This paper investigates the penetration and diffusion characteristics of liquid kerosene jets undergoing phase transition in a supersonic crossflow. A series of numerical simulations of gas-liquid two-phase flows were conducted in various supersonic flow fields using the Euler–Lagrangian method. Through post-processing of simulation data, the penetration and diffusion behaviors of kerosene in different supersonic inflows were analyzed, revealing the key influencing factors and governing laws. The results indicate that in high-enthalpy supersonic inflows, the primary drivers of kerosene diffusion shift from injection mass to evaporation mass and temperature gradients, as compared to low-enthalpy inflows. The penetration and diffusion capabilities of kerosene are jointly determined by the evaporation momentum ratio and the injection momentum ratio induced by phase transitions. Furthermore, the presence of the heat release zone significantly enhances kerosene's penetration and diffusion, with the kerosene boundary closely aligning with the heat release zone. Enhancing the heat and mass transfer between the supersonic inflow and kerosene droplets to accelerate the phase transition rate emerges as an effective strategy to improve kerosene's penetration and diffusion performance. These insights provide a theoretical foundation for optimizing the performance of supersonic combustors.
Online detection of micro-leakage in complex high-pressure gas pipeline networks is difficult to achieve using conventional methods. A high-pressure pipeline micro-leakage detection method based on background-oriented schlieren measurement and feature matching was proposed in this study to address this issue. Through the BOS measurement, the density distribution of the leakage fields was reconstructed through cross-correlation calculation and Poisson equation solving, which was further compared with numerical simulation results under specific operating conditions. The morphological characteristics of the jet field at different leakage pressures were revealed by comparing the density fields from different experimental conditions. Subsequently, the displacement field data were compressed into one-dimensional feature representations for structure-oriented matching of leakage-field characteristics, with temporal smoothing and a dual-threshold hysteresis strategy incorporated to improve matching robustness. The results show that the error in the peak density remains below 10%, which indicates good consistency between the background-oriented schlieren measurements and the numerical simulations. Meanwhile, the one-dimensional feature curve accelerates computation while retaining the dominant characteristics of the leakage field. The proposed framework achieves an area under the ROC curve of 0.992 and an average precision of 0.998. At the selected threshold of 0.650, the overall evaluation metric reaches 0.972, reflecting a favorable balance between sensitivity and reliability. Furthermore, the temporal stabilization strategy improves alarm continuity and suppresses chattering during detection.