Heat transfer coefficient model for three-phase fluidized bed based on neural network

A three layer artificial neural network model with the structure of 3×5×1 has been established by the MATLAB neural network toolbox to simulate the heat transfer coefficient of a novel evaporator for wheat straw black liquor - a three-phase circulating fluidized bed evaporator. The LM algorithm has been selected from four kinds of training functions: BFGS quasi-newton, LM, BR and GD by comparing their training error. All the experimental data have been classified into two groups: one group, accounting for 2/3 of the whole data, is for network training and the other is for network testing. Satisfying matching of the model and the experimental data has been achieved; the maximal deviation is less than 5%. Rapid learning speed and few training steps are also the advantages of the established model. The weight matrix and bias matrix have been given in the paper. The neural network toolbox offered by MATLAB is an advantageous tool to build neural network model.