Coal Preparation » Gravity Separation
This report considers the reasons for the existence of low density tails that have recently been measured with alarming regularity in the Australian coal preparation plants employing large diameter dense medium cyclones. Two aspects of the issue are addressed: precision of partition coefficient measurement and the influence of particle breakage. A mathematical procedure has been developed to allow partition coefficient curve error bars to be estimated. A nine step process is provided along with a worked example. Whilst a pragmatic approach is described to allow these error bars to be estimated from washability data sets incorporated in the ACARP data base, a more rigorous approach could be used if complete sample acquisition details were also available.
For the limited number of cases considered in this report, the presence of low and high SG tails fell well within the estimated 95% confidence level error bar limits, but was readily apparent at the 30% confidence level. This is an indication that supportive evidence is required for such a phenomenon.
The errors associated with partition coefficient curve determination are sufficiently high that, in many cases, empirical data cannot be relied on to detect small efficiency changes due to operational or equipment changes. It may be better to determine such effects through fundamentally-based phenomenological modelling.
Significant breakage and liberation were determined for the 6 cases examined in detail from the ACARP data base displaying significant low density tails. There is important evidence that material in the 60 x 4mm and 4 x 1.4mm size fractions underwent breakage and liberation predominately into the 4 x 1.4mm size fraction. The measured reject washability data for the 4 x 1.4mm dense medium reject screen oversize broadly contained higher amounts of low density material than that predicted by applying the Whiten Equation to the raw feed data (using appropriate cut point and Ep values). This suggests that breakage and liberation may be the cause of the low density tails present in the partition coefficient curves. The breakage appears to have occurred after dense medium cyclone entry at a point either after the cyclone has been exited or where there was insufficient residence time to permit proper separation to occur. To have occurred prior to the dense medium cyclone is likely not to have resulted in the low density tails observed.
The consequences of determining where the breakage event occurs are, however, not considered important
for anything other than confirmatory purposes. There are only likely to be two corrective actions possible,
and neither relies on knowing the point of breakage:
- Further liberate prior to the dense medium cyclone. This is currently practiced in some sites where, for example, coarse circuit rejects is crushed prior to the dense medium cyclone circuit. This can have the benefit of reducing losses/increasing yield and can be achieved so as to generate minimum fines.
- Scavenge from dense medium cyclone reject streams. It is inherently difficult to separate small amounts of product from a reject stream. In this situation, other options may be viable such as further crushing.
The actual approach adopted would have to be determined on a site by site basis following metallurgical and economic justifications.
The extent of the liberation and breakage determined means that the commonly used application of size and washability data adjustment via mass balancing procedures prior to partition coefficient curve determination may, in some cases, be inappropriate. A simpler ash-balance procedure is recommended in which partition coefficients are determined using a reconstituted feed washability and size data set. A three step process is provided along with a worked example. This also obviates the need to float sink feed samples, with commensurate reduction in exposure to hazardous chemicals and analysis costs.