Coal Preparation » Process Control
Currently there are over approximately 30 gravity concentrators comprising both the Advanced Separation Engineering TBS design and the FLSmidth, Reflux Classifier design installed and operating in Australian CPP's treating raw coal plant fines.
It is well established that both the effective Gravity Concentrator cutpoint and efficiency are related to both the selected operating variables, namely the bed density setpoint and the fluidising water flowrate, as well as the feed coal quality and how well the Gravity Concentrator units are maintained. Incorrect selection of these process settings by poorly trained operators may lead to less than optimal fine coal circuit metallurgical performance.
It is proposed that a range of point density readings at different elevations within the gravity concentrator sorting column be measured (rather than a single point density as is currently the case), and the resulting density profile or gradient be utilised to provide a usable correlation with respect to the unit processing cutpoint and processing efficiency. Further, if a usable correlation exists an expert control system can be developed to optimise the operation of the Gravity Concentrator by automatically adjusting the bed density setpoint and fluidising water rate to match an "optimum" density gradient profile.
The first phase of this project focussed on determining whether a usable correlation exists between the measured density profile within the gravity concentrator sorting column and measured processing efficiency of the unit.
A custom density probe was designed and fabricated through a collaborative effort between Burkert Fluid Controls and Quality Coal Consulting. This custom density probe included 10 probe sensors equispaced at 150mm centres so that the density profile within the sorting column of the Gravity Concentrator could be measured to a depth of approximately 1400mm.
The most recent TBS installation at Drayton CTU was used as a reference site for the site testing program. The coarse 2.4m diameter TBS which processes the nominal -2.2w/w+0.6mm fines was selected on which to undertake the sampling and density profile mapping.
A sampling campaign of 16 runs was devised and undertaken through November 2011 where the bed density setpoint and fluidising water rate were adjusted to manipulate the processing cutpoint and efficiency of the TBS.
During each 20 minute test run the TBS feed, overflow product and underflow reject were sampled and the sorting column bed density profile via the new multipoint density probe was logged.
Washability testing was carried out on the -2.2w/w+0.5mm sample composites and partition curves for each run were developed using a new curve fitting procedure. The new curve fitting procedure uses the standard Whiten with bypass approach but fits 2 separate curves to the raw washability partition data: a curve segment fitted to the low density data points and a different curve segment fitted to the high density data points. Constraints were developed to ensure that the individual curve segments were continuous and smooth at the transition point. This procedure resulted in a significant improvement in matching a "best fit" partition curve to the raw washability data.
Efficiency parameters such as Ep, cutpoint, low and high density bypass and combustibles recovery were derived from the laboratory data and fitted partition curves.
The density gradient data was analysed over each sampling period and a density gradient or density profile curve was fitted to the average point density data. The modified Hoerl equation (which has the general power function form) was found to provide the best fit of the average point density data.
The washability partition curve data and the measured point density data was examined and sampling runs which corresponded to poor TBS operation were discounted and removed from the final data set.
The three curve fitting parameters from the density gradient curves of best fit were correlated against the measured efficiency data derived from the TBS sampling testwork.
The shape and form of the density gradient curves were very similar for the data sets examined and although a few minor relationships (between the TBS efficiency data and the density gradient curve fitting parameters) were observed, the correlations were very poor and would not form a suitable basis to develop a workable control strategy to optimise TBS performance.
Considerable variation in the raw coal feed quality and the fine coal circuit operation (flooding sieve bends) which possibly impacted on the TBS metallurgical performance, may have also masked any correlations between the data sets.
Accordingly it is recommended that the second stage of this work (to develop an expert control system for Gravity Concentrators based on optimising the sorting column density profile) be not pursued.