St.

Anthony Falls Hydraulic Laboratory Technical Paper 283- Series A

Hydrodynamic Considerations in the Design of the Hydrodynamics and Cavitation Tunnel (HYKAT) of HSVA

Roger E.A. Arndt, St. Anthony Falls Hydraulic Laboratory Ernst-August Weitendorf, Hamburg Ship Model Tank liSVA

1. Introduction

After the installation of several new test facilities abroad, especially in the Far East, the Hamburg Ship Model Tank (lISVA) felt the need for a new water tunnel with a wide spectrum of capabilities in order to maintain its competitiveness in cavitation research, hydroacoustics, and flow investigations of ships 'and underwater bodies. Considerations for the new HSVA facility, designated Hydrodynamics and Cavitation Tunnel (HYKAT), are contained in a review and comparison of the capabilities of the HYKAT with 14 other existing or planned facilities (Weitendorj et al, 1987). The test section was chosen to be 2.8 X 1.6 m, and the maximum test section velocity was selected to be about 12 m/s. A high tunnel velocity is known to be important: Bubble dynamical considerations and comparisons between model and full scale results Indicate that model propellers operating at high rotational speed in a closed test section tunnel can yield better model full-scale cavitation correlations than can be obtained in free surface facilities with slowly turning model propellers that satisfy the Froude modelling criterion. (This statement may be less true for high speed craft.) These considerations and the need for a balance between size and cost led to the following requirements for the HYKAT:

- Large test section resulting in small blockage effects with whole or shortened ship models,

- Excellent test capabilities for well developed cavitation phenomena at high test section speeds (V :::: 12 m/s), i.e. good cavitation correlation,

- Low turbulence level in the test section, and

- Very good hydroacoustic quality.

To ensure these and further requirements (Weitendorj and Friescli 1990), detailed investigations were carried out at the Saint Anthony Falls Hydraulic Laboratory.

2. General Considerations

After having established the overall configuration of the HYKAT goals were set for the detailed hydrodynamic and hydroacoustic design. Most important are the requirements for velocity uniformity of ±1 % and a turbulence level of less than 0.2% in the test section. More subtle hydrodynamic criteria are established for hydroacoustic testing. A quiet, recirculating flow places severe design requirements on various components in the flow circuit. A slight amount of cavitation acceptable in earlier water tunnel designs is no longer tolerable. It has been determined from noise measurements in many different water tunnels that cavitation of the main drive pump is a major source of noise. Therefore, the design of the pump becomes a critical item. The pump inflow velocity distribution must be made as uniform as possible.

One of the problems encountered in meeting the design goals is the need for cavitation testing over a, wide range of velocity. The particular needs of the HYKAT dictated that the highest elevation of the configuration be at the roof of the test section. In the absence of any other hydrodynamic factors, the minimum cavitation number that can be achieved is 2gh/V2, where h is the maximum elevation of any water passages above the test section and

Schiffstechnik Bd, 37 - 1990

95

V is the test section velocity. Thus, a nonsymmetrical contraction and diffuser are required to meet this criterion. These components are critical for achieving the desired test section flow characteristics as well as for the pump inflow. The turning vanes in the elbows and the turbulence management system are other important components.

In addition to achieving uniform pump inflow characteristics, accurate head loss data are desirable for the pump design. In summary, the special needs of the HYKAT design dictate careful analysis of several components with special attention being paid to any interaction effects between components. The design objectives are given in Table I.

Table I: Design Objectives

Component Requirements Considerations
Contraction 1.6m x 2.8m test section Cavitation
Contraction ratio = 4 Boundary layer separation
Non-symmetrical contraction
±1 % deviation
from velocity uniformity
Turbulence 0.2% turbulence level Head loss
Management Minimum length
Diffuser Non-symmetrical Sensitivity to inlet velocity profiles
Area ratio = 2.16 Exit velocity profiles
Pressure recovery
Separation
Turning Vanes 90° Turning angle Optimize profiles
Pressure distribution
Boundary layer separation
Cavitation susceptibility 3. Approach

3.1. Interactive Physical-Numerical Modelling

Initial studies of the proposed tunnel {Weitendorf et al. 1981} recommended shape of the contraction, the turning vanes in the first elbow and the turbulence management system, and gave overall headloss estimations (Song et al. 1981). The second elbow was changed from round to square cross-section, and the number of turning vanes in the elbows has been increased (Song et al. 1988).

Recognizing the stringent time requirements for completion of the design and the degree of uncertainty in the interactions between components, an interactive physical-numerical modelling approach was selected. A physical model can show interaction effects and is useful in verifying mathematical models. However, it is incapable of simulating flow at full scale Reynolds numbers. This can be overcome by using it to "tune" a numerical model of the diffuser flow. The numerical model can be used to calculate changes in the diffuser outflow with an increase in Reynolds number. The diffuser was selected for special study since its outflow is the primary factor influencing the inflow to the pump. The mathematical models of other components were also useful in evaluating necessary changes that could then be verified in the physical model.·

3.2. Physical Model

A physical model (Fig.l) was constructed to a scale of 1:5 using air as the working fluid to reduce the structural loading as well as simplifying turbulence measurements. The Reynolds number ratio between model and prototype was about 1:18 at the design flow velocity. Not the entire flow circuit was modelled: the components between the pump and the elbow No 4 were not considered (The numbering of the elbows starts with No 1 behind the test section and follows the flow direction). The pump was substituted by an axial flow fan. The fan motor speed was constant, and the fan blades were adjustable at rest. The inlet consisted of a bellmouth.

96

Schiffstechnik Bd. 37 - 1990

t

NOmenc~tlt\,lre

V mean vel.poHy P mean pxeeaur e

'r tt,l)':'bulence !ntenaity

l3J? me~m boundary pr es sur s

V.T. ~)pow ).

Elbow 2 Fa;',ing conr Lcuret Lone

Fig. 1: Sketch of 1 to 5 Physical Model

2.5 ~---T----75~s---'---~

d~'/ IX'

./ be

. ~s

d' ft .>

...... 6"'"

10~~

0.8

1.5 I--

CALCULATED

2.0 r-

0.6

RQ

0.4

, o,

.._.

0.5 f-

o

1.0

0.2

0.4 x/L

0.6 0.8

Fig. 2: Boundary Pressure Distribution on Centerline of contraction

-

-

-

0····· .. AI = 0.6667 6- - AI = 1.0 0- AI = 1.5

-

AI = or/os

8 = (Os + 0r)/H( 1 +n)

S = SEPARATED FLOW

O~----~I ~I ~I __ ~

o 0.05 0.10 0.15

0.20

B

Fig. 3: Calculated Blockage Effect on Diffuser Exit Discharge Ratio

Schiffstechnik Bd, 37 - 1990

97

A filter Was placed at the inlet to prevent the ingestion of dust particles that otherwise could rapidly destroy the hot wire probes used for velocity measurements. The contraction, test section, and diffuser were constructed of transparent plastic to permit flow visualization. A large number of pressure taps were drilled in these components, usually along axial centerlines. Several viewing ports were also installed in the second elbow to permit inspection of the flow in the region of the shaft fairing.

As shown in Fig.l, measurements of mean velocity, turbulence intensity and pressure were made at various positions in the flow circuit. A hot wire anemometer was used for all velocity measurements. Pressure was measured with either a micromanometer or with a pressure transducer coupled with a 30 position scanning valve, depending on the number of measurements involved. An A/D converter and a minicomputer were used to process and store the readings from the transducer and the hot wire anemometer.

3.3. Mathematical Model

An Euler equation model was used for the design and analysis of the contraction and the turning vanes (Song 1987). This model has the advantage over potential flow models of being able to take into account flow distortion, which is an important design consideration for both the nozzle and the turning vanes. A more sophisticated, two dimensional Navier-Stokes equation model was necessary for studying the diffuser flow. As will be described in the latter portion of this paper, the flow in the diffuser is extremely sensitive to inflow velocity distribution, Reynolds number, and free stream turbulence level, and thus must be studied carefully. Because the diffuser has a very sm:all lateral diffusion (1.50 included angle), a two dimensional flow model was judged to be adequate (Song 1988). The scheme is similar to other large eddy simulations reported in the literature using the closure method of (Smagorinsky 1963). It uses the instationary finite volume method with the predictor-corrector scheme. The velocity distribution at the upstream end of the diffuser and the mean pressure at the downstream end were specified as boundary conditions.

For large Reynolds number flows, like the diffuser flow, extremely small grids are needed to completely resolve the flow in the wall region. To make the computation economically feasible, a coarser grid system was used with a partial slip condition based on an assumed logarithmic velocity distribution near the wall. Measurements in the physical model were used to calibrate the mathematical model.

4. Component Design 4.1. Contraction

Based on previous experience with water tunnel designs, a fifth-order polynomial contraction shape was selected for the bottom and the side walls. This profile is more flexible and smoother than the commonly used matched third-order polynomial profile. The contraction matches the entrance section and the test section in terms of position, slope, and curvature.

Until quite recently the analysis of such flows was largely based on the assumption of potential flow. With an Euler equation model the effect of vorticity can be included in the analysis. Because the inflow velocities are likely to be nonuniform, an Euler equation model was chosen for this work. A number of geometrical parameters were varied in search of the most desirable configuration. Three configurations were examined: (1) a nonsymmetrical design with flat top, (2) an intermediate design with 1/3 contraction from top and 2/3 contraction from the bottom, and (3) a symmetrical contraction. The basis of comparison was cavitation characteristics, velocity uniformity, and boundary layer characteristics. The Stratford criterion was used to evaluate the potential for boundary layer separation.

All three configurations were almost equally good in terms of velocity uniformity and bound-

98

Schiffstechnik Bd. 37 - 1990

a~y layer characteristics. From the computed pressure distribution along the boundary, the location of maximum cavitation susceptibility was determined. Cavitation should occur first at the top of the test section. Hydrostatic pressure prevented cavitation from occurring at the lower surface of the contraction. With a flat top surface, cavitation would start in the test section for any test section velocity. For the other two cases, cavitation may start near the contraction entrance depending on the velocity. Therefore, the nonsymmetrical contraction with a flat top was chosen. The results of the mathematical models were compared with those of the physical model and showed good agreement, as can be seen for the bottom wall of the nonsymmetrical contraction (Fig.2). Further details can be found in Weitendorf et at. (1987).

4.2. Turbulence Management System

While considerable information on turbulence management exists for wind tunnels, infermation on water tunnels is sparse. Screens are the most commonly used device for turbulence control in wind tunnels, but they are not suitable for low noise water tunnels because of their tendency to '(sing" in water flow. Several retrofits of turbulence management systems for exist. ing water tunnels have been carried out. In these cases, contrary to our situation, the incoming turbulence was known.

The HYKAT utilizes honeycombs based 011 the theory of Lumley and McMahon (19(7) and Batchelor (1953). This theory was experimentally verified by Robbins (1978) during the retrofit of a turbulence management system in the 48 inch water tunnel at the Pennsylvania State University.

The turbulence intensity in the test section is primarily influenced by three factors; a) the scale and magnitude of the incoming turbulence, b) the attenuation of turbulence by the honeycombs and the generation of new, finer scale turbulence in the flow leaving the honeycombs, and c) the stilling length and the contraction ratio of the nozzle. After considerable analysis, a coarse and a fine honeycomb in tandem were selected having cell diameters of 2 em and 1 cm and lengths of 40 em and 20 em, respectively. It was estimated that this system could reduce the turbulence level to about 0.2% with a 4;1 contraction ratio nozzle.

Verification of the turbulence management system design in the physical model is not possible because of the mismatch in Reynolds number, the difficulty in obtaining scaled down honeycomb, and the fact that not the entire flow circuit was modelled. Therefore it was de" cided to verify the design procedure instead. Two sections of honeycomb, properly spaced, and having a length to cell diameter ratio of 24, were used in the model. As the turbulence level of the inflow was quite low, additional turbulence was created by placing a coarse grid in the entrance section of the model. Turbulence levels were measured upstream of the first honeycomb, between honeycombs, downstream of the second honeycomb, and at the entrance of the test section. These data were used to check the predicted reduction in turbulence intensity at each position. The results indicated that the honeycombs were less effective than predicted, but the stilling length behind the fine grid in combination with the contraction was more effective than predicted, resulting in an overall performance of the management system that is better than expected.

An incoming turbulence level of 13% resulted in a 0.15% turbulence level in the test section.

Unfortunately the actual turbulence level downstream of elbow No 4 is not known, since not the entire flow circuit was modelled. However, Robbins [7] measured a level of 10% at this position in the large water tunnel at the Pennsylvania State University. So the turbulence level in the HYKAT should be close to the design target.

4.3. Diffuser

The diffuser flow is very sensitive to inlet flow conditions. Considerable effort went into the development of a two-dimensional numerical model to analyze this flow. This was deemed to be important because of the lack of symmetry in the primary diffuser. The main diffuser was

Schiffstechnik Ed. 37 ~ 1990

99

designed to have a horizontal top wall and a bottom wall inclined downward by 5.4°. To study the effect of diffuser asymmetry on the flow quality, a group of diffusers with different angle 8t between the top wall and a horizontal line, and with constant total diffuser angle 8 = 5.4° was considered.

One of several factors analyzed was the blockage produced by the boundary layer flow at the inlet. Let liT and liB represent the boundary layer thicknesses at the top and bottom wall of the test section exit. A blockage factor can then be defined as

B = liT + liB

(1 + n)Hi

where Hi is the diffuser height at the inlet, and n is the power law exponent of the boundary layer velocity profile. The measured value of n = 5.125 is used in the calculations. In addition, the asymmetry parameter

has been found to be important.

A series of numerical experiments were carried out to study the effect of Band Af on the flow quality expressed as the discharge ratio

RQ = QT

QB

where QT and Q B are the discharge through the top half resp. bottom half of a cross section. The value of RQ at the diffuser exit calculated as a function of B for three values of Af are plotted in Fig.3. The symbol S in the figure indicates flow separation or stall. Clearly stall can occur with sufficiently large values of B (blockage). Stall is more likely to occur with boundary layer thickness ratios Af less than unity (top wall boundary layer thinner).

The computed diffuser exit velocity distributions for five different values of 8t/8 are plotted in Fig.4. Note that 8t = 0, 0.5, and 1.0 represent horizontal top wall, symmetrical diffuser, and horizontal bottom wall cases, respectively. The same inflow velocity profile was used for all cases. Also shown is the measured velocity profile for the HYKAT design, i.e. 8t/8 = O. The agreement between theory and experiment was deemed to be adequate.

4.4. Vaned Elbows

Three types of turning vane profiles were considered for the first elbow located at the downstream end of the diffuser. The first was based on the N ACA four digit airfoil series; the other two were circular arcs and differed in thickness. A cascade of 24 vanes was selected for study with a space-chord ratio of one-half.

Of particular interest was the pressure distribution on the vanes and the turning action of the vanes. The peak negative pressure near the leading edge of the vane must be small to prevent cavitation. The adverse pressure gradient on the suction side should not cause boundary layer separation.

The calculated pressure distribution on a circular arc vane with 14% thickness ratio is shown in Fig.5. The pressure coefficient is

P~Po Cp = 1/2pVJ

where Po and Vo are the average pressure and velocity at the elbow entrance.

Computations of the critical cavitation number at the diffuser exit indicate that -Cp should be < 1.8. As shown in Fig.5 the pressure distribution of the vanes show a considerable margin against cavitation. One of the most important performance criteria, the uniformity of the exit flow, was studied in detail. The vanes gave good turning action.

100

Schiffstechnik Bd. 37 - 1990

Inside y

r-In_sid_e =~:~~:_V---+-~_Ins_ide __ ~_--rt---:..V / \V

V

Y/H

0.8

0.6

o MEASURED - MODELED

0.4

0.2

0~-----0~~------~------~----~

o

0.5

1.0

1.5

TIME AVERAGED VELOCITY

Fig. 4: Effect of Asymmetry on Exi t Distribution

Diffuser velocity

y

"2.01""""""--------....,...-----------------------..-,

\

I I

I

I .

CL •• ' ••••• ,

w ~i\":// ,'

0.0 -.-.-.~,L._._._._._._._._"'_,.

-"~" .. j "-

',,\ r, ,: . ....",......... ."- .>: "-

\ : ._ . -\Lo;;- s'U:ce

\:

Upper Surface

1. 0 1--,........."--~---f--r--.---,-- ....... --~.-~-.,,.,--~"""

"0.4

0.0

0.2 0,4

SIC

0.6

0.8

2.0

Fig. 5: Pressure Distribution on Recommended Vane

Inside

»>:

V...---..,...--..., Inside

Hub

Inside

1.0

Fig. 6: Variation of vertical Plane Velocity Profiles Around circuit

Schiffstechnik Bd. 37 - 1990

V

101

4.5 Pump Inlet Velocity Profiles

Considerable interest was directed towards the analysis of the pump inflow. Initial measurements indicated that the pump inflow was unacceptable. In order to determine possibilities for improvements, velocity profiles were taken at various planes in the flow circuit. It was found that the non-symmetry of the velocity profile originated at the exit of the main diffuser and was transmitted through the two elbows without significant change (Fig.6).

It was decided that modifications to the diffuser would be prohibitive; therefore, both the physical and the numerical models were used to arrive at a modified design. An appropriately modified mathematical model of the turning vanes was used to guide the direction of the physical model tests. The considerable improvement, Fig.7, was obtained with the design angle for the turning vanes in the first elbow, and 1.5° less turning of the vanes in the second one. The measured velocity distribution was corrected to full scale conditions using the numerical model to simulate the diffuser outflow at both model and prototype Reynolds number (Song et al. 1988).

Finally a ship model like those to be used in routine cavitation testing was installed in the roof of the model test section in order to determine its influence on the pump inflow velocity distribution. These effects were found to be small, Fig.8. In fact, the additional mixing due to the ship model wake improves the diffuser outflow.

5. Conclusions

The HYKAT represents the state of the art in hydrodynamic and hydroacoustic test facilities. Extensive studies assured that the design goals - uniform flow with small turbulence in the test section, very low noise levels - were met. Utilizing an interactive numerical and physical modelling technique, the main pump was designed. The pump operates cavitation free over a wide range of operating conditions, at least up to 8 mls with minimum pressure in the test section. A similar approach should be adopted for future designs.

References

BATCHELOR, G.K. (1953), Homogeneous turbulence, Cambridge University Press

LUMLEY, J.L. and McMAHON, J.F. (1967), Reducing water tunnel turbulence by means of a honeycomb, ASME, J. of Basic Engineering Vol. 89, pp. 764-770

ROBBINS, B. (1978), Water tunnel turbulence measurements behind a honeycomb, J. of Hydronautics, Vol. 12 No.3, pp. 122-128

SMAGORINSKY, J. (1963), General circulation experiments with the primitive equations, r. The basic experiment, Monthly Weather Review, Vol. 91, No 3

SONG, C.C.S., WETZEL, J.M., KILLEN, J.M., and ARNDT, R.E.A. (1988) Physical and mathematical modelling of the HYKAT, St. Anthony Falls Hydraulic Laboratory, Project Report No. 282

SONG, C.S.S., WETZEL, J.M., YUAN, M., ARNDT, R.E.A., and KILLEN, J.M. (1987), Hydrodynamic analysis of the HYKAT, St. Anthony Falls Hydraulic Laboratory, Project Report No 261

WEITENDORF, E.-A., FRIESCH,J., and SONG, C.C.S. (1987), Considerations for the new hydrodynamics and cavitation tunnel (HYKAT) of the Hamburg Ship Model Basin (HSVA), ASME Int. Symp. on Cavitation Research Facilities and Techniques, Boston

WEITENDORF, E.-A. and FRIESCH, J. (1990), Der HYKAT, die neue Versuchsanlage der HSVA - Einsatsmoglichkeiten und erste Ergebnisse, proceedings of Schiftbautechnische Gesellschaft

102

Schiffstechnik Bd. 37 - 1990

TOP

R/Ro

• 0.5 Final Design

o 0.95 Final Design .... 0.5 Initial Design -o- 0.95 Initial Design

v ~ Axial Velocity

VTS ~ Test Section Velocity


r--~ ~. . __ . ~--~
• ....
~ rl~ • 1t- • • ~ • •
~ • •
-< I • --- ~
~ • • •
..( • rl~
0 0 ( rv I~ rv ~
0 0
0 0 0 0 0
0 0 rl O.B

0.6 V/VTS

0.4

0.2

o

45

135

1BO

f) in Deqrees

225

270

90

315

360

Fig. 7: Pump Intake Axial Velocity Distribution Comparison Between Initial and Final Designs

100 0 0 I I 'JU o~cb
0 0 ' Cboo
0 0
i=' 0 0 00 0
0
z 0 0 0
UJ 80 I- 0 0
o
a: 00 0 0
UJ 00
e:- o
00 0
..J 0 0 0
..J 8
-c 60 00
~ 00
• WITHOUT MODEL
UJ Q)
o o WITH MODEL
iii (')
~ 0 ~
:E 40 I- 0
0 0
a: 0
I.L o
UJ 0
o 00 0
z 00 0
j:f 20 l- e 0
(/) 0:::> 0
Ci
0 0 0
~ 0 0
e I Or" o 0 0 I
0 c:;) I
0 0.2 0.4 0.6 0.8 1.0
VIVTS -

-

-

1.2

Fig. 8: comparison Between Vertical Velocity Distributions at Entrance to the Diffuser (Right Hand Curves) and at the Diffuser Exit Plane (Left Hand Curves) With and Without a Ship Model in the Test Section

Schiffstechnik Bd, 37 - 1990

103