Since coal ash is a large portion of the bed inventory in a coal-fired circulating fluid bed (CFB) boiler furnace, the ash formation data for the feeding coal, i.e., the initial ash size distribution and the fragment and attrition rates, are of significance for predicting the material balance in CFB boiler furnace. A laboratory test procedure for determining the coal characteristics was developed early with ash size distribution data published for six kinds of coal. An analysis of this ash size distribution data showed that the ash particle size distribution from narrow size cut samples of coal has a bimodal nature, with a fine soft component and a coarse hard component both following the Rosin-Rammler (R-R) distribution. This work presents a six-parameter model to simulate the ash size distributions. The cumulative size distribution of ash from coal samples with wide size cuts was well predicted by the model. The cumulative size distribution of the soft ash closely followed the R-R distribution, while the hard ash size distribution resembled the coal samples size distribution with both distributions being similar to a R-R distribution. A semi-empirical technique combined with theoretical analysis was developed to predict the ash size distribution. (C) 2003 Elsevier B.V. All rights reserved.