Prediction of ash fusion temperatures by using the chemical composition of the ash has previously been conducted only with linear correlations. In this study, a new technique is presented for predicting the fusibility temperatures of ash. Non-linear correlations are developed by using the chemical composition of ash (eight oxides) and coal parameters (ash content, specific gravity, Hardgrove index and mineral matter content). Regression analyses are conducted using information for Turkish lignites. Regression coefficients and variances of non-linear and linear correlations are compared. The results show that the non-linear correlations are superior to linear correlations for estimating ash fusion temperatures. (c) 2005 Elsevier Ltd. All rights reserved.