To achieve sustainable combustion in scramjets, struts or cavities are typically introduced as flame holders, but they also contribute to complex combustion instabilities. This study employs a hybrid Large Eddy Simulation (LES) and Reynolds-Averaged Navier-Stokes (RANS) technique to investigate the German Aerospace Center’s (DLR) strut-based supersonic combustor, providing detailed insights into the combustion processes with a focus on the near-field flame development and overall combustion oscillations within the combustor. The high-resolution simulation reveals that two pairs of recirculation zones emerge behind the strut. The expansion of the main recirculation zone induces a forward movement of the high-temperature region, triggering auto-ignition and causing a flame reverse movement phenomenon. The interaction between turbulence and chemical reactions results in two types of reverse flame movement in the near-field region, collectively forming a complex flame-switching behavior to induce upstream combustion oscillations. The pressure oscillation frequencies in the near- and far-field regions are essentially identical and exhibit strong cross-correlation characteristics, suggesting a potential influence of near-field combustion oscillations on the far-field combustion oscillations.

The subgrid species-temperature correlation term significantly affects the accuracy of large eddy simulation of reactive flows. A new filtered pressure model was developed by using the basic characteristics of probability equivalence in the probability density function method, which can well close the subgrid species-temperature correlation term. The probability density function and its coupling solution method were introduced. Then, a new filtered pressure model was derived based on previous models. Numerical tests of different filtered pressure models were carried out in a three-dimensional supersonic hydrogen/air temporally evolving reactive mixing layer. The results show that, compared with the traditional filtered pressure model, the new filtered pressure model can significantly improve the simulation accuracy of the reaction mixing layer. In particular, based on the new filtered pressure model, the large eddy simulation coupled probability density function method can better simulate minor species in the chain reaction such as hydrogen peroxide, which is expected to more effectively reproduce complex combustion phenomena such as self ignition.