Efficient Prediction of Supersonic Flowfield in an Isolator Based on Pressure Sequence

The prediction of flowfield evolution can provide valuable reference information for the development of hypersonic technology. Flowfield prediction with the introduction of deep learning techniques is a promising method to provide future flowfield evolution in scramjet isolators. A multipath flowfield prediction model has been proposed to achieve flowfield prediction based on wall pressure sequence. The prediction model is mainly constructed with a convolutional neural network. An experimental dataset was built with supersonic experimental data under different evolution laws in an isolator. The flowfield prediction model is trained and validated using independent experimental data. The proposed model's prediction performance under different prediction spans is discussed in depth. The results demonstrate that the predicted flowfield is in good agreement with the ground truth, and the background wave and shock train structure are basically restored, even when the shock train leading edge changes intermittently. The influence of pressure sequence length on the proposed model's prediction performance is also analyzed.

Automated summary

The prediction of flowfield evolution is crucial for the development of hypersonic technology. Flowfield prediction using deep learning techniques is a promising method to accurately predict future flowfield evolution. A multipath flowfield prediction model based on wall pressure sequence has been proposed and trained using experimental data. The model demonstrates good prediction performance in capturing the flowfield evolution even under intermittent shock train leading edge changes.