Numerical and physical analysis of liquid break-up and atomisation relating to pressure-swirl gasoline direct injection

This thesis presents detailed fuel spray investigations relating to an automotive Gasoline Direct Injection (GDI) pressure-swirl injector, employing a combination of numerical and physical analyses. The emphasis is placed on the near-nozzle in recognition that all later flow processes are dominated by this critical region. To enable the technology to maximise its potential, it is essential to further our understanding of the fundamental flow physics that govern the injection process, which remain largely unknown. The complexity of the spray process has led to many avenues of research. Simplified models are particularly suitable for parametric studies, allowing fast computation of some of the most important design parameters, such as nozzle discharge coefficient, cone angle and initial velocity. More complex methods such as Computational Fluid Dynamics (CFD) offer significantly more detail including the temporal and spatial evaluation of the flow field and fuel distribution, but at the cost of often lengthy computational time, and the need to tune models against physical evidence. Unfortunately none are able to describe all aspects of the injection event simultaneously. A considerable body of existing experimental data gathered under atmospheric conditions has been condensed and carefully presented to provide a comprehensive picture of injector operation. This comprises global spray performance data, spray imaging, and droplet velocity and size maps as a function of time after the Start Of Injection (SOl). These serve to provide a means to develop physical models and to correlate model predictions. Particular attention is drawn to the challenges faced by numerical methods to successfully predict the complex spray behaviour. A fundamental computational study employing the Volume Of Fluid (VOF) method describes droplet break-up under controlled conditions. By varying the Weber number of the flow the expected break-up mechanisms are recovered, and the numerics and case set-up tuned to offer a practical balance between the resource burden and solution accuracy. This paved the way to a detailed 3-D transient analysis of the near-nozzle region of a pressure-swirl injector. Computed results clearly identify the consecutive phases of the fuel spray development, from the initial unsteady jet through to the stable, swirling hollow cone formation. Comparison with experimental measurements revealed that the computational approach is able to capture the main qualitative features of the spray process.