Vous êtes sur la page 1sur 6

764

Experimental investigation and numerical prediction of cavitation incurred on propeller surfaces


Yu-chi Chang 1, Chin-ning Hu 1, Jing-chin Tu 2, Yi-chih Chow 1*
1

Department of Systems Engineering & Naval Architecture National Taiwan Ocean University, Keelung,China 2 United Ship Design & Development Center Tamshui, Taipei, China *Corresponding Author: ycchow@ntou.edu.tw propeller at inclined-shaft conditions and qualitatively described the cavitation patterns without performing any image analysis. Konno et al. [4] even used handmade sketches to illustrate the process of the tip vortex cavitation bursting. However, this kind of data or illustrations is becoming more and more unsatisfactory with the present research need and industrial demand. In other words, it is the recent trend that quantitative descriptions of cavitation are being more regularly sought. For example, Pereira et al. [5] analyzed their images of a cavitating propeller by cross-correlating them with a template image (the image of the propeller at a noncavitating condition) to quantitatively identify the extension of the sheet cavity area on the propeller surface. It can be noted that the advantage of using the cross-correlation technique is its accuracy and robustness in pattern identification. Its operation, however, is considered to be relatively more timeconsuming because it mainly involves operations of multiplication. The present paper attempts to use an alternative approach to achieve similar goals by combining several simpler algorithms such as thresholding, linear filtering (Laplacian, Sobel, average), nonlinear filtering (maximum, median, minimum), etc [6] . Numerical simulations have become increasingly popular and proved to be an efficient tool for designing machineries such as propellers. There are three major numerical methods for computing propeller flows: the vortex-lattice method [7], the panel method [8], and RANS (Reynolds Averaged NavierStokes equation Solver). The first two methods are

ABSTRACT: Multiple types of cavitation appearing on the surface of a marine propeller when the ship advances with high speed cause the decline of the propellers performance and/or damages (erosions) to the propeller. Therefore, better understandings of the cavitation formations and the ability to predict them are important to a successful propeller design. This paper presents image results of cavitation incurred on the suction surfaces of a specific propeller and analyzes these images to obtain the statistical parameters of cavitations such as their sizes, locations, and occupied areas. These experimental results are used to evaluate the accuracy of the associated RANS simulations in order to identify issues of turbulence and cavitation models commonly used in such simulations. KEY WORDS: marine propeller; cavitation; image processing and analysis; RANS.

1 INTRODUCTION Marine propellers often generate cavitation due to their high rotational speed. Several types of cavitation can be induced by a propeller, e.g., tip vortex cavitation, sheet cavitation, bubble cavitation, etc [1-2]. Among these types, the sheet cavitation is mainly attributed to the high angle of attack of a propeller blade and has significant impact on the performance of the propeller. Therefore, it is important and necessary to investigate in detail the cavitation characteristics or responses to a certain propeller design. Majority of previous cavitation experiments associated with marine propellers focused on qualitative descriptions of the phenomena for a quick answer or solution to the engineering problem at hand. For instance, Kehr et al. [3] took random snapshots of digital images of cavitation incurred on the root of a

765

based on the assumption of inviscid fluid, using only empirical formula to account for the effect of viscosity. On the other hand, RANS fully computes the viscous flow and as a result is capable of simulating (at least qualitatively) viscous phenomena such as flow separation and tip vortex [9]. For cavitation computations, for example, Rhee et al. [10] and Liu et al. [11] have obtained feasible results computed using RANS. However, it is well known that the performance and results of RANS simulations are sensitive to the turbulence and the cavitation models chosen. Therefore, numerical results need to be verified with experimental data (at least in benchmark tests). This serves as another purpose of this paper that we address this issue by comparing results obtained from the experiment and the RANS simulation of a specific propeller. 2 EXPERIMENTAL SETUP AND CONDITIONS We perform experiments of a model propeller in a cavitation tunnel with a cavitation imaging system including a CCD camera, a stroboscope, and a data acquisition system. These major experimental elements are briefly described as follows: 2.1 Cavitation tunnel The cavitation tunnel we use is the MCT facility (Median Cavitation Tunnel) at the National Taiwan Ocean University. It is a close-loop, circulation type of water tunnel. The cross-sectional area is 0.5 m 0.5 m. The maximum flow speed in this facility is 11.7 m/s, and the inflow uniformity can be maintained quite well as long as the cavitation number () of the experiment is greater than 0.4. An adjustable inclinedshaft dynamometer is integrated with this tunnel for measuring torque (Q), thrust (T), vertical force and horizontal force experienced or delivered by the model propeller. 2.2 Cavitation imaging system We use the LaVision Imager ProX cross-correlation CCD camera originally designed for PIV (Particle Image Velocimetry) measurements. It has the resolution of 1600 1200 pixels and is capable of acquiring two images with a very short time separation (0.1 s). The light source is a high-power stroboscope with the maximum flashing frequency of 500 Hz. As shown in Fig. 1, the image acquisition synchronizes all the devices with the encoder signal of the motor that drives a model propeller.

Fig. 1 Schematic of the integrated experimental setup

We test a model propeller numbered P4012. It has four blades whose shape is designed with the NewSection profile (Eppler and Shen [12]). This propeller is expected to demonstrate the capability of unloading root and suppressing sheet cavitation [13]. Its performance chart is shown in Fig. 2.

Fig. 2 K-J chart for P4012. The symbols KT, KQ and J denotes thrust, torque and advance coefficients, respectively

2.3 Experimental conditions The inflow velocity is 7 m/s. The cavitation number is set to be 0.6 and the advance coefficient J = 1.14, corresponding to the design point of P4012. One thousand cavitation images are acquired at the same phase angle of the propeller. These images are processed and analyzed using techniques aforementioned and the resulting data are phaseaveraged for comparing with that of the RANS simulation (details follow).

766

3 IMAGE PROCESSING AND ANALYSIS A sample image is shown in Fig. 3a. We first subtract the non-cavitating image (not shown) from it and obtain Fig. 3b. Then the contrast of Fig. 3b is enhanced by thresholding the image into a binary plot of white objects and black background, as shown in Fig. 3c. However, the white area representing the cavity extension is fragmented due to the thresholding process. The remedy for this problem is to dilate all the fragments by maximum-filtering the image to make them inter-connected, and then contract the outer boundary to its original size by further minimum-filtering the image, finally resulting in Fig. 3d. Figure 3e shows Fig. 3d on top of Fig. 3a and it is clearly evident that the cavity extension is quantitatively identified using the image processing and analysis procedure just described.

(c)

(d) (a)

(e) (b) Fig. 3 (a) Sample image of the cavitating propeller, (b) the image after the subtraction of non-cavitating image from (a), (c) the binary image after thresholding (b), (d) the image after maximum-filtering and then minimum-filtering (c), and (e) (d) on top of (a)

767

One hundred images are randomly picked up out of all the images we obtain as samples for the error analysis. We manually inspect a sample cavitation image and the corresponding image mask resulting from the procedure stated above (e.g., Fig. 4a) to check if the latter covers the former correctly. If the cavitation information carried by a pixel in a sample image (i.e., whether the pixel is within a cavity extension or not) is not consistent with the mask (e.g., the pixel is not within a cavity extension but the mask still covers it), the pixel will have one count. The blue parts in Fig. 4b cover all the pixels in error in the sample image. The error (or error probability) can be estimated using
(a)

Err (x, y )

E ( x, y )
n n =1

(1)

where x and y are the coordinates in pixel unit, En is the error indicator for the nth sample image that En = 1 when error occurs and otherwise En = 0, and N = 100 is the number of the sample images. As shown in Fig. 4c, it is evident that the maximum error is less than 5% and the error spots correlate with the locations where cavitation frequently occurs (details follow). 4 RANS SIMULATION We also conduct RANS simulations using FLUENT for predicting extension of cavitation. Grids are generated using GRIDGEN as shown in Fig. 5. We use k- model [14] as the turbulence model. After the pressure distribution on the propeller surface is computed, mixture model (see FLUENT 12.0 manual: Theory Guide) is used to determine the cavity extension.

(b)

(c) Fig. 4 (a) Mask resulting from the image analysis procedure on top of the original sample image, (b) the error spots marked with blue color, and (c) the contour plot of the error (or error probability) Err(x,y) defined using Eq. 1. In (c), the white dashed line with the r/R = 0.3 line indicate the boundaries of the region of interest, outside of which we dont present data for clarity

Fig. 5 Grids used in the RANS simulation

768

6 RESULTS Similar to how we calculate the error associated with the analysis of our experimental data (Eq. 1), we can define the cavitation occurrence probability, COP(x,y), using

COP(x, y )

C (x, y )
n n=1

(2)

where x and y are the coordinates in pixel unit, Cn is the cavitation indicator for the nth sample image that Cn = 1 when the pixel is within a cavity extension and otherwise Cn = 0, and N = 1000 is the number of all the images we acquire. Figure 6a shows the contour plot of COP(x,y). The region where COP(x, y ) > 10% starts from about the last quarter of the cordlengths to a little bit outside of the trailing edge, in-between r/R = 0.4 and 0.7. The contours outside r/R = 0.9 are contributed by the tip vortex cavitation, whereas the red strip (COP > 20%) just above r/R = 0.3 is due to an unexpected occurrence of cavitation probably caused by some impurity of the water suddenly attaching to the surface in the second half of the experiment. The RANS simulation result is shown in Fig. 6b where the lightgray masks are the predicted cavity extensions. The transverse extension of the predicted cavity matches that of the measured contour of COP(x, y ) > 10% , i.e., the region between r/R = 0.4 and 0.7. The longitudinal extension of the predicted cavity, however, starts at about the mid-chord of r/R = 0.55 (i.e., earlier than that of the experimental result) and never makes it to the trailing edge. There may be several contributors to this deviation, e.g., the bubble dynamics may play an important role, which is not considered in our RANS simulation. 7 CONCLUSIONS We have established an experimental setup to optically measure the extension of cavitation incurred on the propeller surface. In order to quantify this characteristic of cavitation, a robust image processing and analysis procedure has been developed to generate a distribution of cavitation occurrence probability. The comparison between the experimental and RANS results show some striking matches, but deviations still exist. The future work will include incorporating bubble dynamics model into RANS simulations and experimenting on other geometries of propeller.

(a)

(b) Fig. 6 (a) Contour plot of COP(x,y) defined using Eq. 2, and (b) the predicted cavity extensions of our RANS simulation result

ACKNOWLEDGEMENTS This project is sponsored by the USDDC (United Ship Design and Development Center) under the grant number USDDC-221-T258(98). The authors are grateful for Prof. Y.-Z. Kehr for the model propeller of his design we used in the experiments. We also wish to thank the graduate students of the PIV Lab and the MCT Lab at NTOU for their assistance in setting up and performing the experiments. REFERENCES
[1] Franc J P, Michel J M. Fundamentals of Cavitation [M]. Dordrecht ; Boston : Kluwer Academic Publishers, 2004: 1-13. [2] Cervone A, Emilio B, Rapposelli E, et al. Thermal Cavitation Experiments on a NACA 0015 Hydrofoil [J]. Journal of Fluids Engineering, 2006, 128: 326-331. [3] Chen S Z, kehr Y Z. The influen of propeller root geometries on cavitation pattern at inclined shaft condition

769

[C]. Master Thesis, Keelung, NTOU, 2008 [4] Konno A, Wakabayashi K, Yamaguchi H, et al. On the Mechanism of the Bursting Phenomena of Propeller Tip Vortex Cavitation[J]. Journal of Marine Science and Technology, 2000(6): 181-192. [5] Pereira F, Salvatore F, Felice F D. Measurement and Modeling of Propeller Cavitation in Uniform Inflow [J]. Journal of Fluids Engineering, 2004, 126: 671-679. [6] Woods R E, Gonzalez R C. Digital Image Processing [M]. PEARSON, 2008: 1-13. [7] Greeley D A, Kerwin J E. Numerical Method for Propeller Design and Analysis in Steady Flow [J]. Transactions of SNAME, 1982, 90: 415-453. [8] Hsin C Y. Development and analysis of panel method for propellers in unsteady flow [C]. Ph. D. Thesis, Massachusetts, MIT, 1990 [9] Stephen B P. Turbulent Flows [M]. Cambridge University ,

2000: 387-457. [10] Rhee S H, Kawamura T, Li H. Propeller Cavitation Study Using an Unstructured Grid Based Navier-Stoker Solver [J]. Journal of Fluids Engineering, 2005, 127: 986-994 [11] Liu D C, Hong F W, Zhao F, et al. The CFD Analysis of Propeller Sheet Cavitation [C]. Nantes, France, Proceeding of 8th ICHD, 2008, [12] Shen T, Eppler R. Wing Sections for Hydrofoils. Part I: Nonsymmetrical Profiles[J]. Journal of Ship Research, 1981, 25: 191-200. [13] Kehr Y Z. On the Development of a Newseries Propeller for High-Speed Crafts [C]. Seattle Washington, U.S.A, Fifth International Conference on Fast Sea Transportation, 1999. [14] Wilcox D. Reassessment of the scale-determining equation [J]. American Institute of Aeronautics and Astronautics Journal, 1988,26: 12991310.

Vous aimerez peut-être aussi