Colloquium: Fanglun Wang (FPT)

25 augustus 2022 09:00 - Locatie: Online | Zet in mijn agenda

Experimental Study on Laser-induced Near-wall Cavitation via Particle Image Velocimetry

In present study, the single cavitation bubbles of hundreds of micrometers are seeded at near a solid wall by laser techniques, and at the mean time a high-speed photography experiment up to 200,000 Hz, a planar PIV experiment combined with shadow method and the first tomographic PIV experiment in the field are meticulously designed and implemented. The imaged taken are firstly calibrated by a high-order anti-distortion algorithm before signal-to-noise ratio enhancement processing. According to the bubble morphology data, two simple formulas describing the bubble collapse time and the bubble radius as the function of the stand-off distance are fitted. The planar PIV shows the flow fields during the bubble oscillation, which helps to explain the mechanism of the formation of the two types of vortex left by the cavitation bubble.  The consistence between the results of planar PIV and the 2D slices of the tomographic data as well as the axial symmetry proved by the 3D velocity field data support the sufficience of the planar PIV for the most flow fields induced by a single cavitation bubble. The technical difficulties encountered and the resulting data in this first tomographic PIV experiment in the field of bubble dynamics may serve as a valuable reference for future research. The wall shear rate exerted by the cavitation bubble are tried to resolved by the SPCC PIV and the AI PIV for the first time, as they both are able to achieve a single-pixel-resolution velocity field. In the case when the bubble do not touches the wall the SPCC algorithm succeeds, while it is effort-consuming, unstable and uncertain. In particular, the flow fields are dominated by the many sources of instabilities for the moment long after the bubble generation, in which case the SPCC algorithm cannot deal with the images ensemble from different flow conditions. Although the AI deep learning method in present work is failed in obtain the correct velocity gradient at the near-wall region, the there is a very bright way out. The governing law NS equations, the physical constrains non-slip conditon, and the high accuracy PIV cross-correlation velocity data can be together considered by the loss function of convolutional neural network. In such way the AI method is supposed to be capable to get the correct cavitation flow field up to single-pixel-resolution.

Supervisor: Woutijn Baars

/* */