The measurement of fluid velocity in the vicinity of wavy interfaces by means of Particle Image Velocimetry (PIV) still constitutes a challenge. Besides the experimental complexities such as appropriate seeding, reflections due to gradients in refractive indices, aberrations, etc., also the image processing phase constitutes a critical component. Ignoring bias errors introduced by laser reflections near the interface, strong velocity gradients are typically encountered near the curved liquid/gas interface and are detrimental to the common cross-correlation analysis. These effects are exacerbated by the use of traditional rectangular static cross-correlation windows. Moreover, dynamic boundaries suffer from a certain loss in reliability due to the difference in number of particles located in a specific physical region between the two frames. In this paper we present an improved algorithm enhancing the accuracy of the velocity vector detection in dynamic wavy flows. This algorithm divides the fluid domain in subregions (i.e. bulk, intermediate, vicinity and interface) and applies to each region different features (such as image patching, window relocation and forward difference finite schemes) and ad hoc predictors. Its assessment is performed on the basis of synthetic images, which reproduce a wavy motion similar to the one encountered when fluid in a cylindrical reservoir undergoes sloshing. When juxtaposed with a standard PIV code, numerical errors are shown to strongly diminish while improving the spatial resolution approaching the interface. The algorithm is finally applied to experimental PIV images of water sloshing in a cylindrical reservoir to provide more reliable results at the interface, contrary to standard PIV codes.
|Number of pages||16|
|Journal||Experimental Thermal and Fluid Science|
|Early online date||13 Dec 2018|
|Publication status||Published - 1 Apr 2019|
- Adaptive PIV
- Dynamic interfaces
- Free surface flows