfully used to control a mill for stability and optimisation. Once implemented, payback periods of less than a week are possible.

### Acknowledgements

The support of many mining companies and their staff for the development of a practical non-contact acoustic system is gratefully acknowledged. In particular the support of the recent work at MMG Century by the management and staff of that operation was appreciated.

**References**

Napier-Munn T J, Morrell S, Morrison R D and Kojovic T, 1996. Mineral Comminution Circuits, Julius Kruttschnitt Mineral Research Centre, University of Queensland, Brisbane.

Pax R A, 2011. Non-contact acoustic measurement of dynamic in mill processes for SAG/AG mills, in Proceedings MetPlant 2011, pp 163-175 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Pax R A, Wynn R, Mann S and Cornish B, 2014. Implementation of Acoustic Arrays for Semi-autogenous Grinding Mill Operation, in 12th Mill Operators Conference, pp 273-282 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Pax R A and Cornish B, 2015a. A Novel Measurement for the Internal Operation of a SAG Mill using Acoustic Sensors: A Case Study, in Proceedings Metplant 2015, pp 120-126 (The Australasian Institute of Mining and Metallurgy: Melbourne.

Pax R A and Cornish B, 2015b. Determination of particle trajectories, toe and shoulder dynamics using a non-contact acoustic array on an industrial SAG mill, in Proceedings SAG 2015 Conference (Eds B Klein, K McLeod, R Roufail and F Wang), (Canadian Institute of Mining, Metallurgy and Petroleum).

Pax R A and Cornish B, 2016a. Understanding size effects of semi-autogenous grinding (SAG) mill operation as a pathway to solving feed disturbances – Case study using the MMG Century SAG mill, in Proceedings 13th Mill Operators Conference, pp 321-329 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Pax, R A, 2016b. Using non-contact acoustic measurement to view inside a tumbling mill, in Proceedings Int Mineral Processing Conference (The Canadian Institute of Mining, Metallurgy and Petroleum).