Image processing for Bubble Image Velocimetry in self-aerated flows

Daniel B. Bung, Daniel Valero

Thursday 2 july 2015

15:20 - 15:35h at Africa (level 0)

Themes: (T) Special session, (ST) Acoustic monitoring of flow, turbulence and river discharge

Parallel session: 12F. Special session: Acoustic monitoring of flow, turbulence and river discharge

In hydraulic engineering, self-aeration often occurs in high-turbulent free-surface flows. From the practical point of view, this is a key phenomenon as it directly affects important design parameters like flow depth and energy dissipation. Design formulas are oftentimes developed on basis of experimental data from hydraulic laboratories. However, gathering of detailed air-water flow data is difficult as very sensitive and cost-intensive measuring equipment is needed. In most cases, intrusive needle probes are applied which detect phase changes by a change of conductivity or light refraction. It was recently shown by Bung (2011) and Bung (2013) that non-intrusive measuring with a high-speed camera can produce valuable results which help to support or even to better understand data from these intrusive air-water measurements. For example, the Bubble Image Velocimetry (BIV) method, introduced by Ryu et al. (2005), gives instantaneous velocity fields which are not possible by classical intrusive point measurements. This technique is similar to Particle Image Velocimetry (PIV), using different gray values of aerated flow for cross-correlation instead of some artificial seeding. However, as demonstrated by Leandro et al. (2014), resulting flow velocities tend to underestimate intrusive probe data from the flumeā€™s center-line. According to Kramer (2004), who compared flow velocities in aerated chute flows, wall friction lead to a velocity decrease of ~10% near the sidewalls. Leandro et al. (2014) thus assume that the low flow velocities are caused by sidewall effects as well as the well-known characteristic of intrusive probes to overestimate velocities. However, the potential improvement by image preprocessing was not investigated. In this paper, different filtering and image processing techniques are evaluated with the aim to fill this gap and to test if a preceding image processing may enhance the BIV results for a skimming flow in a stepped spillway model. For this purpose, constant settings for BIV calculations are applied (i.e. interrogation window size, overlap, correlation method) although it was shown by Leandro et al. (2014) that these settings are of high significance for BIV accuracy. It is thus not intended in the present study to optimize the absolute BIV results but to evaluate the relative effect of the different image processing techniques.