An analysis of the kinetics for oil agglomeration of coal

Abstract

<p>In batch tests with graphite, 2 min. were required to produce a maximum shift towards larger particle sizes in the distribution curves for a stirring speed of 21,000 rpm, slurry concentration of 2 wt%, and oil dosage of 10 v/w%. For a stirring speed of 650 rpm, 4 min. were required for the maximum shift. When the oil was pre-emulsified and then added to the slurry, the maximum shift required 1 min. at 650 rpm. In all these cases, the distribution curves showed particle size reductions after their maximum shift. In the case of 650 rpm, the shift was all the way back to nearly the feed size distribution;For emulsion characterization, the nonpolar oils (tetralin and a heptane/heptanol mixture) were more stable than the polar oil (heptane) and an ultrasonic treatment produced the most turbid emulsions;In batch studies with Australian coal from the Ulan mine, emulsification decreased the time needed for agglomeration and created slightly more compact spherical agglomerates. Although the use of heptane (a nonpolar oil) created more selective agglomeration of the organic matter, its use caused the agglomeration to proceed at a slower rate. Of the three oils used, heptane, tetralin, and the heptane/heptanol combination, the mixture of heptane/heptanol produced the fastest agglomeration;In continuous runs employing Upper Freeport coal, there were runs where the size of the agglomerates was limited to about 1000 [mu]m or smaller. In these runs, application of the layering mechanism in the population balance did a good job in modelling the particle growth. For runs where agglomeration produced particles that were much larger than 1000 [mu]m, the layering model was able to fit the size distribution for small particle sizes, while a Gaussian distribution was able to fit the distribution for large particle sizes. The Gaussian distribution was useful for two purposes: (1) for determining the influence of the operating parameters on the particle size distribution data, and (2) to serve as an explicit functional form when employing a coalescence/breakage model. Thus, some of the continuous runs produced distributions that were explained by assuming growth by layering, while other continuous runs produced distributions that were explained by both layering (small particle sizes) and coalescence/breakage (large particle sizes).</p>