Prediction model of flow field in an isolator over various operating conditions

The flow characteristics in a scramjet isolator are quite complicated, so it is difficult to achieve precise operating state detection on a scramjet isolator by wall pressure only. Flow field prediction using deep learning is a promising method to provide a rich source of information for isolator operating state detection. A data-driven model is proposed for the prediction of the flow field in an isolator by fusion convolutional neural networks using measurements of the pressure on the isolator. Numerical simulations of flow in an isolator at various Mach numbers and backpressures are carried out to establish the dataset. A convolutional neural network architecture composed of a fusion of paths is designed. The convolutional neural network is trained using the computational fluid dynamics dataset to establish the mapping relationship between the wall pressure on the isolator and the flow field in the isolator, including the Mach number field, the pressure field, and the temperature field. The trained model is then tested over various Mach numbers and backpressures. The relative error of the prediction results in most areas does not exceed 5%, and the correlation coefficients between prediction results and computational fluid dynamic results are all over 0.99 in the whole testing set. The predictions of this model are found to agree well with the computational fluid dynamics results, i.e., the data-driven model successfully learns the relationship between internal flow field and pressure experienced on the wall of an isolator.