Optimization and backpressure mechanisms of a ramjet flowpath in an over/under turbine-based combined cycle inlet

Abstract

Two-dimensional variable-geometry inlets are key to wide-envelope operation of over/under turbine-based combined-cycle engines, but their ramjet flowpaths must balance aerodynamic compression, installation constraints, and flow regulation during mode transition. For a single fixed ramjet-flowpath geometry operating over Ma∞=3.5–7.0, an aerodynamic-prior-guided gated deep neural network (AP-GDN) is coupled with differential evolution to perform weighted multi-condition optimization, followed by a low-Mach-number backpressure analysis. AP-GDN combines freestream-condition gating with a total-pressure-recovery-to-throat-Mach-number cascade to capture cross-condition nonlinearities; R2 is at least 0.985 for all five test-set outputs. With the adjustment-preference coefficient α=0.10, the selected configuration increases the multi-condition aerodynamic-performance component and the composite objective by 5.0% and 11.7%, respectively, relative to the baseline. At Ma∞=7.0, it achieves φ=0.939, raises PRth by 10%, and increases δ2 from 5.39° to 7.60°. Quasi-steady backpressure loading shows that upstream motion is controlled jointly by an asymmetric X-type shock train and separation-induced blockage along the lower wall; the critical backpressure ratios are ∼18 and 24 at Ma∞=3.5 and 4.0, respectively. The framework quantitatively reconciles wide-range aerodynamic performance, second-ramp adjustment requirements, and low-Mach-number backpressure tolerance under a prescribed mission preference.