COPING WITH STRONG CROSS FLOWS AT PUMP INTAKES

 

TATSUAKI NAKATO

 

Iowa Institute of Hydraulic Research

The University of Iowa

Iowa City, Iowa 52242, U.S.A.

(319)-335-5228 -- telephone, (319)-335-5238 -- fax

tatsuaki-nakato@uiowa.edu -- e-mail

 

MARKUS KRÜTTEN

Universität Kaiserslautern

Fachbereich Maschienenbau und Verfahrenstechnik, Lehrstuhl für Strömungs- und Verdrängermaschinen

Postfach 3049 Gottlieb-Daimier-Straße

67653 Kaiserslautern, Germany

 

 

ABSTRACT

In order to improve nonuniform pump-approach flow distributions under strong cross-flow conditions, various flow-turning vanes were tested using a 1:10-scale rectangular pump-sump model. The area-averaged mean velocity ratio between the channel flow and pump intake flow was varied from 1.31 to 5.28. Simple vanes installed at the intake entrance were found to rectify nonuniform pump-approach flow distributions induced by flow separation. These vanes can be incorporated as part of the trashrack design.

 

Keywords: Pump Intake, Pump Sump, Pump Throat, Cross Flows, Velocity Distributions, Flow-Straightening Devices, Trashracks, ADV

 

INTRODUCTION

Many large-scale circulating-water (CW) pump intakes withdraw water at right angles from rivers where strong cross-flow conditions exist. This is particularly true when power plants are built along large steep rivers, such as the Missouri River and the Ohio River. CW pumps generally withdraw large volumes of water from rivers, e.g., up to 10 m³/s each, and pump capacities are becoming even larger. However, because of the site-specific nature of pump-intake design, such as space limitations, different types of travelling screens (single or dual flow), the need for warm-water re-circulation lines for de-icing during winter months, the need for combining CW-pump bays with service-water pump bays, etc., it is extremely difficult to produce a generic intake design for large pump installations. Furthermore, effects of cross flows on pump-approach-flow distributions are extremely difficult to evaluate without resorting to physical models. In order to ensure that CW pumps deliver reliable sources of condenser-cooling water, most intake designers consult with hydraulic laboratories for physical model studies. Laboratories generally build and test proposed intake structures on the basis of Froude's law using geometrically-undistorted scaled models whose scales typically range from 1:8 to 1:12. Iowa Institute of Hydraulic Research (IIHR), The University of Iowa, has conducted numerous hydraulic model studies on pump intakes in the past two decades. Presented in this paper are recent IIHR research results in improving pump-approach flow distributions within a rectangular pump sump under a variety of cross-flow conditions (Krütten and Nakato 1998).

 

EXPERIMENTAL SET UP AND TEST PROCEDURE

In order to produce cross-flow conditions, the intake model was provided with a 6,934 mm long, 914-mm deep, and 914-mm wide channel section, as shown in figure 1. A 607-mm wide and 1,067-mm long rectangular pump sump model was attached to the channel section. Ameren Corporation's Labadie Plant CW-pump bell was built at an undistorted model scale of 1:10 and attached to a 152-mm diameter suction column, as shown in Section A-A in figure 1. The bell diameter was 233.4 mm, the floor clearance was 117.5 mm, and the backwall clearance was 58.7 mm. Water discharge supplied to the model channel was varied between 28.3 l/s and 113.3 l/s and the pump discharge was maintained at 14.2 l/s. Water, supplied from a constant-head tank, passed through a diffuser box before entering the model channel basin upstream of the pump bay. Flow through the model suction line was simulated by means of a siphon and controlled by a butterfly valve.

 

Figure 1 : Schematic of model testing facility

 

The required model channel flow rate was first established and the siphon for the pump-suction line was activated next. The pump-sump water level was set to 478.5 mm above the sump floor by adjusting the tailgate located at the downstream end of the model basin, simulating the lowest water level at Labadie Plant. In each test, velocity distributions within the channel section and pump sump were measured using a three-dimensional Acoustic Doppler Velocimeter (ADV); velocity distribution at the model pump throat was measured by means of a pitot tube; and swirl angle was determined by means of a vortimeter (zero-pitched four-vane meter).

 

TEST RESULTS

Tests were conducted in four different cases to investigate pump-approach flow characteristics, pump-throat velocity distributions, and effectiveness of flow-straightening and vortex suppressors, including Case 1 - without any flow-training devices; Case 2 - with various flow-straightening devices at the pump-sump entrance; Case 3 - with only a floor splitter, sidewall floor-corner fillets, and a backwall floor-corner fillet; and Case 4 - with combinations of flow-straightening vanes and vortex-suppressing devices used in Case 3. In all four cases, the pump-intake discharge, Q_i, was maintained at 14.2 l/s and the model channel flow rate, Q_c, was varied between 28.3 l/s and 113.3 l/s.

 

In Case 1 (Q_c=113.3 l/s and Q_i=14.2 l/s), three-dimensional velocities were measured at twenty one locations along Sections A, B, and C in the model channel section and twenty two locations along Sections D and E within the pump sump. At each location five velocities were measured at equal intervals along the vertical, i.e., at z/h = 0.16, 0.32, 0.48, 0.64, and 0.80, where z = distance measured from the sump floor and h = sump water depth. Figures 2 and 3 show planar velocity distributions obtained near the sump floor (z/h = 0.16) and near free surface (z/h = 0.80), respectively. As can be seen in these figures, planar velocity vectors near the sump floor in front of the intake bay within the model channel were directed toward the pump sump, and there were strong reverse flows along the right (looking downstream) sidewall. Near the free surface, planar cross-flow vectors adjacent to the intake entrance were directed more or less parallel to the channel streamwise direction, and strong streamwise velocity components developed along the left sidewall of the pump sump. Transitions of velocity distributions from the sump floor to the free surface took place smoothly, as verified by the velocity data acquired at z/h = 0.32, 0.48, and 0.64. Pump-throat velocities measured at forty five locations along four different diameters were normalized and plotted in figure 4. The minimum and maximum normalized velocities were 0.60 and 1.23, respectively, and the swirl angle, defined as arctangent of the ratio of streamwise velocity and tangential vortimeter-tip velocity, was 16.9°. The commonly accepted swirl angle is 5°. Similar tests were also conducted for three different cross-flow discharges (Q_c=28.3 l/s, Q_c=56.6 l/s, and Q_c=85.0 l/s). However, the test results were very similar to those with Q_c=113.3 l/s. In summary, both pump-approach flow distributions and pump-throat velocity distributions were unacceptably nonuniform under strong cross-flow conditions.

 

In order to improve the nonuniformity in pump-approach flow, five configurations of vertical flow-turning vanes were placed at the entrance of the intake in Case 2. The first series of tests were conducted with forty-eight equally spaced 25.4-mm deep vanes spaced at 12.7-mm on centers. A significant improvement in the sump flow distributions was achieved. Reverse flows along the right sidewall were completely eliminated and lateral nonuniformity in velocity distributions along Section D and E became remarkably small when compared with those tested without vanes. Next the vane depth was further extended to 38.1 mm by maintaining spacing at 12.7 mm. No improvement with deeper vanes was recognized. Three other vane configurations, including a vertically segmented vane array with narrower vane spacing for the lower half of the flow depth and wider vane spacing for the upper half, were tested. Among them, 12.7-mm deep vanes placed on 12.7-mm centers were found to produce the best flow distributions for all four different discharge conditions, i.e., Q_c=28.3 l/s, Q_c=56.6 l/s, Q_c=85.0 l/s, and Q_c=113.3 l/s. Despite significant improvements in pump-sump flow distributions with these vanes, no improvement in pump-throat velocity distributions was obtained.

 

Case 3 studies involved suppressing formation of floor-attached and sidewall- and backwall-attached subsurface vortices. For this purpose, a floor splitter, and sump-floor corner fillets, shown in figure 5, were employed in the model. No boundary-attached vortices were found to form under these modifications.

 

Case 4 tests were conducted in conjunction with the vortex suppressors developed in Case 3 and 12.7-mm deep vanes placed on 12.7-mm centers. Figures 6 and 7 show planar velocity distributions obtained with the largest cross flow (Q_c=113.3 l/s). These figures show significant improvements in pump-approach flow distributions, as compared with those shown in figures 2 and 3. The distribution of pump-throat velocities obtained in this case is shown in figure 8. As compared with the distribution in figure 4, a significant improvement in the pump-throat velocity distribution can be seen. The minimum and maximum normalized velocities were 0.94 and +1.06, respectively, and the swirl angle was 8.0°. Although a backwall splitter would have reduced the swirl angle, no further modifications to reduce the swirl angle to an acceptable level were conducted.

 

Figure 2 : Planar velocity distributions obtained in Case 1 at z/h=0.16

(Q_c=113.3 l/s and Q_i=14.2 l/s)

 

Figure 3 : Planar velocity distributions obtained in Case 1 at z/h=0.80

(Q_c=113.3 l/s and Q_i=14.2 l/s)

 

Figure 4 : Three-dimensional plots of normalized pump-throat velocities

in Case 1 test with Q_c=113.3 l/s and Q_i=14.2 l/s

 

Figure 5 : Floor splitter and sump-floor corner fillets

 

Figure 6 : Planar velocity distributions obtained in Case 4 at z/h=0.16

(Q_c=113.3 l/s and Q_i=14.2 l/s)

 

Figure 7 : Planar velocity distributions obtained in Case 4 at z/h=0.80

(Q_c=113.3 l/s and Q_i=14.2 l/s)

 

Figure 8 : Three-dimensional plots of normalized pump-throat velocities in Case 4 test with Q_c=113.3 l/s and Q_i=14.2 l/s

 

CONCLUSIONS

Vertical, flow-turning vanes, 127-mm deep spaced at 127-mm on centers in full scale, can function as an excellent flow-straightening device for a prototype unit river discharge ranging from 0.98 m²/s to 3.9 m²/s when they are installed at the entrance of the rectangular pump intake operating at a mean prototype velocity of about 0.15 m/s. However, issues related to formation of free-surface and subsurface vortices must be treated individually with skimmer walls, splitters, corner fillets, etc. Pump-throat velocity distributions should be obtained for individual near pump-bell modifications and their uniformities should be checked by qualified engineers.

 

REFERENCES

Krütten, M., and Nakato, T. (1998). "Experimental studies of flow-straightening devices to improve approach-flow distributions at pump intakes under cross flows," (in press), Iowa Institute of Hydraulic Research, The University of Iowa, Iowa City, Iowa.