Assessment of mill lifter bar deflection measurements using wavelets and discrete element methods.

This paper shows how Partial Least Square Regression (PLS) methods can be used to model sensor data of spectral character. The modelling approach has been applied on a tumbling mill where a strain gauge sensor measures the deflection of a lifter bar when it hits the charge. The deflection of the lifter bar during every mill revolution gives rise to a characteristic signal profile that is shown to contain information on both the charge position and grinding performance. As a signal pre-processing method the discrete wavelet transform is used. It distinctly shows a capability of signal feature extraction where both time and frequency are of interest. Its well-known ability to achieve good data compression without loss of information is also demonstrated, a data reduction ratio of 20:1 is obtained here. Modelling results demonstrate that different operating conditions are well distinguishable from each other and by that the finding of proper operating regimes are highly feasible. Grinding parameters that are normally measured in the laboratory are now readily modelled from the on-line signal. A further objective of this paper is to link the experimentally obtained strain gauge sensor data with computational data from a discrete element mill model (DEM). This enables to visualise the charge motion and helps to interpret the complex phenomena that take place inside a grinding mill measured by the strain gauge sensor. The approach taken is to simulate the behaviour of a rubber lifter when it is exposed to forces from the grinding charge in a two-dimensional DEM mill model using a particle flow code. The deflection profile obtained from the DEM simulation shows a reasonably good correspondence to pilot mill measurements. [ABSTRACT FROM AUTHOR]