Prediction of impinging spray penetration and cone angle under different injection and ambient conditions by means of CFD and ANNs

This paper presents a numerical study on 3D non-reacting isothermal impinging spray penetration and cone angle under different injection pressure and ambient conditions. The selected ambient conditions include different nozzle diameters, injection profiles, backpressure and ambient temperature. Transient Eulerian–Lagrangian multiphase solver in open source software, i.e., OpenFOAM, has been used for modeling fuel discrete phase interacting with compressible gaseous continuous phase. In addition, hybrid breakup model of KH–RT and k–εhave been applied as the standard model in Reynolds averaged Navier–Stokes for liquid fuel core breakup and turbulence modeling, respectively. Numerical results of macroscopic spray characteristics have been validated against the experiment. On the other hand, supervised feed-forward artificial neural networks (ANNs) with back-propagation learning algorithm have been designed to predict the average impinging spray penetration and cone angle under five different injection and ambient conditions. Likewise, 51 different CFD patterns have been used as the input evidence in the selected ANNs. To optimize the learning procedure, Levenberg–Marquardt algorithm has been employed. Based on the iterative algorithm, our optimal ANNs have been selected from 260 different architectures. The selected ANNs are able to predict the impinging penetration length with mean square error (MSE) lower than 0.00097 and correlation coefficient higher than 0.98921. Moreover, the selected ANNs for predicting the spray cone angle have maximum MSE of 0.0004 and minimum correlation coefficient of 0.99439. According to weights and bias of the designed ANNs, two set of equations are proposed for predicting the impinging penetration length and cone angle. Moreover, Based on CFD results, longer impinging penetration length and larger cone angle are achieved using higher injection pressure and ambient temperature and lower backpressure.