Applied Mathematics Seminar

Tumbling mills are large-scale grinding devices commonly used in mineral processing. The operation of some of these devices, such as Autogenous/Semi-Autogenous Grinding (AG/SAG) mills is difficult to control and optimise. Direct monitoring of the comminution process is not feasible due to the hostile environment inside the mill. However, surface vibration monitoring of the external shell has been shown to be a valuable tool for soft-sensor monitoring of AG/SAG mill "hidden" process and performance variables. Collision events associated with the AG/SAG mill charge (ore slurry and grinding media) motion and resultant comminution processes strongly contribute to acoustic emissions (transient stress waves generated by deformations in a body) that propagate throughout the mill structure. Accelerometers are mounted such that they detect the component of surface displacement normal to the mill shell, predominantly due to the propagation of surface vibration waves. A key problem is the localisation and characterisation of sources that emit the detected vibrations in circumstances of low wave attenuation characteristics and hence, strong propagation of waves around a mill liner/shell. This is important in order to gain a clear understanding of the location and nature of the source acoustic emission events in such mills, which in turn could be used as a means to monitor and optimise comminution in tumbling mills.

This talk focuses on the development of an automated signal analysis system for source event location and characterisation based on multiple sensor surface vibration monitoring of tumbling mills. Surface vibration data is simultaneously acquired from three piezoelectric accelerometers mounted in a triangular array on the liner/lifter bolts of the rotating mill outer shell, coupled to analogue radio transmitters/receivers. The vibrations are analysed in the context of seismological source location methodology and acoustic emission inspection techniques used in the testing of structures. Wave trains associated with individual source impact events are identified and the differences in arrival time at each accelerometer of such events are calculated. The location and characteristics of each large source event are estimated on the basis of an assumed propagation path (surface geodesic) and a fixed propagation speed for "Rayleigh-like" surface waves.

Formulation and solution of the inverse problem for multiple collision event location on a rotating cylindrical surface based on arrival time differences for each event at each transducer is discussed. Vibration event characterisation techniques, signal filtering and feature extraction for the accurate calculation of individual event arrival times, event matching for the identification of the arrival of a specific vibration event at each transducer and optimisation techniques for solution of the source location inverse problem are also briefly discussed. The techniques are demonstrated using multiple sensor surface vibration data collected at an industrial-scale AG/SAG mill, with single impact tests on the mill liner used to establish the characteristics of individual vibration events. An analysis of the stability of the inverse problem for source location is also presented.

An automated signal analysis system for location of impact events in tumbling mills based on multiple- sensor surface vibration monitoring is established. Such a system provides insights into both the efficiency of the grinding process and the propensity for liner wear as a function of mill operating conditions.

(Joint work with J.J. Campbell, V. Sharp, K.J. Davey, P.L. Phillips, D.G. Barker and R.J. Holmes. This work was originally presented at the Intelligent Processing and Manufacturing of Materials conference in Sendai, Japan (May, 2003).)