Statistical methods of images analysis as an essential tool in the assessment process of computer methods intended for numerical simulations of cavitating flows

Abstract

[EN] Cavitation, because of its negative effects, like e.g. erosion, noise or vibration, is usually an undesirable phenomenon. However, in devices where spraying and atomization are expected, cavitation is required. This group of devices includes e.g. diesel injectors. Appearance of vapour bubbles results in increase of the maximum flow velocity. It is possible for the following reasons. Firstly, bubbles start to form in the throat, so its diameter reduces. Secondly, appearance of vapour bubbles along the wall results in a slip boundary condition. Moreover, cavitation has a positive influence on a spray cone angle. However, regardless of the place of occurrence, research on cavitation bases primarily on numerical simulations. The area of numerical methods intended for cavitating flows includes many solutions which differ not only in the basic assumptions, i.e. considering flow either as a multiphase mixture with the average density or just as two independent liquids, fluid and vapour, with a distinct boundary between them, but also in many methods applied in particular approaches. Currently, to choose the best way of the prediction of cavitation phenomenon for the undertaken issue, many aspects should be considered. The most important factor is the assessment level between the results of numerical simulation and experimental data. Secondary are computing time, requirements for the hardware, price of software and additional costs connected with the selected software. The final decision about the chosen way of the cavitation prediction results from all the above-considered elements. The main aim of the work is to present the methods of image analysis, which can be very helpful in this process. The main advantage of these methods is the quantitative answer about the correlation degree between analysed images. It eliminates subjective decisions based solely on a raw imaging material. The image material used in the work was obtained via numerical simulations performed in ANSYS Fluent. Presented methodology bases on their statistical analysis that considers the shape and intensity of cavitating area, as well as on basic methods of image processing and analysis. The conclusion is that the obtained results demonstrate the usefulness of the proposed methods in the aspect of a reliable comparison of images obtained in the numerical studies of the cavitation phenomenon.