Prediction of attrition rate of coal ash for fluidized bed based on chemical composition with an artificial neural network model

Abstract

Attrition of ash has a significant effect on the performance of a circulating fluidized bed (CFB) combustor, and the attrition rate coefficient Kaf is widely used to analyze the mass balance of the CFB combustor. In this study, a four-layer artificial neural network (ANN) model is developed to estimate the value of Kaf according to the chemical components of the ash based on a training database consisting of 40 sets of samples. An optimum structure comprising two hidden layers with ten neurons in each layer is adopted, and the mean square error for training and validation stages is reduced to 0.00175 and 0.13842, respectively. To verify the validity of the model, field tests are conducted in a large-scale CFB boiler by burning two types of coal with different blending ratios. The Kaf of the two kinds of coal is estimated using the trained ANN model. The ratio of the coal ash discharged via fly ash to the total coal ash decreases, while the Sauter mean diameter of the circulating material increases with an increase in the blending ratio of the coal with a smaller value of estimated Kaf; this exhibits good validity of the model.