Flow structure captured in a scramjet combustor: Turbulent combustion flow inversion driven by sparse observations

Using sparse sensor data to reconstruct the global combustion flow state under supersonic combustion conditions is a novel problem, which is of great significance for breaking through the measurement and diagnosis technology of high-speed aircraft in extremely complex environments. However, the problem is particularly challenging when the number of sensors is extremely scarce. At present, the common end-to-end learning model usually has extremely strong generalization problems when reconstructing the combustion flow in the whole space, especially for the sparse pressure measurement system common in the actual ground wind tunnel test. To this end, this study adopts an encoder–decoder structure to construct an inversion method for turbulent combustion flow driven by sparse observation data. The channel aggregation algorithm is used to aggregate the sparse observation information obtained by sparse pressure sensors in different spaces, and the scale factor is used to adjust the numerical scale of the aggregated channel to prevent local channel information from being covered. In order to fully express the obtained feature information, the feature fusion attention network is used to enhance the detailed features of complex flow areas. The effectiveness of the proposed model is demonstrated on a ground test dataset acquired at multiple jetting backpressures. The results show that the proposed model can still effectively predict the global flow field, given 14 sparse observation points.