For many gas turbine architectures a failure modes and effects analysis identifies a potential mode in which failure of an oil transfer pipe could result in oil leakage into the secondary air system. Such an event would result in a complex two-phase interacting flow. The atomisation and transport of the oil within the air system is of interest, but is difficult to predict. Available data for the droplet size resulting from jet breakup in crossflow are limited. A dimensional analysis shows jet breakup in a crossflow to involve many factors. The atomisation process has been shown experimentally to include many physical processes and is still not completely understood. Currently, the most practical method of modelling these breakup processes in sprays is by using a CFD package with a set of sub-models within an Euler-Lagrangian (discrete-droplet) approach. The strengths and weaknesses of each of these sub-models cannot reasonably be tested when used in combination with other approximations to model a spray in crossflow. The purpose of this study was to assess various submodels for liquid breakup with a series of simple test cases.

The flow in the secondary air system of a gas turbine engine passes over numerous oil supply and scavenge pipes and a fracture in such a pipe will cause a jet of oil to be ejected as a spray. This spray will disperse in the surrounding flow. Accurate and reliable numerical modelling of these sprays presents significant problems due in part to their complexity, but also the lack of experimental data available for model validation. This paper describes the design, manufacture, testing and results from an experimental test rig aimed at spray characterisation. The sprays considered were produced through a round sharp edged nozzle with a 0.57 mm diameter and a length to diameter ratio of 1.61. The spray was introduced normal to the cross flow. Phase Doppler Anemometry was used to determine droplet size and velocity for Weber numbers within the range of 13 < Weg < 580 and Momentum Flux Ratio within the range of 0.8 < q < 136, resulting in 19 different spray fields. Each of these spray fields has been characterised at three axial locations. Contours of droplet size, mass flux distribution, axial droplet velocity and transverse droplet velocity are presented. In addition, a pulsed laser sheet and CCD camera were used to analyse the jet behaviour in terms of break up length and jet trajectory.

The leakage and transport of oil within secondary air system cavities is of interest in oil and air system design, for which CFD can be used as a predictive tool. This paper focuses on the leakage of oil from cracks into a high speed crossflow, idealised as round nozzles at Weber numbers and momentum flux ratios relevant to those in an aero-engine. Simulations were performed using the Euler-Lagrangian approach implemented in a commercial CFD code (FLUENT), including sub-models for breakup, deforming droplet drag, collisions/coalescence and turbulent dispersion. CFD predictions were compared with experimental data from two independent studies. The calculated position of the centre-of-mass of the spray plume agreed well with experiment in all cases, but the penetration was found to be under-estimated. Differences in droplet sizes between experiments could not be explained by variations in the the gas Weber number alone, and a review of the literature has highlighted the importance of the liquid to gas viscosity ratio in determining droplet size trends. Experimental trends in droplet size with changing viscosity ratios were captured by CFD simulations, and droplet SMD was predicted within 20% of experiment. It is concluded that the sub-models used within an Euler-Lagrangian approach can be useful tools for the prediction of droplet size, although further improvements in breakup and coalescence modelling will be necessary if greater accuracy is required.