Development of image processing methods for extracting quantitative parameters from x-ray images of spray flows

Analysis of the fluid dynamic processes that govern spray formation is critical for characterization of fuel atomization in jet engines. The physical and chemical properties of liquid fuels can impact the nature of atomization, which can in turn impact the combustion characteristics and ultimately the engine performance. Experimental imaging of spray behavior in the region close to the nozzle exit can be valuable for understanding the breakup dynamics. X-ray phase contrast imaging is uniquely suited for this purpose because it is possible to image through optically opaque materials. However, quantitative analysis of phase contrast images of liquid flows continues to be challenging for several reasons such as the presence of multiple overlapping fluid structures of varying scales, and the phase effects at the edges of a single fluid features creating sharp intensity gradients. The focus of this work is to develop an image analysis method to identify and track fluid features from the phase contrast images of a complex spray flow field. In this work, we develop a multi-pass algorithm to detect fluid features of varying scales and flow speeds using a non-parametric adaptive kernel density method to compute a background model to isolate the dynamic flow features of interest. We apply this method to an experimental dataset of an impinging liquid jet in crossflow. Our results show that this image analysis method can be used to detect the fluid regions in a range of flow events, and the extracted spray dynamics can then be applied towards higher level tasks such as velocity estimation, and computation of droplet statistics.