Scramjet engine flame field intelligent prediction method based on deep learning

The invention discloses a scramjet engine flame field intelligent prediction method based on deep learning, and belongs to the technical field of supersonic flow combustion and artificial intelligence cross fusion. Experimental data of a hydrogen fuel scramjet engine is utilized, and wall surface temperature and a time sequence thereof are used as input variables; a supersonic combustion chamber flame field prediction model is constructed, by introducing a deformable convolution kernel, the capture capacity of the supersonic combustion chamber flame field prediction model for irregular form changes of a flame field is remarkably enhanced, meanwhile, the model calculation amount is reduced, and the gradient disappearance problem is effectively solved by adopting a residual network. The prediction precision and the reasoning speed of the combustion flow field are further improved, real-time cross-time prediction is realized, and the modeling capability of the network for complex spatial transformation is enhanc