Steady-state CFD Simulations of a Small-scale Turbojet Engine from Idle to Cruise Conditions

Gas turbine engines have conventionally been designed following a component level approach. Each component is modeled and simulated assuming boundary conditions that may not represent the actual boundary conditions of an operating engine. Therefore, a system-level approach is a logical progression that accounts for component-to-component interactions while removing uncertainties on boundary conditions. To this end, a multi-point compressible flamelet/progress variable (FPV) model is presented for performing steady Reynolds-averaged Navier-Stokes (RANS) simulations of a gas turbine engine from idle (off-design) to cruise (design) conditions. A new penta-dimensional probability density function (PDF) table is a function of mixture fraction, mixture fraction variance, progress variable, and combustor inlet temperature and pressure. A small-scale turbojet engine exhibiting complex geometries with multiple periodicities is utilized to demonstrate the capabilities for modeling and simulating a compressible reacting flow with rotating turbomachinery. Results indicate that by increasing shaft rotational speed (ω_Z^*), the diffuser residual swirl, combustor inlet static pressure (P ̃_(S,3.1)^*) and temperature (T ̃_(S,3.1)^*) increase. This yields larger regions of high temperature in the combustor that raise both T ̃_(S,4)^* and T ̃_(S,4.1)^* at the combustor exit, increasing air mass flow rate (m ̇_air^*) and engine thrust (F^*). Sonic flow is observed in some regions of the turbomachinery component when ω_Z^*>=0.816. But only at ω_Z^*=1.0 (design point) is the engine exit flow sonic in some regions. These qualitative observations are consistent with expectations. Numerical results are compared with measurements. There is good qualitative and quantitative agreement for T ̃_(S,3.1)^*, P ̃_(S,3.1)^*, and F^* vs. ω_Z^* with larger discrepancies for m ̇_air^* vs. ω_Z^*. Increasing ω_Z^* promotes non-uniform flow at component interfaces, which is not typically accounted for in component-level approaches. The sensitivity of fuel flow rate (m ̇_fuel^*) is explored. By reducing m ̇_fuel^* by 5%, less than 1% variation in m ̇_air^* and F^* is observed. This work demonstrates the state-of-the-art in full-engine simulation capabilities, accuracy, and computational efficiency.