Methodological Study on Prediction of Detonation Cell Width with Gaussian Process Regression

Detonation cell width is an important parameter to quantify the detonation risk of a flammable gas mixture.In this study,detonation cell width,λ,is correlated with the characteristic chemical reaction zone width,δ.Regression is performed by using experimental data,with non-dimensional activation energy and non-dimensional temperature as independent variables and the logarithm of λ/δ as a dependent variable.Considering the disadvantages of traditional parametric regression methods,a Gaussian process regression based on machine learning,is applied in this study to fit the experimental data.By the comparison of the measured data and predictions of the regression result,it is found that the Gaussian process regression method can precisely predict the detonation cell width of flammable gas mixtures with different compositions and at different initial conditions.Meanwhile,the Gaussian process regression method takes an advantage in fitting precision comparing to the traditional parametric regression methods.