| Summary: | 博士 === 國立臺灣海洋大學 === 系統工程暨造船學系 === 105 === Propellers generate multiple types of cavitation when operating. Cavitation affects propulsion efficiency, and even causes erosion and vibration problems. Therefore, avoiding cavitation becomes the main design objective for a propeller. As the CFD (computational fluid dynamics) advances significantly, propeller designers often adopt it for numerically simulating and analyzing cavitation problems of propellers. Nowadays, the RANS (Reynolds-Averaged Navier–Stokes Equation Solver) is the mostly-used computational model in CFD. Due to the fact that the most of propeller experiments of cavitation are qualitative, however, there are no experimental data to quantitatively validate the numerical results.
This dissertation firstly establishes a phase-locked imaging system to take images of unsteady sheet cavitation generated by a propeller during its open water test, and then develops an image processing and analysis procedure to transform the cavitation images to quantitative data with physical meaning: the cavitation occurrence probability (denoted as COP). Secondly, the fact that the vapor volume fraction (denoted as VVF) in the RANS can be approximated by the experimental COP is shown, enabling quantitative validation and calibration for numerical models that can’t be achieved with qualitative comparisons between numerical and experimental results.
The RANS computations for the same propeller are further conducted with the commercial code ANSYS FLUENT. Three distributions of VVF are generated using three corresponding cavitation models: the models of Singhal, ZGB and SS. They are quantitatively compared with the corresponding COP based on two indicators: the correlation coefficient (the similarity level between distributions) and the average value of VVF (the overall value difference to the COP). The results show that the VVF distribution of the Singhal model has the highest correlation coefficient value of 0.69, while that of the ZGB model has the value of 0.60 and that of the SS model has the lowest value of 0.14. Their average values of VVF are all much larger than that of COP (0.011).
Finally, A sensitivity analysis for the adjustable parameters in these three cavitation models is performed to assess their impacts on the results. The results show that except the bubble number density of the SS model, all other parameters yield VVF distributions less correlated with the COP distribution as their values deviate from the default values; except the bubble diameter of the ZGB model, the average value of VVF increases with the values of all other parameters.
|
|---|
