Droplet Velocities in a Liquid Spray

Sammanfattning: A method to look into a dense liquid spray has beenevaluated. The spray was mechanically sliced to remove theobscuring parts of it, thus gaining optical access to theinternal regions and thereby allowing for visualization ofpreviously unavailable parts of the spray. The width of themechanical slit was 100 µm, which was significantlysmaller than earlier experiments performed by otherresearchers. The approach was applied to a diesel fuel spray,and velocities were measured by means of image analysis appliedto double-exposed images of droplets in the spray.To perform the evaluation a test rig wasdeveloped, whereliquid sprays could be optically studied under controlledconditions. In the test rig diesel fuel was injected into apressurized vessel with optical access, which was achievedusing two glass windows, one for the light from the flash, andthe other for the camera. The necessary optics and the camerawere fitted on a rail, to allow for easy adjustment of thefocus. Personal computers controlled the injection sequence andthe image acquisition. The injector used was a prototype withan internal hydraulic pressure booster, which deliversapproximately 100 MPa peak injection pressure. The nozzle usedwas a specially manufactured one with an axial orifice, inorder to fit into the available vessel. The diameter of thenozzle orifice was 110 µm, and the orifice length was 700µm. The temperature in the vessel was 300 K and the gaspressure 3 MPa. The gas in the vessel was air, and consequentlythe gas density was about 35 kg/m3.Enlarging optics in conjunction with a solid-state camerawas used to acquire the images. The magnification of the opticswas about ten times, the useful resolution 10 µm, and thedepth of field approximately 100 µm. The double-exposureswere achieved in two different ways, in the beginning by meansof a spark discharge flash unit able to deliver multiplesparks, with a flash duration of approximately 300 ns, butlater with the shutter of the camera in combination with axenon discharge flash lamp, to achieve shorter exposure times.The spray was lighted up from behind by the flashlight, andconsequently the droplets were visible as dark regions againsta bright background.It was considered convenient to work with an imagevelocimetry software for which the source code was accessible,and for that reason primarily in-house developed codes wereused. Three different velocimetry approaches were evaluated,namely cepstrum, cepstrum with the low frequency pedestalremoved from the intermediate spectrum, and finallyauto-correlation. It was concluded that the most powerfulalgorithm was the cepstrum with pedestal removal. However, thismethod showed significant sensitivity to variations in inputdata, and was therefore rejected. Auto-correlation wasconsidered the most rugged algorithm, and as the images were tobe processed in an automated manner, it was decided thatauto-correlation was to be used.In order to make the final algorithm able to discard imagesfrom which it failed to get correct velocity data, a number offilters and thresholds were used. The filter parameters and thethreshold levels were adjusted manually to treat a smallreference group of images correctly before they were applied tothe real data.Keywords:Liquid spray, diesel spray, dropletvelocities, droplet sizes, spray visualization, sliced spray,particle image velocimetry, spectrum, cepstrum,cross-correlation, auto-correlation.