On-Conveyor Belt Analysis of Coal

A laboratory feasibility study has been carried out on new and advanced neutron and gamma-ray analysis systems for the direct on-conveyor belt analysis of ash in coal without the need for sample by-lines. Such an analysis system could deliver the combined advantages of a direct on-conveyor configuration with new and accurate analysis techniques. An industry survey of 18 coal companies carried out in early 1996 has indicated that accurate on-belt ash analysis is of the highest priority. Subsequent laboratory work has focussed on the investigation of methods with the potential for improving the accuracy of ash content measurement relative to existing on-belt ash analysers, the most widely-used of which are based on dual energy gamma-ray transmission (DUET), which is sensitive to variations in ash composition.

The main challenge in developing bulk material analysers for application to direct on-belt analysis is to develop techniques which are capable of measuring the parameters of interest independent of both horizontal and vertical segregation and independent of changes in belt loading. CSIRO Minerals have developed a new technique that has the potential to meet this challenge and a provisional patent application has been lodged. The new technique shows promise for development into an on-belt ash analysis system which is significantly less sensitive to composition changes than DUET and which analyses a much larger proportion of coal on the belt, thus eliminating some key sources of analysis error.

In the present project, the new method has been tested in the laboratory using 24 bulk (about 70-kg) coal samples of thickness from 100 to 300 mm and ash from 7 to 31 wt.%. The samples comprised 12 samples of run-of mine (ROM) coal from Mt Thorley and 12 samples of product and ROM coal from Camberwell Coal. The analyser comprised a fast neutron source (241Am-Be) and backscatter and transmission bismuth germanate (BGO) detectors to measure simultaneously neutron inelastic scatter and thermal neutron capture gamma-rays from bulk coal samples of variable thickness. Data from the CSIRO laboratory analyser were analysed using four different methods, namely spectral windows / multiple linear regression; spectral windows / artificial neural networks; partial least squares; and principal components analysis / artificial neural networks. Of these methods, the first two currently give the best results. However, the measured accuracy of any experimental technique will be limited by the accuracy of the chemical laboratory measurements on which the calibration is based. In the present case, calibration and cross-validation r.m.s. errors of between 0.46 and 0.54 wt.% ash have been achieved using either of the first two methods on a suite of coal samples for which the standard deviation of the chemically measured ash values has been estimated to be 0.5 wt.%. The present results suggest that these spectral analysis techniques are capable of significant improvement in accuracy but are currently limited by the sampling and chemical laboratory error. For comparison, the calculated accuracy of the DUET technique for these 24 samples is 1.8 wt.% ash.

In addition, the same 24 bulk samples were analysed for comparison using a commercial COALSCAN 9500 analyser at MCI. The relative precision of results from the COALSCAN 9500 , the CSIRO gauge and three independent chemical laboratories were compared using the Grubbs Estimator method. The Grubbs analysis indicates errors of about 0.40 wt.% ash for the CSIRO method on variable thickness samples and 0.30-0.50 wt.% ash for static fixed-volume samples in the COALSCAN 9500. However it should be pointed out that to yield accurate error estimates, Grubbs requires at least 30 sets of independent measurements, a condition which is not met in the present work.

In summary, the present laboratory feasibility study has indicated that a new generation of on-belt ash analysers based on the new CSIRO on-belt ash analysis technique will have the following characteristics:

* Much improved accuracy of analysis compared to on-belt DUET gauges and a similar level of accuracy to the commercial PGNAA gauges such as the COALSCAN 9500
* Significantly reduced sampling errors compared to existing on-belt and by-line gauges, as the new analyser interrogates a much larger coal sample;
* A relatively low installation cost compared to by-line neutron/gamma gauges, primarily because of the deletion of the by-line sampling equipment to supply coal to the analyser;
* Reduced installation time through factory pre-calibration; and
* Improved availability as a sample by-line is not required.

The survey results have shown that on belt moisture is the second highest priority. The CSIRO low frequency microwave technique has been assessed for the determination of moisture in the same set of bulk coal samples from Mt. Thorley and Camberwell Coal. It was found that the sample moisture could be determined with an r.m.s. error of 0.35 wt.% over the range 4.6-10.4 wt.% moisture and 100-300 mm height. The measured microwave attenuation of <0.2 dB/cm (at 0.8 GHz) for all coal samples analysed indicated that the technique is well suited to the analysis of these coals at bed depths ranging from 100 mm to well in excess of 1000 mm. The new generation moisture monitors offer the promise of a wider range of operating conditions, improved accuracy and a reduced cost.

It is recommended that further work be carried out to develop these techniques to a stage of plant testing and commercialisation.