Pyrolysis of different rank fuels: characteristics and kinetic parameter study using nonlinear optimization and artificial neural network

During the pyrolysis process, solid fuels decompose at different degradation rates and temperature ranges owing to their complex structure and composition. Therefore, predicting the thermal degradation of a wide range of solid fuels is paramount for understanding the thermal stability of fuels to effectively utilize them. In this study, the proposed artificial neural network (ANN) (NN-5-22-22-2), with three layers and tansig–logsig transfer functions, predicts the thermogravimetry (TG) and the derivative thermogravimetry (DTG) curves based on 24,750 experimental data points as input parameters (six different heating rates, 825 temperature points, and moisture, volatile, and fixed carbon content of five types of samples). In this exploratory study, we applied a nonlinear optimization method, namely the generalized reduced gradient (GRG) algorithm, to determine the activation energy (E) and frequency factor (A) in a nonlinear equation instead of using the conventional slope and intercept method. Five types of rank fuels were calculated using a distributed activation energy model (DAEM) at six heating rates (5, 10, 20, 30, 40, and 50 °C min−1). The results of the highest average correlation factor, low mean square error values, and low normalized mean square error values of the prediction and experimental data were in agreement. Thus, we demonstrated that the GRG algorithm is an appropriate method for analyzing kinetic data.