INVESTIGATION ON SWIRL IN PUMP SUMP INTAKES AND ITS CONTROL

Abstract

Water pumps used in drainage, agriculture, and industrial process often experience operational problems such as vibration, impeller damage caused by cavitations, and excessive wear of bearings. This not only results in severe deterioration of their performance but also leads to a significant increase in operational and maintenance costs. These problems probably result from non-axial flow of water in the suction pipe because of poor pump-sump design. These problems and solutions have been investigated extensively and are widely reported in the literature (Denny 1956; Gordon 1970; Tullis 1979; Melville et al. 1994). Unfortunately, the flow phenomena near the suction pipes are so complex and diverse that there is no comprehensive theoretical model to predict them. Existing design guides usually contain little more than rules of thumb for design of pump sump intakes and control of swirl. In general, the intake structures are designed based on Hydraulic Institute Standards (HIS, 1998), British Hydromechanics Research Association (BHRA, Prosser 1977) standards, and the Japan Society of Mechanical Engineers (JSME, 1984) standards. These standards recommend construction of physical scaled models in a laboratory to design a problem free pump sump intakes. Padmanabhan and Hecker (1984) found that scaled models typically have unavoidable scale effects as swirl flow in pipe and inlet losses show scale effects. Moreover, the similarity laws of the model testing are not clear yet, and there are different opinions for velocity setting in the model test (Turbomachinery Society Japan 2002). Some research works recommend to perform the model study of pump sump for higher Froude number (up to three times of Froude number corresponding to Froudian conditions). However, other studies indicate that the pump sump model study is to be carried out for Froudian condition only. The time and cost involved in sump model studies for design and optimization of sump geometry can be reduced to a large extent through CFD modelling. A number of CFD studies are reported in the literature such as Pradeep et al. (2012), Tang et al. (2011), Kang et al. (2011), Skerlavaj et al. (2009). They found that floe simulation in the pump sump intake can be accurately carried out using CFD. Various types of anti-vortex devices (AVD) are used to control the swirling flow in pump sump intake (HIS 1998; BHRA 1977). However, the suitability of these vortex breakers in the light of flow conditions approaching the bell mouth are not discussed in literature. vii Swirl produced by vortex flow decays with distance along the suction pipe due to shear at the pipe wall. As the tangential velocity increases with distance from the pipe centre, wall friction acts on the near maximum tangential velocities, producing maximum dissipation forces (Knauss 1987). Knowledge of the decay of swirl with distance is important in determining possible effects of swirl on pump and turbine performance. Limited studies are reported in literature in respect of decay of swirl. Baker and Sayre (1974) showed that the swirl decay follows an exponential function. The present study is being taken up keeping in mind the above gaps in the design and model testing of multiple pump sumps. The specific objectives of the present investigation are as follows: 1. To study swirl flow in the pump sump intake through experimentation and CFD under different approach flow to the bay. 2. To study the effect of Froude number on swirl in the pump intake through experimentation and CFD. 3. To study the effect of different types of anti-vortex devices and their dimensions on the swirl in the pump intake and to standardized the shape and size of AVDs taking into considerations the approach flow conditions. 4. To study the decay of swirl in suction pipe column of the pumps. A 1:12 scaled model of a prototype of eight bays pump sump intake was chosen for the present study. Froude number dynamic similarity law was followed for the model study. The bell entrance Reynolds and Weber numbers were kept higher than 6104 and 320 respectively for neglecting the effect of viscosity and surface tension. The model was consisted of an intake chamber, fore bay, eight bays of the pump sump and suction pipe with bell mouth. Suitable arrangements in form of perforated walls, flow straighteners, and wave suppressors were provided in the intake chamber to stabilize the flow without any disturbance in the inflow. The model was made of concrete except the floor and the front view of the bays, which were of transparent glass to visualize the flow. The bell mouth and suction pipe were also fabricated of transparent plastic. A vortimeter was fitted in the suction pipes to measure the swirl angle of the flow. Acoustic Doppler Velocimeter was used to measure the temporal and spatial variation of velocity in bays and approach channel. An ultrasonic flow meter was used to measure the discharge in the suction lines for calibration of bend meters fitted in the suction lines. A pointer gauge of 0.1 mm accuracy was viii used to measure the water level in the sump. An OTT acoustic digital current meter (ADC) was used to measure one dimensional (1D) velocities in the bays. The following observations were taken in each experimental run: 1. The possible existence of subsurface or Free surface vortices were explored by visualization. 2. The revolutions per minute of the vortimeter were counted, and the swirl angle was calculated for each suction line. 3. Velocity distributions in the bays were measured for selected runs. The experimental runs were conducted for maximum discharge (13.69 L/s) and minimum water level (666 mm) for all combinations of pump operation. Experiments were also run for higher discharges at minimum water level for studying effect of Froude number on swirl. Different combinations of the pumps were chosen considering the symmetry of flow in the system and the time required running the model. Swirl and velocity in bays for design discharge, maximum and two times of design discharge were also observed to validate the CFD model results. Cross type anti-vortex device (AVD) of three heights i.e, 25mm, 50 mm and 75 mm have been tested for control of swirl in suction pipes under minimum water level and maximum discharge conditions in the pump sump model. It was found that as height increases, the effectiveness of the breaker increases with respect to swirl in combination A. Three different sizes of cross type anti-vortex devices are used in the physical model below the bell mouth intake for the control of swirling flow in the suction lines. Out of the three sizes of AVD, the middle size is found more effective in controlling the swirling flow. To find the most optimum AVD, a suitability index is developed and used in present study. From this AVD of 50 mm height is found most suitable in all the combinations of pumps and flow conditions. Nevertheless the higher size of the AVD breaks the swirling nature of the incoming flow, however, due to blockage effect it induces additional swirl which results in higher swirl compared to medium size of the AVD. A CFD model was developed for the pump sump intake model. A 3 m length of approach channel, forebay and eight bays configuration were modelled. The walls of suction pipes were ix considered as thin walls. A multi-block structured mesh with about 5.7 million hexahedral cells was generated using ANSYS ICEMCFD Hexa mesher. The sump model dimensions were same as in the physical model. The orthogonal quality of the mesh is of 0.37 on scale from 0 to 1 and the maximum aspect ratio was 13. O-Grid technique was used to create smooth mesh near the intakes. Two O grids were used - one inside the suction line and one outside the suction lines. The finite-volume method (FVM) was used to discretize the governing equations. QUICK scheme was used for convective term of the momentum equations. Second order discretization was employed for pressure equation. Pressure and velocity coupling were enforced using SIMPLEC algorithm. Steady state solutions were obtained by iterative method in pseudo time using Fluent solver. Accuracy of the results obtained from the FVM increases as the grid spacing decreased. To find the optimum spacing, grid dependency tests for three sizes of grid (coarse, fine and finer) have been performed. Standard k-ε turbulence model has been used in this study. At the inlet, a uniform velocity was taken as upstream boundary condition. At the outlets of suction pipes, outflow boundary condition was specified in which the gradients of all the variables were assumed to be zero. No-slip and no-penetration conditions were employed at the solid walls (bed, suction-line walls, side and back-walls). Standard wall function approach (Launder and Spalding 1974) was used to model the flow near solid walls. For the free surface, not considering shear stresses caused by the wind and heat transfer from the atmosphere, a rigid lid assumption was employed (i.e. symmetrical conditions). CFD model is verified by conducting mesh dependency test and the mesh dependence of the solution was estimated by running simulation for three hex meshes. To validate the CFD model, the swirl angle in suction lines and velocity distribution in bays are compared to the experimental results however results for one combination are discussed in detail. Results show good agreement with the experimental results. CFD modelling has been carried out to investigate the effectiveness of three shapes of the AVD i.e. vertical backwall splitter, L shape splitter and cross type. For each shape, three sizes of the AVD have been investigated. For each AVD, swirl angle in the suction line, flow pattern near the bell mouth, subsurface vortices have been investigated. It is found that L shape AVD of medium size is more effective in having favourable flow conditions in the suction lines. x Swirl angle for different Froudian conditions by changing flow rate in suction line were observed. It was found that as the Froude number (based on suction line diameter and axial velocity) increases, the swirl angle first increases then become constant. Therefore, it can be concluded that there is no need to perform model test at higher than Froude number conditions. Swirl angles produced by vortex flow decay with distance along the suction pipe due to shear at the pipe wall. The decay of swirl has been also found in CFD results. The decay of swirl was exponential in the present study which is same as reported by Baker and Sayre (1974). A new relationship between decay of swirl along the pipe and Froude number is proposed from the CFD results. The present study shall contribute immensely for the design of pump sump intakes, Flow simulation in the sump, and suction line, understanding of swirling flow and their control.