2022 Volume 17 Issue 1 Pages 21-00390
This study deals with different machine learning algorithms modeling of small-scale burner to predict the temperature from the top and bottom of the flame from combustion of ethanol-diesel fuel blends. The data used for training and testing of the proposed algorithms was acquired by combusting ethanol-diesel fuel blends at different fuel mixing proportions, inner diameters of quartz tubes, flow rates of air and volume flow rates of blend in small-scale burner. Three models for machine learning algorithm based on back propagation (BP), generalized regression neural network (GRNN) and support vector machine (SVM) for small-scale burner were established employing the experimental data for training and testing. The feasibility of these algorithms in predicting the flame temperature of ethanol-diesel combustion in small scale were contrasted by performance correlation coefficient (R) and mean absolute percentage error (MAPE). The results showed that all three algorithms could well identify the complicated nonlinear relationship between the flame temperature and related variables. In addition, for the top temperature, the R value of the SVM model was the largest, which was 0.98513, and the MAPE value was the smallest, which was 1.60%. And, for the bottom temperature, the R value of the SVM model was the largest, which was 0.98135, and the MAPE value was the smallest, which was 1.84%, similarly. Therefore, the SVM model could predict the flame temperature of small-scale burner well, and its performance was the best among the three machine learning algorithms.
