Predicting Pulveriser Performance

This project aims to demonstrate that the Hardgrove Grindability Index (HGI) coupled with standard Petrographic Analysis can be used to greatly improve the prediction of mill power requirements, mill throughput and product size distribution using existing data generated in a pilot scale mill.

The HGI is included in most thermal/PCI coal specifications as the indicator of mill performance. While HGI is adequate to characterise the grinding of most coals, some Australian coals can be disadvantaged as the HGI will indicate higher power requirements, lower throughputs and /or a coarser size distribution than actual.

The evaluation of a coal's behaviour in the thermal or PCI coal market requires knowledge of the size distribution of the organic and inorganic components of the coal to enable the determination of performance parameters such as combustibility, slagging and fouling and handleability.

This project examined the mill test data from ACIRL's pilot scale vertical spindle mill on 96 coals. A total of 360 mill tests, conducted under a wide range of throughputs, roll pressures and classifier settings, was included into the data set.

The mill performance of maceral groups or microlithotypes was assumed to be additive, that is, each maceral group or microlithotype behaved independently and a size fraction of the product PF was the mass weighted sum of the petrographic components of that size. Based on this assumption it was possible to determine the size distribution of the product PF for a wide range of milling conditions based solely on petrographic analysis. As microlithotypes were not determined directly they were estimated from the maceral analysis. The size distribution of individual maceral groups or microlithotypes can also be estimated. Size distribution based on petrographic analysis proved to be a better estimate than that obtained based on the HGI.

Mill power can also be estimated from petrographic analysis, but the HGI is a better predictor of mill power.

Mill feed sizing has a significant impact on PF size distribution and mill power requirements. Feed sizings were not available for the majority of coals examined. It is felt that the departure from a "natural" feed sizing for many of the coals contributed to the scatter of the data.

CoalTech's mill model was shown to fit the mill performance of an individual coal. The lack of feed size distribution and the interdependence of the breakage function, used by the mill model, on mill operating conditions and coal properties meant that this model could not be used in the fitting of petrographic analysis to mill data. A revised breakage function that properly separates mill conditions and coal properties will allow petrographic analysis to incorporated into the model. The mill model could then be used to predict a coal performance in full scale mills with different types of classifiers and should produce better estimates of mill power. This was beyond the scope of this project.