Field Testing of a Zero Offset Dense Medium Separator

Approximately sixty percent of Australian Coal is washed in dense medium cyclones, yet instantaneous module efficiencies always indicate a higher plant yield than that obtained in practice. This difference is extremely significant - experienced coal metallurgists estimate the difference to be the equivalent of up to four percent of yield at constant product ash (Williamson, 1993).

The best available dense medium cyclone simulation model is that of Wood (1991), developed from data collected from the samplilng of individual circuit modules. However, if the actual yield form the in-plant campaign washing of a coal needs to be estimated, most metallurgists still use the original DSM Guanrantee data and calculations - often modified by company of personal correction factors - because the historical data has been shown to accurately predict long term plant performance.

The Wood cyclone model is in fact an accurate and useful prediction of module performance. Dense medium cyclones are efficient separators at the unit level, but operate significantly less efficiently at the plant level, because different modules sampled over an extended period will exhibit an apparent lower separation efficiency, because of variations in cut-point with time. This effect has long been recognised in dense medium circuits, and is the basis of the requirement for tight control of the correct medium density.

The effect of module cut-point differences on overall separation efficiency is demonstrated in Figure 1, showing that even efficient module partition curves result in inefficient overall plant separation if the cutpoints are sufficiently dissimilar.

The loss of yield noted in Figure 1 is due to the fact that conventional dense medium cyclones effect a separation at a density significantly higher than the feed medium density; and the magnitude of th edensity offset (the difference between the cut-point andthe feed medium density) in conventional dense medium cyclones is:

• large and variable, ranging from 0.06 to 0.14 RD units
• implicit in the design of current dense medium cyclones, and
• affected by cyclone and pump wear, errors in density gauge calibration, and feed consist

A significant improvement in overall plant yield will result if the different plant modules can be operated so as to separate at the same cut-point. This could be achieved by:

• measuring the module cut-points on-line, and making appropriate adjustments to the feed medium density in each module;
• ensuring that each module receives the same feed (size, tonnage, and density distribution), the same feed pressure, and operates in the same wear condition, so that the density offset will be the same in each module; or
• developing a separator which will separate at the feed medium density, regardless of changes in module wear or feed consist.

Most of the dense medium cyclone research reported overseas has addressed the issue of individual unit, rather than overall plant, performance. Where research on alternative dense medium cyclonic devices has been reported, little hard data have been presented on coal applications (Ruff, 1983; Ferrara et al, 1993). Little work has been reported that either recognises or attempts to address the problems outlined above.

In Australia, the issue is being addressed in a number of ways.

BHP Steel has developed a system for inferring the cut-point after determining the feed, overflow and underflow medium densitites around a dense medium cyclone module (option 1). This approach is promising (Ellison et al, 1993), but requires a considerable investment in both instumentation and control equipment for each module.

Simulations of dense medium cyclone plant performance using the Wood model (Craney, 1993) have demonstrated that normal levels of pump wear, cyclone wear and density gauge miscalibration account for approximately one percent loss of yield in a multi-module plant. Therefore optimisation of maintenance practices will not, by itself, resolve the problem.

Feed distribution biases therefore have a significant impact on cut-point variability, and a significant number of Australian projects are currently examining ways of balancing circuit loads and consist. However, none of these are considered viable medium term options.

An alternative and potentially simpler approach is to operate a dense medium separator without an offset. The density gauges already in use in all dense medium cyclone plants provide a reliable mearsurement of circulating medium density. If a cyclonic separator could be made to separate efficiently at the feed medium density - that is, with a zero offset - the problem would be overcome in a cost effective manner by the retrofitting of such separators.

It should be noted that conventional DSM type cyclones can be operated as zero offset separators by opening the apex to a sufficiently large size (0.35 to 0.40 Dc ) (Davis, 1987). However, doing so results in a decrease in coarse particle separting efficiency to an extent which makes this option unattractive for operating washeries.

Of the alternative cyclonic separators which have the potential to operate as zero-offset devices, the Dynawhirlpool and Larcodems were disregarded because of their poor separating efficiency on finer coal (Davis 1994). However, the Tri-Flo separator is reported to have superior performance at finer sizes (Ruff andFerrara, 1982) and was therefore the device proposed for evaluation.

Objective

The objective of the project was to test a Tri-Flo separator operating in parallel with a conventional DSM type cyclone to see if:

• the Tri-Flo is as efficient as a DSM type cyclone,
• the Tri-Flo can operate as a constant cut-point zero-offset separator and, if so, at what efficiency.

The Tri-Flo is effectively two Dynawhirlpool separators operating in series, with the floats from the first separator feeding directly into the second. The cylindrical body of the Tri-Flo is divided by a partition wall into two consecutive chambers communicating with each other through an axial orifice, and each chamber is equipped with an involute media inlet and sink discharge.

The feed is sluiced with a small amount of medium and added to the first chamber of the vessel, where a separator takes place giving a 'sink 1' product. The float from the first stage is the feed to the second chamber where a second separation, generally at a somewhat lower density, produces a 'sink 2' product and the final product (Ruff 1983).

Conclusions

• The Tri-Flo can operate as a zero-offset separator but is unstable at those conditions.
• The Tri-Flo is stable when operated at a small negative offset.
• When so operated, the Tri-Flo has similar separation efficiency to a higher DSM type dense medium cyclone althouth it operates at a higher medium to coal ration.
• The Tri-Flo's offset is as stable as that of a conventional dense medium cyclone with normal variations in plant conditions with the exception of pressure to which it appears more sensitive.
• There is not compelling technical advantage of a Tri-Flo over a conventional dense medium cyclone.