Interpretation of mill vibration signal via wireless sensing Sarada Prasad Das, Debi Prasad Das, Santosh Kumar Behera, Barada Kanta Mishra ⇑ Institute of Minerals & Materials Technology, Council of Scientific & Industrial Research, Bhubaneswar 751 013, India article info Article history: Available online 15 September 2010 Keywords: Tumbling mill Vibration analysis Fourier transform Digital filtering abstract Tumbling mills emit vibration that can be captured to assess its performance. We interpret the vibration signature of a one meter diameter mill in response to changes in mill speed, rock/particle size, quantity of balls, and slurry viscosity. A ±500 g tri-axial accelerometer with a wireless transmitter located on the sur- face of the mill shell transmits vibration signal to a receiver connected to a PC. The quality of the vibration signal is preserved due to the use of wireless transfer of data. The filtered vibration signal in the fre- quency domain is averaged which is then used as a parameter for comparison purpose. This approach to signal analysis is most suitable to compare vibration response of a mill to changes in operating param- eters. The mill shell vibration signature turned out to be an excellent indicator to establish the differences in mill performance under wet versus dry grinding conditions, coarse versus fine grinding, changes in mill speed, ball load, etc. The broader implication of the main observations in the context of development of a diagnostic tool to assess the mill performance is highlighted. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction In mineral industry, the size reduction task is most commonly achieved by tumbling mills. The material inside the mill is contin- uously subjected to rotational motion which is transmitted from the mill shell. The motion of the charge is affected by mill speed, mill filling, ball quantity and size, type of ore, slurry viscosity, lifter shape and configuration, etc. The mill performance can be assessed for a set of operating and design parameters. These parameters can be adjusted to arrive at the optimal values. Any shift from the opti- mal values of the parameters affects the charge motion which in turn affects the mill performance. One could predict the mill per- formance up to a certain degree of accuracy based on power and throughput but quite often, recourse to additional information per- taining to mill operation would be required in order to predict more accurate mill performance. It turns out that mechanical vibration of tumbling mills can be picked up by suitably installing accelerometers (vibration sensors) on the mill surface and the sig- nal can be interpreted to link it to the mill operational parameters. The simplest type of mill vibration is a forced vibration, typically due to variation in the center of mass of the charge at any rota- tional frequency. In order to correlate the vibration response of the mill to charge motion and other events that take place inside the mill, it is quite important to understand the dynamics of the charge. This way an alternative method to assess the mill perfor- mance could be established. In recent years, many researchers have attempted to develop mill diagnostic tools to study the dynamics of the charge and other processes that affect grinding efficiency. Liddle and Moys (1988) measured the toe and shoulder positions of the load by the use of conductivity probe mounted in the mill shell. Zeng and Forss- berg (1993, 1994) generated power spectra of the vibration signals and established a relationship between signal characteristics and key grinding parameters like power draw, feed rate, pulp density and product size. Later, Zeng (1994) used accelerometer at nine locations on the trunnion bearing of an industrial ball mill and ana- lyzed the data by considering the power spectra of vibrations using principal component analysis. It was determined by Zeng (1994) that in order to collect maximum amount of pertinent vibration data, the vibration sensor should be placed on the bearing of the pinion axis. Zeng and Forssberg (1995) showed that some of the key grinding parameters in ball mill grinding can be predicted by a combination of principal components derived from vibration and acoustic spectra. In another approach Kolacz (1997) used a pie- zoelectric strain transducer to evaluate weight of the mill charge. Gelle et al. (2000) applied Blind Sources Separation (BSS) to acoustic and vibration data of rotating machines using two acceler- ometers and two microphones. Behera et al. (2007) used an accel- erometer mounted on mill shaft and interpreted the data to correlate with charge motion. Su et al. (2008) installed two accel- erometers on the bearing house of an industrial tubular mill and correlated the energy amplitude of the vibration signal to predict over-load condition. Most recently, Huang et al. (2009) studied the correlation between fill level and angular position of the maximum vibration point on the mill shell by collecting vibration 0892-6875/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.mineng.2010.08.014 ⇑ Corresponding author. Tel.: +91 674 2581126; fax: +91 674 2581160. E-mail address: bkm@immt.res.in (B.K. Mishra). Minerals Engineering 24 (2011) 245–251 Contents lists available at ScienceDirect Minerals Engineering journal homepage: www.elsevier.com/locate/mineng