TY - JOUR
T1 - Prediction of combustion promotion effect of high/low-frequency AC electric fields based on machine learning method
AU - Duan, Hao
AU - Yin, Xiaojun
AU - Kou, Hailiang
AU - Zeng, Ke
N1 - Publisher Copyright:
© 2023 Elsevier Ltd
PY - 2023/8/15
Y1 - 2023/8/15
N2 - Electric field-assisted combustion has become a trend in the past several decades. The effects of low-frequency and high-frequency AC electric fields on the flame propagation and combustion process of CH4-N2-O2 lean burning flame are compared experimentally, and different machine learning methods are applied to predict the flame propagation and combustion characteristics in this research. The experimental results show that both the fields can affect the flame significantly in the electric field direction, while the flame front under a low-frequency field is more stable. The promotion effect of the fields on the average flame propagation speed is in the same order, and the high-frequency fields increase the combustion pressure and pressure rise rate more than the low-frequency one. Support vector machine and artificial neural network are used to establish the correlation model of the average flame propagation and combustion pressure, respectively. It is found that the generation ability and the prediction performance of the models are very impressive and reliable. The correlation coefficient of each model established by SVM method is over 0.998, and the values of MAPE and Theil IC are below 1.093% and 0.007, respectively. The calculation speed of the ANN model is much higher than that of the SVM one. Besides, the results of the SVM method are more deterministic than that of the ANN method. Present work illustrates that machine learning methods provide time-saving research approaches for the forecasting of the flame propagation and combustion characteristic parameters in the electric field-assisted combustion field.
AB - Electric field-assisted combustion has become a trend in the past several decades. The effects of low-frequency and high-frequency AC electric fields on the flame propagation and combustion process of CH4-N2-O2 lean burning flame are compared experimentally, and different machine learning methods are applied to predict the flame propagation and combustion characteristics in this research. The experimental results show that both the fields can affect the flame significantly in the electric field direction, while the flame front under a low-frequency field is more stable. The promotion effect of the fields on the average flame propagation speed is in the same order, and the high-frequency fields increase the combustion pressure and pressure rise rate more than the low-frequency one. Support vector machine and artificial neural network are used to establish the correlation model of the average flame propagation and combustion pressure, respectively. It is found that the generation ability and the prediction performance of the models are very impressive and reliable. The correlation coefficient of each model established by SVM method is over 0.998, and the values of MAPE and Theil IC are below 1.093% and 0.007, respectively. The calculation speed of the ANN model is much higher than that of the SVM one. Besides, the results of the SVM method are more deterministic than that of the ANN method. Present work illustrates that machine learning methods provide time-saving research approaches for the forecasting of the flame propagation and combustion characteristic parameters in the electric field-assisted combustion field.
KW - Artificial neural network
KW - Electric field-assisted combustion
KW - Lean burn
KW - Machine learning
KW - Support vector method
UR - https://www.scopus.com/pages/publications/85152143825
U2 - 10.1016/j.fuel.2023.128348
DO - 10.1016/j.fuel.2023.128348
M3 - 文章
AN - SCOPUS:85152143825
SN - 0016-2361
VL - 346
JO - Fuel
JF - Fuel
M1 - 128348
ER -