CEE April-June 2012

be analysed as influencing variables. A test plan comprising eight te~ts and three centre– points were used. As this plan was intended as a first step, replications were not performed and a mixing of effects accepted. Test planning on the basis of the pressure drops led to no "impossible" test combinations. All eleven tests could be performed and fulfilled the requirements for the stability conditions. A subsequent abort test demonstrated that, with increased feed, as described above, it was possible to meet the stability conditions either only briefly or not at all. Effect evaluation and selection were performed on the basis of the analysis of variance (ANOVA) and graphical procedures such as the half-normal plot. In the ANOVA analysis, it is necessary to adhere to an error probability of a maximum (p-value) of 5 per cent in respect of the alternative hypothesis (effect present - but not identified) in order to attain the validities of an effect. The coefficient of determination R 2 should be above 0.7. Evaluation based on the example of particle size at 50 per cent screen undersize (kso%) by means of the half-normal plot method is shown in Figure 7. All factors located on the straight lines are not significant; all those not located on the straight lines are potentially significant influencing factors which are further validated by means of ANOVA with appropriate software. A total of more than twenty responses, such as product fineness, surface area and system energy consumption were investigated and evaluated. Two of these evaluations are plotted logarithmically on the basis of a 30 plot of the influencing factors in the form of a prediction in Figure 8 and Figure 9. By definition the screening plan only applies to a linear system, which , however, in this case was logarithmically adjusted. The quality of the prediction stands and falls with the actual suitability of the assumption of a logarithmic adjustment. Table 3 qualitatively summarizes the essential process parameters and their influence on product dispersity and specific energy consumption . The results obtained up to now and shown in Table 3 largely support the correlations already known from the relevant literature. To obtain further knowledge of the mill and of internal product circuits, in particular, subsections of the mill are now studied analytical ly. e -= ., . . 0 "' .X A Au flow lm'ih) 8 Effect evaluilt1on for the 1nfluenong factors of class1f1er speed and classtfymg a1r flow rate on the k.., A. Aor flow [m /h) D· Spez. grondmg roller pre~sure (kN/m' ) 9 Effect evaluation for the influenong factors of roller gnndmg pressure and dassifytng a1r flow on total energy consumptton Table 3: Selected parameters and their influence Kso% (K7s% K2s%) [mm] Kso% [1] [A) ~ ~ Classifying air flow rate [B) ~~ ~~ Classifier speed [D) 0 ~ Grinding roller pressure [AB] ~ 0 Air flow* Classifier speed *red = negative/green = positive, •• =significant effect/ 0 = none significant effect Courtesy: Spec. Energy [kWh/t] ~ ~ ~ 0 ATMineral Processing English Edition 1-2012, Vol.53, Pp 48- 54. 29 l