|
Grain size composition in lamella classification processes |
|
|
|
|
Grain size composition in lamella classification processes Author: J. Marciniak-Kowalska
In lamella classification processes statistical grain size distributions in the final products of the classification processes are determined on the basis of grain size distribution in the feed material. Computed grain size distributions in the fine product in overflow and the coarse-grained product in underflow depend on the employed statistical distribution pattern. Grain size distributions in the feed materials are mostly unimodal distributions covering the interval. Two most widely applied distribution patterns are the log-normal distribution and RRB (Rosin-Rammler-Bennett distribution). The main aim of the present study is to compare the results of approximation of grain size distribution in lamella classification processes. Grain size distribution in the feed is approximated with the log-normal and RRB distribution patterns; on that basis grain size distributions in the products of lamella classification are obtained. The paper provides the comparison between the two distribution patterns employed in approximation: the long-normal and RRB distributions. The log-normal distribution function was computed for the predetermined parameters m and s(sigma) and thus obtained data were approximated with the RRB distribution. Subsequently, the parameters d0 and n were found. The correlation coefficient and F-Fisher statistics were the tools employed to evaluate the goodness of fit. Rosin-Rammler-Bennett distributions were next used to generate distribution functions which, in turn, were approximated with the log-normal distribution. The results of approximation of grain size distributions in the final products of the lamella classification processes are presented and discussed. There was admitted, that when the feed grain size composition displays little scatter (low values of standard deviation s(sigma) in log-normal distribution or high values of shape parameter n in the RRB distribution), the differences in the value of distribution functions might be even as high as 0.2.
|
2011