Reduced-order modeling of supersonic fuel–air mixing in a multi-strut injection scramjet engine using machine learning techniques

Dual-mode ramjet/scramjet engines promise extended flight speed range and are the commonly preferred air-breathing propulsion system from within family of hypersonic aircraft concepts. One main challenges that should be hurdled in their design is modeling fuel–air mixing process to provide optimal fuel distribution yield best engine performance. Injecting into high-speed air stream along multiple parallel struts can augment penetration improve efficiency. Mixing intensity increased with turbulence by shock-expansion waves post-strut regions. However, this enhancement bring about detrimental effects on aerodynamic performance increasing losses total pressure. Designing working configuration requires testing interaction between many variables. This a tedious computationally costly task. Machine learning models thus appear well-suited for multi-objective optimization variables elusive designers. In particular, non-linear regression built over available sparse simulation data predict unseen conditions. present work, we carry out detailed investigation effect multi-strut parameters three objective functions: efficiency, length, pressure recovery (TPR) factor. These functions linked most relevant physical phenomena supersonic flow field. We first generate CFD database solving compressible, non-reactive, Reynolds-averaged Navier–Stokes (RANS) filtered equations 2D scramjet domain varying variables: location, strut wedge angle V-settlement angle. then apply various – artificial neural network (ANN), Gaussian (GPR), kernel formulate surrogate model injection utilized reduced-order studies estimate find generally more difficult vicinity (due turbulence/shock effects) easier further downstream struts, but ANN performs better than other models. Thus, our tool incorporates efficiency predicted ANN. It computes thrust less 10% error. also discussion insights gained database; link earlier findings machine tools. sensitivity study, influencing parameter properties losses.