An improved Kelvin-Helmholtz Rayleigh-Taylor (KH-RT) breakup model with wide fuel applicability based on data-driven techniques

The Kelvin-Helmholtz Rayleigh-Taylor (KH-RT) model, widely used for spray breakup simulations, relies on some idealized assumptions (e.g., neglect fuel viscosity and surface tension), necessitating case-specific manual tuning that introduces substantial prediction uncertainties in spray simulations. This is critical because spray simulation accuracy directly impacts the in-cylinder mixture formation, combustion, and ultimately the performance predictions of power equipment. To address this issue, this study establishes a data-driven correlation between the breakup length constant Cb and two key parameters (i.e., ambient density ρg and fuel kinematic viscosity νl) through Non-dominated Sorting Genetic Algorithm II (NSGA-II), which can be expressed as Cb = 0.075⋅νl3⋅ln(ρg)+4.74. The proposed correlation is first verified by the analytical solution of Cb, demonstrating good agreement in both magnitude and trend. Furthermore, large-scale validation using 563 cases was conducted, confirming that spray features of the eight fuels (i.e., diesel, gasoline, biodiesel, DME, methanol, PODE, ammonia, and n-butanol) under various conditions can be accurately captured by the improved KH-RT model without manual parameter tuning. The impacts of fuel properties and ambient conditions on breakup length, along with their influences on spray development, were finally investigated. The improved KH-RT model eliminates empirical parameter tuning, significantly reducing both computational time and spray prediction uncertainties, thereby improving power equipment performance predictions.