- Email: [email protected]
Experimental and numerical investigation of axial single scroll integral pumping devices for dual mechanical seals
Alexandria Engineering Journal (2018) 57, 2719–2728
H O S T E D BY
Alexandria University
Alexandria Engineering Journal www.elsevier.com/locate/aej www.sciencedirect.com
ORIGINAL ARTICLE
Experimental and numerical investigation of axial single scroll integral pumping devices for dual mechanical seals H.A. Warda, E.M. Wahba, A.B. Rashad, A.A. Mahgoub * Mechanical Engineering Department, Alexandria University, Alexandria 21544, Egypt Received 30 May 2017; revised 24 September 2017; accepted 21 October 2017 Available online 15 November 2018
KEYWORDS Dual mechanical seal; Axial pumping ring; Numerical study; Experimental; Computational fluid dynamics
Abstract Radial flow pumping rings receive the flow radially which is perpendicular on the shaft while axial flow pumping rings receive the flow in a direction parallel to the shaft (axially). An experimental investigation is carried out to evaluate the single axial pumping ring performance with a non API design radial clearance. The experimental setup is constructed according to API plan 53B with an accumulator-based seal support system in the barrier circuit loop. Furthermore, a numerical investigation is conducted to provide a reliable numerical model represented by ANSYS Fluent by creating an acceptable meshing and turbulence modeling using the experimental results as a validation method. As for the numerical investigation, it is indicated that K-epsilon realizable turbulence model is the best numerical model in terms of the least error deviation between the numerical model’s performance curve and the experimental performance curve. Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction Any simple mechanical seal has three basic components; rotating part, stationary part and secondary seals. The rotating part is the section which is attached to the shaft and rotates along with the shaft. This part mainly consists of primary ring which is usually made of carbon. The rotating assembly’s position is maintained by means of springs to uphold the small clearance required for sealing. The stationary part consists of a mating ring which is made of a harder material normally ceramics. * Corresponding author. E-mail addresses: [email protected] (H.A. Warda), akram. [email protected] (A.A. Mahgoub). Peer review under responsibility of Faculty of Engineering, Alexandria University.
While the secondary seals are crucial to leakage prevention in any metal to metal contact inside the mechanical seal. Single mechanical seals may not be enough to achieve the proper seal to some process fluids, since single mechanical seals have minimal lubricating film in the primary for lubrication and necessary facilitation for heat losses from the rotating primary ring. Even though the leakage may not be visible for some process fluids due to its flashing, this may even lead to further complication safety wise if the flashed process fluid is either toxic, flammable or highly corrosive which jeopardizes the safety of personnel and field operators in contact with that pump. Furthermore, the most obvious reason for double mechanical seal is that process fluid is quite expensive and losing this fluid decreases the plant efficiency. From a mechanical point of view, this type of seal is considered a backup option in case the inboard mechanical seal is damaged; this application
https://doi.org/10.1016/j.aej.2017.10.007 1110-0168 Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
2720
H.A. Warda et al.
Nomenclature BI BO DAQ e FI I k MRF P PI PS PT Q
Barrier Inlet Barrier Outlet Data acquisition turbulent dissipation rate Flow meter indicator unit tensor turbulence kinetic energy Multiple reference frame Pressure Pressure Indicator Pumping Scroll Pressure Transmitter Flowrate
can be applied to critical pumps in the plant where high reliability is required in order for the pump to operate for higher running hours. A single seal will not function correctly if the process fluid is not suitable for lubricating the running faces. Therefore, a separate fluid must be supplied to cool and lubricate the mechanical seals, especially if the fluid changes state or is highly unstable. Dual mechanical seals have both inboard and outboard seals. Inboard seals are lubricated by means of the process fluid normally as any single mechanical seals while the outboard seals are lubricated by means of separate fluid in a closed loop. This fluid’s condition is being specified according to the planned arrangement for the dual mechanical seal at hand. Dual mechanical seals have two arrangements; Arrangement two in which the outboard fluid’s pressure is lower than the inboard fluid’s pressure so as the outboard pressure acts as a buffer (which makes the common name known for it is ‘‘buffer fluid”), while as in Arrangement three ,the outboard fluid’s pressure is higher than the inboard fluid acting as a barrier fluid [1]. In fact, the performance of integral pumping devices and the flow dynamics inside the dual seal cavity is considered to be one of the topics that not many researchers tackled especially the axial pumping scroll since the double mechanical seal was introduced to industrial applications. Clark and Azibert [2] concluded by utilizing FLUENT to simulate flow field in the barrier fluid domain that large radial gap can increase the mechanical seal leakage risk despite the fact that large radial clearance is quite essential for the barrier fluids axial circulation for cooling the primary seal between the rotating and stationary rings to improve the seal’s lifetime and reliability. Moreover, Carmody et al. [3]as well utilized FLUENT to simulate the barrier fluid flow characteristics of double mechanical seals in API 682 sealing arrangements aiming to improve the flow characteristics through the internal cavities of standard seals by providing adequate head and flow rate requirements. Roddis and Carmody [4] researched the impact of using Bi-directional premium pumping rings and demonstrated the possibilities of saving energy and resources by using the pumping rings as well as testing the effect of different barrier fluid types on the rings performance.
RO RSM TI TO TT VI l ! m ! vr q ¼ s ! x
Radial Outlet port Reynolds stress model Tangential Inlet port Tangential Outlet port Temperature Transmitter Virtual Instruments molecular viscosity aboslute velocity relative velocity density stress tensor angular velocity
Furthermore, Smith [5] has investigated the challenge of using the highest reliable configuration of the mechanical seal by reviewing the advantages and the disadvantages of different types of arrangements for the double mechanical seal as well as the importance of using the integral pumping devices for cooling and circulation of the barrier fluid inside the outboard mechanical seal chamber. Richard also used FLUENT to simulate the baffle deflector in order to investigate the efficiency of the deflector in providing cooling effect for the inner seal thus improving the lifetime of both seals inner and outer. Finally, Warda et al. [6] experimental study was done different types of integral pumping devices radial and axial in different operating conditions - i.e. rotational speed, barrier fluid inlet temperature, barrier fluid pressure as well as outlet port configuration and radial clearance value - and they concluded that varying these conditions can in fact has an impact of the pumping ring performance in most cases. Added to that, they have conducted a CFD study utilizing FLUENT software to simulate the flow characteristics of radial pumping rings. It is concluded that k-epsilon standard equation produced acceptable results and radial outlet configuration has a lower performance output than the tangential due to separation phenomenon created at the radial outlet port causing eddies to form. Warda et al. [7] have conducted as well the effect of the kinematic viscosity on the radial integral pumping rings performance by measuring experimentally the ring’s performance against three barrier fluid types (Propylene Glycol-water mixtures and Diesel fuel (Grade D2) which impacted both maximum flow and differential pressure for the same pumping ring type and configuration and the significance of fluid density in the total differential pressure deduction. In the present study, an experimental investigation will be taking place in order to address the accuracy of the axial pumping scroll measurement as well as an improvement method in the pressure reading instrumentation in order to produce more reliable and accurate results. A numerical study will be performed as well in order to figure out which turbulence model would produce the numerical performance curve which matches the best with the experimental performance curve with by observing the least error deviation between both curves deduced in the experimental study. This numerical study will be performed by using ANSYS FLUENT software.
Experimental and numerical investigation 2. Experimental study The experimental setup is aimed at providing an indication of the integral pumping rings’ performance in Arrangement three configuration. In the following section, the barrier loop system which supplies the outboard seal with barrier fluid for the seal’s lubrication and cooling will be demonstrated in details. 2.1. Barrier loop system This system specifies the barrier flow from and to the outboard mechanical seal. As it was established the barrier loop will flow up to the specifications of API 682 plan 53B. Fig. 1 demonstrates a photographic and schematic overview of the barrier loop system. One of the barrier fluid loop’s main functions is to provide adequate flow to the outboard mechanical seal to ensure efficient cooling and lubrication supply to the seal. Therefore, the barrier loop system consists of several components:
Ball bladder type accumulator Shell and tube heat exchanger Coriolis mass flow meter (FI) Barrier Inlet and Outlet Pressure and Temperature transmitters (PT and TT) Accumulator pressure indicator (PI) Ball valves Interconnecting pipes Mechanical seal
API 682 has used the term ‘‘Accumulator-based seal support system” in describing both bladder type accumulator and a shell and tube heat exchanger. This indicates the importance of these components in supplying barrier fluid to the outboard seal, ensuring adequate conditions is being given to the barrier fluid, thus extending the Mechanical seal life time as well as its reliability.
Fig. 1
2721 The ball bladder accumulator main function is to provide the barrier loop the pressure needed to exceed the process fluid pressure thus maintaining the mechanical seal assembly to work in the arrangement three configuration. The pressure should be high enough to overcome the leakage rate in the pumping ring itself which is being achieved with a precharged ball bladder accumulator which provides a constant pressure. Another challenge the outboard mechanical seal encounters are the barrier fluid rising temperature once the pump runs. This is due to the rotation of the rotating assembly with the pumping ring thus heating up the primary seal clearance between the rotating and stationary part same as any conventional single mechanical seal. The ‘‘Accumulator-based seal support system” has provided the user to control the fluid temperature or more specifically the inner barrier temperature. This is achieved by controlling the heat load on the shell and tube heat exchanger by controlling the supply of the shell fluid into the heat exchanger which is in this case water. The barrier fluid passing through the tube side and throughout the barrier loop system in this experimental investigation is water as well (density 998.2 kg/cm3 at 25 °C). 2.1.1. Mechanical seal The test seal is John Crane type 48. It is 1.87500 diameter, multiple-spring dual pusher seal with Carbon/SiC/Carbon/ SiC faces and Fluorocarbon Rubber elastomers throughout. The mechanical seal is following the schematics of API 682 Arrangement 3 in which two assemblies (inboard and outboard seal) in one cartridge face to back position. As it was established, the barrier fluid is passing through the outboard seal in the cartridge where the pumping ring is installed. In the present study, the pumping ring type in which the performance test is investigated is the single scroll ring shown in Fig. 2. This ring is classified as axial ring type in which fluid enters the seal chamber in tangential direction through tangential inlet TI port. The flow then travels axially to the pumping ring, providing the fluid with pressure energy and flow. The fluid then exits the seal chamber through outlet port. The outlet port can be either radial or tangential port. Radial port
Photographic view of the barrier fluid loop setup (left) and Schematic drawing of the barrier fluid loop (right).
2722
Fig. 2 Three dimensional drawing of the single pumping scroll PS which is installed on the rotating part of the outboard seal using Solid works software.
indicates that the port is radial or perpendicular on the shaft direction while the tangential port is inclined at a specific angle which can be considered to be the best angle to align with the flow direction as it exits from the scroll ring in the seal chamber. In the present study, experimental investigation will be performed on the pumping ring bearing a radial clearance of 0.5 mm (non-API pumping scroll).
H.A. Warda et al. the different outlet port configuration tangential outlet (TO) and Radial outlet (RO) to ensure more reliable results. In experimental setup, the pressure transmitter was replaced and installed on the mechanical seal’s inlet and outlet sides as shown in Fig. 3. This action was due at that the differential pressure output is at the range of 0.01 to 0.25 bar which considered being too low in relative to the transmitter installed so the transmitter is replaced with another one with a lower ranged pressure transmitter and lower error span (measurement range 0–10 bar with accuracy ±0.10%) as well as inverted U-tube to improve reading which are in the second decimal place. The Coriolis flowmeter (measurement range 0– 20 lit/min with accuracy ±0.15% of actual measured flow rate) is not replaced in the current study [7]. As expected, the rotational speed is considered a significant factor that affects the performance output of the pumping ring either in terms of the maximum flowrate value or overall pressure output value [6,7]. The results match the deductions of Warda et al. [6,7] that the tangential outlet offers higher flow rate and pressure in the same speed than the radial outlet. In the radial outlet port configuration, the fluid has to change direction in order to exit the seal chamber through the radial port leading to flow separation and subsequent flow reduction while the tangential outlet port is aligned with the fluid direction as it exits the pumping ring which means neither flow separation nor flow reduction occur in this configuration. 4. Numerical study
3. Experimental accuracy and validation method Repeatability is a term that indicates overall consistency of the results extracted from a specific instrument or experimental procedure. It is imperative that the experimental setup which was discussed in the previous section is both accurate and reliable without the need to rebuild a new laboratory setup. This can be achieved by the test-retest reliability method. The concept that user has ensured that the laboratory setup piping is cleaned up and all the conditions are the same since it was first built Warda et al. [6] including recalibrating the transmitters and the data acquisition system. The first form of repeatability test is purely instrumental. Each and every single transmitter connected to the National Instrument’s Data Acquisition system, i.e. DAQ assistant in LABVIEW software, has been put into its settings that the signal which will be received on the user’s computer has 10,000 sample readings which means that the user has ten thousand readings by the time the signal is received. Afterwards, this signal is to be averaged in another Virtual Instruments (VI) file to receive only one last reading. Furthermore, another form of repeatability test retest method is to measure a previously known and measured pumping ring performance curve and validate it by the previous measurements. The selected pumping ring is the non API single pumping scroll since this will be the pumping ring type in which the numerical investigation will take place. The previous results have been acquired as a courtesy from Warda et al. work [6]. The same barrier pressure transmitter (measurement range 0–16 bar with accuracy ±0.10%) used in the repeatability test-retest method so as to maintain all variables constant and recreate the same conditions as the previous experiments had on investigating
In the present study, a CFD model will be formulated to deduce the most accurate representation to the flow in the outboard seal in terms of the turbulence modeling equations. Each turbulence modeling equation has its own characteristics with advantageous and disadvantageous components in predicting the flow fields with adverse conditions (i.e. separation, swirling). In the current numerical study, the outboard mechanical seal with a non API single pumping scroll ring has been subjected to the k-epsilon group stated in the next section for both outlet port configurations tangential and radial outlet. The validation reference is the experimental results which were acquired from the experimental setup described in the previous section. These results are shown in Fig. 4.
Fig. 3 . Photographic view of the mechanical seal area illustrating the two pressure transmitters.
2723
0.3
Pressure difference (bar)
Pressure difference (bar)
Experimental and numerical investigation
0.25 0.2 0.15 0.1
TO 3600 rpm TO 3000 rpm TO 1500 rpm TO 1000 rpm
0.05 0
0
1
2
3
4
5
6
Barrier flow rate (L/min)
0.2 0.15 0.1 RO 3600 rpm RO 3000 rpm RO 1500 rpm RO 1000 rpm
0.05 0
0
1
2
3
4
5
6
Barrier flow rate (L/min)
Fig. 4 Graphical representation of the experimental results of the single pumping scroll performance curve tangential outlet (left) and radial outlet (right).
4.1. Governing equations and numerical methods The CFD model is based on the Reynolds–averaged Navier Stokes equations: @q þ r q! mr¼0 @t @ ! ¼ q m þ r q! m r! m þq ! x ! m ¼ rp þ r s @t 2 ¼ T m rI s ¼ l r! m r þ r! mr r! 3
ð1Þ ð2Þ ð3Þ
Eq. (1) represents the conservation of mass where qq : is the density. Equation two represents conservation of momentum which consists of pressure gradient term, two additional acceleration terms: the Coriolis accerleration (2! x ! m r ) and the ! ! ! centripetal acceleration ( x x r ) which can be represented in absolute velocity formulation by q ! x ! m and the conventional stress tensor term on the equation’s right side. Equation three represents stress tensor term identification where l is the molecular viscosity, I is the unit tensor, and the second term on the right hand side is the effect of volume dilation [8]. In modeling, the turbulence equations are classified into both time and space averaging. In the time-averaged approach, the equations are classified into both turbulent stress model (RSM) which has 5 to 7 equations to solve and Eddy viscosity approach which has variety of equations from zero equation model to two equation model either k-epsilon or k-omega. In this study the k-epsilon approach is being used. There are three forms of k-epsilon approach equation: k-epsilon standard, k-epsilon realizable and k-epsilon RNG. According to Jones and Launder [9], the standard k-e model is a semiempirical model based on model transport equations for the turbulence kinetic energy (k) and its dissipation rate (e). The model transport equation for k is derived from the exact equation, while the model transport equation for e was obtained using physical reasoning and consists of similarities to its mathematically exact counterpart. This approach has several known limitations: separation prediction, swirling flows and flow with strong streamline curvature. The RNG k-e model was derived using a rigorous statistical technique (called renormalization group theory). It is similar in form to the standard k-epsilon model, but includes an additional term in its epsilon equation that significantly improves the accuracy for rapidly strained flows as well as effect of swirl on turbulence is
included in the RNG model, enhancing accuracy for swirling flows. According to Shih et al. [10], the realizable k-epsilon model differs from the standard k-epsilon model in two ways; new equation expressing the turbulent viscosity as well as the dissipation rate, e, which has been derived from an exact equation for the transport of the mean-square vorticity fluctuation. An immediate benefit of the realizable k-epsilon model is that it more accurately predicts the spreading rate of both planar and round jets as well as providing superior performance for flows involving rotation, boundary layers under strong adverse pressure gradients, separation, and recirculation. The commercial software ANSYS FLUENT [8] is used in this numerical study employing finite volume discretization of the governing equations. Convective terms are discretized using a second-order upwind scheme [8]. The SIMPLE algorithm [11,12] is used for pressure–velocity coupling. In the present study, the scalable wall function [13] is used as well. 4.2. Pre processing In the following section, the pre-processing stage of the CFD model for the non API single pumping scroll for both tangential and radial outlets will be discussed in details. The preprocessing is the first crucial stage in creating a viable CFD numerical model and it consists with both geometry as well as meshing. One of the difficulties encountered while creating the geometry is its complexity which was in two separate volumes; tangential inlet ports as well as scroll geometry. One of the unique properties that the axial integrated pumping device has is the tangential inlet in order to adjust the flow direction towards the scroll part of the ring in order to minimize the eddy losses which may occur otherwise. The tangential inlet geometry is quite tricky when drawing since the inclination of the port is not in the XY plane only but in both XY and YZ planes as shown in Fig. 6. Tangential inlet also contains another unique characteristic which is that the inlet area in the outer gland of the outboard mechanical seal contains two inclined ports to accommodate any different designs the scroll ring may have either left hand scroll or right hand scroll which is considered to be due to John Crane’s interest in making its outer gland case for this type of pumping ring generalized to be used in more than one application with any type of scroll. As for the scroll ring geometry, its complexity lies in two different challenging aspects. First of all, the scroll geometry
2724 has unique geometrical dimensional input in order to draw the correct scroll ring geometry i.e. (helix angle, cutting tool width, axial lead as well as scroll teeth depth). All of the previous properties helps define the actual scroll ring which provided the performance curve deduced from the experimental setup in Fig. 4. The geometry has been drawn using Solid Works 2015 software as shown in Fig. 2 utilizing the actual pumping ring as a reference. Furthermore, the non– API pumping scroll radial clearance is 0.5 mm. Therefore, when drawing the fluid volume as a part of the pre-processing stage the scroll cavities as well as the 0.5 mm radial clearance is the fluid volume representing the rotating scroll ring volume which increased the complexity of the three dimensional model drawn. This is considered to be the primary reason for author’s choice to use tetrahedral mesh cell type for representing such complex volume. In meshing, the geometry’s complexity is taken into consideration as well as the necessity to represent in a valid way the change of fluid volume alongside the assembly. Soft behaviors as well as the proximity and curvature properties are being implemented in the CFD model in order for the mesher to adapt against complex geometry conditions. Figures 5 and 7 illustrate the Meshing geometry in non API pumping scroll in both tangential and radial outlet port utilizing the proximity and curvature function to deduce a high quality mesh. Table 1 states some of the statistics of the Mesh for both outlet port configurations for the fine mesh. It is noticed that due to the
H.A. Warda et al.
Fig. 6 A screenshot of 3-D model of the tangential inlet port volume using Design Modeler software in ANSYS 14.0 software.
different outlet port geometry, the mesher deduces a higher number of elements to represent the radial outlet geometry than that of the tangential outlet configuration while using the same body sizing value as an input. This data has been deduced after performing the grid independence test successfully which be described in details in the next section. As shown in Figures 5 and 7, unstructured meshing technique was used for adapt meshing and to allow mass distribution on complex geometry of the scroll part in the CFD domain. Furthermore, similar to the previous work of Warda
Fig. 5 (a) Grid generation view of the radial outlet port assembly indicating the fluid volume meshing around the pumping ring, (b) Frontal view of the assembly revealing the grid generation around the radial port, (c) sectional view indicating the grid generation around three interfaces of three different subdomains in the radial outlet assembly.
Experimental and numerical investigation
2725
Fig. 7 (a) Grid generation view of the tangential outlet port assembly indicating the fluid volume meshing around the pumping ring, (b) Grid generation view of the tangential inlet port while suppressing the rest of the assembly (c) Frontal view of the assembly revealing the grid generation around the tangential port, (d) sectional view indicating the grid generation around three interfaces of three different subdomains in the tangential outlet assembly.
Table 1 Summary of the significant statistics of the fine mesh used to represent both tangential and outlet port configuration.
rate) is added as an input and pressure outlet where the pressure output value is deduced numerically.
Outlet configuration
Tangential
Radial
4.3. Grid independence test
Body sizing Number of Nodes Number of elements Max. Aspect ratio Min. Element quality Max. Skewness
5e-004 m 2,132,625 11,176,869 10.484 0.204 0.8214
2,353,325 12,333,240 11.257 0.2035 0.8240
In the following section, the validity of the meshing grid in representing both tangential and radial outlet will be tested by using grid independence test. This test is a measure of the discretization accuracy of the numerical model. This is achieved by running the numerical model on several different grid specifications: coarse, medium, fine meshes and comparing the numerical model outcome results. Once the results are constant among the meshes, this indicates that the grid is not considered a variable factor that affects the discretization accuracy of the numerical solution. In the current study, the grid independence test is performed on both tangential and radial outlet port numerical models. The results are shown in Table 2 and 3 where the meshes results are being deduced utilizing the maximum flow-
[6] in the radial integrated pumping ring meshing representation, the multiple reference frame (MRF) technique has been used in this study to represent the zone domains of the CFD model. The rotating zone or frame has been selected for the scroll ring zone. The distinct difference in that approach is that since the non-API radial clearance from the scroll teeth to the wall of the outer gland is too small (0.5 mm), MRF approach was used in such a way that the whole zone above the ring is in fact the rotating zone while the wall outer gland is the stationary zone as well as stationary wall which will be considered as a boundary condition. The rest of the ring geometry which has no teeth in it is considered to be a rotating wall as an additional factor in the boundary condition input. The rest of the zone starting from the inlet and the outlet piping, fittings as well as ports are to be stationary frame. The approach used in the numerical study is velocity inlet where velocity (i.e. flow
Table 2 Details of grid used in grid independence test in tangential outlet numerical model. Grid
Number of nodes
Number of elements
Pressure output (bar)
Coarse Medium Fine
1,025,845 1,492,879 2,132,625
5,234,189 7,601,932 11,176,869
0.230957 0.211589 0.206957
2726
H.A. Warda et al.
Table 3 Details of grid used in grid independence test in radial outlet numerical model. Grid
Number of nodes
Number of elements
Pressure output (bar)
Coarse Medium Fine
1,123,562 1,578,462 2,353,325
5,565,754 8,024,186 12,333,240
0.084596 0.089112 0.087807
rate value of the tangential (5.2 lit/min) and radial outlet (3.8 lit/min) as an input. It is deduced that as the number of cells increases, from coarse to fine mesh, the steady state value of the pressure output almost remains constant for the same flow rate input value which indicates the fine mesh size stated in Table 1 converges on the most accurate results. Therefore, the fine mesh approach will be used in the current study to deduce numerically the best turbulence model that represents the flow field inside the outboard seal. The numerical investigation is performed on a computer with specifications of 32 GB RAM and processor IntelÒ XeonÒ Processor E5-2630 v4 with run time reaching 9–12 days till convergence. 5. Results and discussions In order to extensively test the compatibility of the turbulence model to the CFD model, three flow points were investigated.
These three points are the zero flow, mid-flow and maximum flow. Furthermore, the CFD model is to be subjected to two different rotational speeds; 1500 rpm and 3600 rpm considering that these are the common used rotational speeds used in the industrial applications. Tangential outlet configuration CFD numerical model results shown in Fig. 8 shows that k-epsilon RNG and kepsilon standard modeling results are significantly close but not accurate while the k-epsilon realizable results are relatively the least error deviation. It is noticeable as well that by decreasing the rotation speed, the numerical models are more successful in producing relatively more accurate results in relation to the experimental performance curve with the k-epsilon realizable being the most accurate turbulence model of the three k-epsilon equations as well. As for the radial outlet configuration, the numerical results in Fig. 9 reveal as well the same conclusion about the k-epsilon realizable turbulence equation being relatively the highest in the accuracy of the three equation with the remaining two being close but not accurate. The three turbulence model is more successful in the deduction the performance curve in three flow points with lower rotational speeds. It is noticeable that radial outlet CFD model has relatively higher error deviation from the experimental performance curve for the same turbulence model equation that the tangential outlet numerical model which was the same deduction in previous work of Warda [6] for the radial integral pumping ring.
Fig. 8 Graphical representation of the tangential outlet performance curve against all three k-epsilon equations in speeds 3600 rpm (left) and 1500 rpm (right).
Fig. 9 Graphical representation of the radial outlet performance curve against all three k-epsilon equations in speeds 3600 rpm (left) and 1500 rpm (right).
Experimental and numerical investigation
Fig. 10
2727
Pressure distribution on the scroll ring wall.
Fig. 11 (a) radial pressure distribution in the non API clearance around the pumping ring, (b) pressure distribution in one of the helical scrolls in the pumping ring, (c) Frontal view of the axial pumping ring exit indicating the barrier fluid’s maximum pressure value on each scroll.
Utilizing CFD-Post software, the numerical investigation is extended to comprehend the pressure distribution along the single axial pumping scroll assembly. One of the main observations on the post processing results, the symmetrical replication of all the flow field properties as it is considered nearly the same for each tooth alongside the circumference of the
ring. Fig. 10 is considered as one of the crucial pressure contours representations to help understand the flow field inside the outer seal. It is noticeable that the pressure is almost constant outside the scroll teeth region on the scroll ring except for certain low pressure spots alongside the circumference of the ring. The number of these spots matches exactly the number
2728 of scroll teeth in the circumference which indicates that due to the fluid flow inside the scroll teeth which rotates with the assembly acting as a rotor pump, these spots are considered to be suction region for each of the scroll cavity. These spots have a concentric nature in pressure distribution decreasing gradually as shown in Fig. 10 reaching its minimum value in its center. As the flow field advances toward the exit side of the pumping scroll, the pressure varies in an exponential increase reaching its maximum at the tip of the teeth midway the scroll lead. The maximum pressure area increases alongside the tooth until it overwhelms the whole side of the tooth at the end of each tooth. As predicted from the results of Fig. 10, the barrier fluid pressure increases exponentially as fluid progresses towards the exit port. This is revealed in Fig. 11a which shows the pressure contour distribution axially as the fluid moves along the ring assembly as well as in side view position revealing the circumferential pressure distribution of the barrier fluid on the ring as it exits. Fig. 11b also illustrates the pressure contour plane captured in the mid area of one of the scroll cavity which indicates an increase of the pressure value as it progresses through the cavity reaching its maximum at the exit end which is shown in Fig. 11c. It is noticed that at the beginning of the scroll’s helical tooth, the radial plane in CFD-Post captures the pressure of its minimum value which matches the previous conclusion in Fig. 10 being the lowest possible pressure value of the assembly. As the fluid progresses axially alongside the radial plane, the barrier fluid pressure value increases normally until it reaches its maximum value on the exit side. It is also noticeable that the pressure at the topside of the radial clearance is higher than the pressure near the wall which is due to the centrifugal pressure building up from the pumping scroll rotation. Furthermore, when observing the side view indication, the pressure distribution of all the exit barrier fluid pressure in each of the scroll cavities in the exit, few conclusions are made: First of all, the pressure in each of scroll cavity is relatively the highest on the left wall side and decreases slightly when progressing to the cavity’s right wall. This occurs due to the stagnation pressure of the barrier fluid as the ring rotates in a clockwise direction. Last but not least, when observing each of the pressure values of each of the scroll cavities, it is observed that the pressure distribution profile is symmetrical on each of the scroll tooth along the circumference. 6. Conclusions The experimental results vary according to the shaft speed as well as the outlet port configuration. The experimental performance curve results have been used as a validation method while undergoing the numerical investigation for both tangential and radial outlet model. Moreover, the numerical results have shown acceptable performance curve using the kepsilon realizable model which has higher accuracy in repre-
H.A. Warda et al. senting the experimental performance curve than the rest of the k-epsilon family. Finally, the post processing data reveals that the barrier fluid pressure increase as it progresses through the scroll teeth until it reaches the maximum at the ring exit. Acknowledgments The authors would like to acknowledge E.A.Selim and M.W Gamal Aldin for their assistance and support in the experimental and the numerical investigations respectively and John crane UK for providing the mechanical seal assembly. References [1] API, Pumps — Shaft Sealing Systems for Centrifugal and Rotary Pumps ANSI / API Standard 682 ISO 21049: 2004, Pumps — Shaft Sealing, 2005. [2] R. Clark, H. Azibert, visualizing fluid flow and heat transfer in rotating shaft seals., In: 15th Int. Conf. Fluid Seal., BHR group conference series publication, 1997: pp. 353–378. [3] C. Carmody, A. Roddis, J. Amaral Teixeira, D. Schurch, Integral pumping devices that improve mechanical seal longevity., In: 19th Int. Conf. Fluid Seal., BHR group conference series publication, UK, 2007: pp. 235–247. [4] C. Carmody, A. Roddis, Saving Energy, Saving Water and Saving the Planet through the use of Affordable, Premium BiDirectional Pumping Rings, In: Des. Optim. through Value Eng. Fluid Mach., IMECHE, 2008: pp. 1–15. [5] R. Smith, Contradicting requirements in API 682 dual seal design configurations, Rotherham, England, 2008. http://www. aesseal.com/en/resources/whitepapers/contradictingrequirements-api-682-dual-seal-design-configurations [6] H.A. Warda, E.M. Wahba, E.A. Selim, Integral pumping devices for dual mechanical seals: experiments and numerical simulations, J. Eng. Gas Turbines Power. 137 (2014) 22504, https://doi.org/10.1115/1.4028384. [7] H. A.Warda, I.G. Adam, A.B. Rashad, M.W. Gamal Aldin, Effect of Kinematic Viscosity of Barrier Fluids on The Performance of a Bi-Directional Integrated Pumping Ring for Dual Mechanical Seals, In: ASME Proc. | 17th Symp. Turbomach. Flow Simul. Optim., Washington, DC, USA, 2016: pp. 1–36. http://doi.org/10.1115/FEDSM2016-7763 [8] ANSYS, ANSYS FLUENT Theory Guide, Release 14.0., ANSYS Inc., Canonsburg, PA, 2011. [9] W.P. Jones, B.E. Launder, The prediction of laminarization with a two-equation model of turbulence, Int. J. Heat Mass Transf. 15 (1972) 301–314. [10] T.-H. Shih, W.W. Liou, A. Shabbir, J. Zhu, A new k-e eddyviscosity model for high reynolds number turbulent flows model development and validation, Computer. Fluids 24 (1995) 227–238. [11] S. Patankar, Numerical Heat Transfer and Fluid Flow, first ed., Hemisphere, Washington, DC, USA, 1980. [12] Z. Luan, M.M. Khonsari, Numerical simulations of the flow field around the rings of mechanical seals, J. Tribol. 128 (2006) 559–565, https://doi.org/10.1115/1.2197845. [13] M. Chmielewski, M. Gieras, Three-zonal wall function for k-e turbulence models, Cmst 19 (2013) 107–114, https://doi.org/ 10.12921/cmst.2013.19.02.107-114.
H O S T E D BY
Alexandria University
Alexandria Engineering Journal www.elsevier.com/locate/aej www.sciencedirect.com
ORIGINAL ARTICLE
Experimental and numerical investigation of axial single scroll integral pumping devices for dual mechanical seals H.A. Warda, E.M. Wahba, A.B. Rashad, A.A. Mahgoub * Mechanical Engineering Department, Alexandria University, Alexandria 21544, Egypt Received 30 May 2017; revised 24 September 2017; accepted 21 October 2017 Available online 15 November 2018
KEYWORDS Dual mechanical seal; Axial pumping ring; Numerical study; Experimental; Computational fluid dynamics
Abstract Radial flow pumping rings receive the flow radially which is perpendicular on the shaft while axial flow pumping rings receive the flow in a direction parallel to the shaft (axially). An experimental investigation is carried out to evaluate the single axial pumping ring performance with a non API design radial clearance. The experimental setup is constructed according to API plan 53B with an accumulator-based seal support system in the barrier circuit loop. Furthermore, a numerical investigation is conducted to provide a reliable numerical model represented by ANSYS Fluent by creating an acceptable meshing and turbulence modeling using the experimental results as a validation method. As for the numerical investigation, it is indicated that K-epsilon realizable turbulence model is the best numerical model in terms of the least error deviation between the numerical model’s performance curve and the experimental performance curve. Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction Any simple mechanical seal has three basic components; rotating part, stationary part and secondary seals. The rotating part is the section which is attached to the shaft and rotates along with the shaft. This part mainly consists of primary ring which is usually made of carbon. The rotating assembly’s position is maintained by means of springs to uphold the small clearance required for sealing. The stationary part consists of a mating ring which is made of a harder material normally ceramics. * Corresponding author. E-mail addresses: [email protected] (H.A. Warda), akram. [email protected] (A.A. Mahgoub). Peer review under responsibility of Faculty of Engineering, Alexandria University.
While the secondary seals are crucial to leakage prevention in any metal to metal contact inside the mechanical seal. Single mechanical seals may not be enough to achieve the proper seal to some process fluids, since single mechanical seals have minimal lubricating film in the primary for lubrication and necessary facilitation for heat losses from the rotating primary ring. Even though the leakage may not be visible for some process fluids due to its flashing, this may even lead to further complication safety wise if the flashed process fluid is either toxic, flammable or highly corrosive which jeopardizes the safety of personnel and field operators in contact with that pump. Furthermore, the most obvious reason for double mechanical seal is that process fluid is quite expensive and losing this fluid decreases the plant efficiency. From a mechanical point of view, this type of seal is considered a backup option in case the inboard mechanical seal is damaged; this application
https://doi.org/10.1016/j.aej.2017.10.007 1110-0168 Ó 2018 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
2720
H.A. Warda et al.
Nomenclature BI BO DAQ e FI I k MRF P PI PS PT Q
Barrier Inlet Barrier Outlet Data acquisition turbulent dissipation rate Flow meter indicator unit tensor turbulence kinetic energy Multiple reference frame Pressure Pressure Indicator Pumping Scroll Pressure Transmitter Flowrate
can be applied to critical pumps in the plant where high reliability is required in order for the pump to operate for higher running hours. A single seal will not function correctly if the process fluid is not suitable for lubricating the running faces. Therefore, a separate fluid must be supplied to cool and lubricate the mechanical seals, especially if the fluid changes state or is highly unstable. Dual mechanical seals have both inboard and outboard seals. Inboard seals are lubricated by means of the process fluid normally as any single mechanical seals while the outboard seals are lubricated by means of separate fluid in a closed loop. This fluid’s condition is being specified according to the planned arrangement for the dual mechanical seal at hand. Dual mechanical seals have two arrangements; Arrangement two in which the outboard fluid’s pressure is lower than the inboard fluid’s pressure so as the outboard pressure acts as a buffer (which makes the common name known for it is ‘‘buffer fluid”), while as in Arrangement three ,the outboard fluid’s pressure is higher than the inboard fluid acting as a barrier fluid [1]. In fact, the performance of integral pumping devices and the flow dynamics inside the dual seal cavity is considered to be one of the topics that not many researchers tackled especially the axial pumping scroll since the double mechanical seal was introduced to industrial applications. Clark and Azibert [2] concluded by utilizing FLUENT to simulate flow field in the barrier fluid domain that large radial gap can increase the mechanical seal leakage risk despite the fact that large radial clearance is quite essential for the barrier fluids axial circulation for cooling the primary seal between the rotating and stationary rings to improve the seal’s lifetime and reliability. Moreover, Carmody et al. [3]as well utilized FLUENT to simulate the barrier fluid flow characteristics of double mechanical seals in API 682 sealing arrangements aiming to improve the flow characteristics through the internal cavities of standard seals by providing adequate head and flow rate requirements. Roddis and Carmody [4] researched the impact of using Bi-directional premium pumping rings and demonstrated the possibilities of saving energy and resources by using the pumping rings as well as testing the effect of different barrier fluid types on the rings performance.
RO RSM TI TO TT VI l ! m ! vr q ¼ s ! x
Radial Outlet port Reynolds stress model Tangential Inlet port Tangential Outlet port Temperature Transmitter Virtual Instruments molecular viscosity aboslute velocity relative velocity density stress tensor angular velocity
Furthermore, Smith [5] has investigated the challenge of using the highest reliable configuration of the mechanical seal by reviewing the advantages and the disadvantages of different types of arrangements for the double mechanical seal as well as the importance of using the integral pumping devices for cooling and circulation of the barrier fluid inside the outboard mechanical seal chamber. Richard also used FLUENT to simulate the baffle deflector in order to investigate the efficiency of the deflector in providing cooling effect for the inner seal thus improving the lifetime of both seals inner and outer. Finally, Warda et al. [6] experimental study was done different types of integral pumping devices radial and axial in different operating conditions - i.e. rotational speed, barrier fluid inlet temperature, barrier fluid pressure as well as outlet port configuration and radial clearance value - and they concluded that varying these conditions can in fact has an impact of the pumping ring performance in most cases. Added to that, they have conducted a CFD study utilizing FLUENT software to simulate the flow characteristics of radial pumping rings. It is concluded that k-epsilon standard equation produced acceptable results and radial outlet configuration has a lower performance output than the tangential due to separation phenomenon created at the radial outlet port causing eddies to form. Warda et al. [7] have conducted as well the effect of the kinematic viscosity on the radial integral pumping rings performance by measuring experimentally the ring’s performance against three barrier fluid types (Propylene Glycol-water mixtures and Diesel fuel (Grade D2) which impacted both maximum flow and differential pressure for the same pumping ring type and configuration and the significance of fluid density in the total differential pressure deduction. In the present study, an experimental investigation will be taking place in order to address the accuracy of the axial pumping scroll measurement as well as an improvement method in the pressure reading instrumentation in order to produce more reliable and accurate results. A numerical study will be performed as well in order to figure out which turbulence model would produce the numerical performance curve which matches the best with the experimental performance curve with by observing the least error deviation between both curves deduced in the experimental study. This numerical study will be performed by using ANSYS FLUENT software.
Experimental and numerical investigation 2. Experimental study The experimental setup is aimed at providing an indication of the integral pumping rings’ performance in Arrangement three configuration. In the following section, the barrier loop system which supplies the outboard seal with barrier fluid for the seal’s lubrication and cooling will be demonstrated in details. 2.1. Barrier loop system This system specifies the barrier flow from and to the outboard mechanical seal. As it was established the barrier loop will flow up to the specifications of API 682 plan 53B. Fig. 1 demonstrates a photographic and schematic overview of the barrier loop system. One of the barrier fluid loop’s main functions is to provide adequate flow to the outboard mechanical seal to ensure efficient cooling and lubrication supply to the seal. Therefore, the barrier loop system consists of several components:
Ball bladder type accumulator Shell and tube heat exchanger Coriolis mass flow meter (FI) Barrier Inlet and Outlet Pressure and Temperature transmitters (PT and TT) Accumulator pressure indicator (PI) Ball valves Interconnecting pipes Mechanical seal
API 682 has used the term ‘‘Accumulator-based seal support system” in describing both bladder type accumulator and a shell and tube heat exchanger. This indicates the importance of these components in supplying barrier fluid to the outboard seal, ensuring adequate conditions is being given to the barrier fluid, thus extending the Mechanical seal life time as well as its reliability.
Fig. 1
2721 The ball bladder accumulator main function is to provide the barrier loop the pressure needed to exceed the process fluid pressure thus maintaining the mechanical seal assembly to work in the arrangement three configuration. The pressure should be high enough to overcome the leakage rate in the pumping ring itself which is being achieved with a precharged ball bladder accumulator which provides a constant pressure. Another challenge the outboard mechanical seal encounters are the barrier fluid rising temperature once the pump runs. This is due to the rotation of the rotating assembly with the pumping ring thus heating up the primary seal clearance between the rotating and stationary part same as any conventional single mechanical seal. The ‘‘Accumulator-based seal support system” has provided the user to control the fluid temperature or more specifically the inner barrier temperature. This is achieved by controlling the heat load on the shell and tube heat exchanger by controlling the supply of the shell fluid into the heat exchanger which is in this case water. The barrier fluid passing through the tube side and throughout the barrier loop system in this experimental investigation is water as well (density 998.2 kg/cm3 at 25 °C). 2.1.1. Mechanical seal The test seal is John Crane type 48. It is 1.87500 diameter, multiple-spring dual pusher seal with Carbon/SiC/Carbon/ SiC faces and Fluorocarbon Rubber elastomers throughout. The mechanical seal is following the schematics of API 682 Arrangement 3 in which two assemblies (inboard and outboard seal) in one cartridge face to back position. As it was established, the barrier fluid is passing through the outboard seal in the cartridge where the pumping ring is installed. In the present study, the pumping ring type in which the performance test is investigated is the single scroll ring shown in Fig. 2. This ring is classified as axial ring type in which fluid enters the seal chamber in tangential direction through tangential inlet TI port. The flow then travels axially to the pumping ring, providing the fluid with pressure energy and flow. The fluid then exits the seal chamber through outlet port. The outlet port can be either radial or tangential port. Radial port
Photographic view of the barrier fluid loop setup (left) and Schematic drawing of the barrier fluid loop (right).
2722
Fig. 2 Three dimensional drawing of the single pumping scroll PS which is installed on the rotating part of the outboard seal using Solid works software.
indicates that the port is radial or perpendicular on the shaft direction while the tangential port is inclined at a specific angle which can be considered to be the best angle to align with the flow direction as it exits from the scroll ring in the seal chamber. In the present study, experimental investigation will be performed on the pumping ring bearing a radial clearance of 0.5 mm (non-API pumping scroll).
H.A. Warda et al. the different outlet port configuration tangential outlet (TO) and Radial outlet (RO) to ensure more reliable results. In experimental setup, the pressure transmitter was replaced and installed on the mechanical seal’s inlet and outlet sides as shown in Fig. 3. This action was due at that the differential pressure output is at the range of 0.01 to 0.25 bar which considered being too low in relative to the transmitter installed so the transmitter is replaced with another one with a lower ranged pressure transmitter and lower error span (measurement range 0–10 bar with accuracy ±0.10%) as well as inverted U-tube to improve reading which are in the second decimal place. The Coriolis flowmeter (measurement range 0– 20 lit/min with accuracy ±0.15% of actual measured flow rate) is not replaced in the current study [7]. As expected, the rotational speed is considered a significant factor that affects the performance output of the pumping ring either in terms of the maximum flowrate value or overall pressure output value [6,7]. The results match the deductions of Warda et al. [6,7] that the tangential outlet offers higher flow rate and pressure in the same speed than the radial outlet. In the radial outlet port configuration, the fluid has to change direction in order to exit the seal chamber through the radial port leading to flow separation and subsequent flow reduction while the tangential outlet port is aligned with the fluid direction as it exits the pumping ring which means neither flow separation nor flow reduction occur in this configuration. 4. Numerical study
3. Experimental accuracy and validation method Repeatability is a term that indicates overall consistency of the results extracted from a specific instrument or experimental procedure. It is imperative that the experimental setup which was discussed in the previous section is both accurate and reliable without the need to rebuild a new laboratory setup. This can be achieved by the test-retest reliability method. The concept that user has ensured that the laboratory setup piping is cleaned up and all the conditions are the same since it was first built Warda et al. [6] including recalibrating the transmitters and the data acquisition system. The first form of repeatability test is purely instrumental. Each and every single transmitter connected to the National Instrument’s Data Acquisition system, i.e. DAQ assistant in LABVIEW software, has been put into its settings that the signal which will be received on the user’s computer has 10,000 sample readings which means that the user has ten thousand readings by the time the signal is received. Afterwards, this signal is to be averaged in another Virtual Instruments (VI) file to receive only one last reading. Furthermore, another form of repeatability test retest method is to measure a previously known and measured pumping ring performance curve and validate it by the previous measurements. The selected pumping ring is the non API single pumping scroll since this will be the pumping ring type in which the numerical investigation will take place. The previous results have been acquired as a courtesy from Warda et al. work [6]. The same barrier pressure transmitter (measurement range 0–16 bar with accuracy ±0.10%) used in the repeatability test-retest method so as to maintain all variables constant and recreate the same conditions as the previous experiments had on investigating
In the present study, a CFD model will be formulated to deduce the most accurate representation to the flow in the outboard seal in terms of the turbulence modeling equations. Each turbulence modeling equation has its own characteristics with advantageous and disadvantageous components in predicting the flow fields with adverse conditions (i.e. separation, swirling). In the current numerical study, the outboard mechanical seal with a non API single pumping scroll ring has been subjected to the k-epsilon group stated in the next section for both outlet port configurations tangential and radial outlet. The validation reference is the experimental results which were acquired from the experimental setup described in the previous section. These results are shown in Fig. 4.
Fig. 3 . Photographic view of the mechanical seal area illustrating the two pressure transmitters.
2723
0.3
Pressure difference (bar)
Pressure difference (bar)
Experimental and numerical investigation
0.25 0.2 0.15 0.1
TO 3600 rpm TO 3000 rpm TO 1500 rpm TO 1000 rpm
0.05 0
0
1
2
3
4
5
6
Barrier flow rate (L/min)
0.2 0.15 0.1 RO 3600 rpm RO 3000 rpm RO 1500 rpm RO 1000 rpm
0.05 0
0
1
2
3
4
5
6
Barrier flow rate (L/min)
Fig. 4 Graphical representation of the experimental results of the single pumping scroll performance curve tangential outlet (left) and radial outlet (right).
4.1. Governing equations and numerical methods The CFD model is based on the Reynolds–averaged Navier Stokes equations: @q þ r q! mr¼0 @t @ ! ¼ q m þ r q! m r! m þq ! x ! m ¼ rp þ r s @t 2 ¼ T m rI s ¼ l r! m r þ r! mr r! 3
ð1Þ ð2Þ ð3Þ
Eq. (1) represents the conservation of mass where qq : is the density. Equation two represents conservation of momentum which consists of pressure gradient term, two additional acceleration terms: the Coriolis accerleration (2! x ! m r ) and the ! ! ! centripetal acceleration ( x x r ) which can be represented in absolute velocity formulation by q ! x ! m and the conventional stress tensor term on the equation’s right side. Equation three represents stress tensor term identification where l is the molecular viscosity, I is the unit tensor, and the second term on the right hand side is the effect of volume dilation [8]. In modeling, the turbulence equations are classified into both time and space averaging. In the time-averaged approach, the equations are classified into both turbulent stress model (RSM) which has 5 to 7 equations to solve and Eddy viscosity approach which has variety of equations from zero equation model to two equation model either k-epsilon or k-omega. In this study the k-epsilon approach is being used. There are three forms of k-epsilon approach equation: k-epsilon standard, k-epsilon realizable and k-epsilon RNG. According to Jones and Launder [9], the standard k-e model is a semiempirical model based on model transport equations for the turbulence kinetic energy (k) and its dissipation rate (e). The model transport equation for k is derived from the exact equation, while the model transport equation for e was obtained using physical reasoning and consists of similarities to its mathematically exact counterpart. This approach has several known limitations: separation prediction, swirling flows and flow with strong streamline curvature. The RNG k-e model was derived using a rigorous statistical technique (called renormalization group theory). It is similar in form to the standard k-epsilon model, but includes an additional term in its epsilon equation that significantly improves the accuracy for rapidly strained flows as well as effect of swirl on turbulence is
included in the RNG model, enhancing accuracy for swirling flows. According to Shih et al. [10], the realizable k-epsilon model differs from the standard k-epsilon model in two ways; new equation expressing the turbulent viscosity as well as the dissipation rate, e, which has been derived from an exact equation for the transport of the mean-square vorticity fluctuation. An immediate benefit of the realizable k-epsilon model is that it more accurately predicts the spreading rate of both planar and round jets as well as providing superior performance for flows involving rotation, boundary layers under strong adverse pressure gradients, separation, and recirculation. The commercial software ANSYS FLUENT [8] is used in this numerical study employing finite volume discretization of the governing equations. Convective terms are discretized using a second-order upwind scheme [8]. The SIMPLE algorithm [11,12] is used for pressure–velocity coupling. In the present study, the scalable wall function [13] is used as well. 4.2. Pre processing In the following section, the pre-processing stage of the CFD model for the non API single pumping scroll for both tangential and radial outlets will be discussed in details. The preprocessing is the first crucial stage in creating a viable CFD numerical model and it consists with both geometry as well as meshing. One of the difficulties encountered while creating the geometry is its complexity which was in two separate volumes; tangential inlet ports as well as scroll geometry. One of the unique properties that the axial integrated pumping device has is the tangential inlet in order to adjust the flow direction towards the scroll part of the ring in order to minimize the eddy losses which may occur otherwise. The tangential inlet geometry is quite tricky when drawing since the inclination of the port is not in the XY plane only but in both XY and YZ planes as shown in Fig. 6. Tangential inlet also contains another unique characteristic which is that the inlet area in the outer gland of the outboard mechanical seal contains two inclined ports to accommodate any different designs the scroll ring may have either left hand scroll or right hand scroll which is considered to be due to John Crane’s interest in making its outer gland case for this type of pumping ring generalized to be used in more than one application with any type of scroll. As for the scroll ring geometry, its complexity lies in two different challenging aspects. First of all, the scroll geometry
2724 has unique geometrical dimensional input in order to draw the correct scroll ring geometry i.e. (helix angle, cutting tool width, axial lead as well as scroll teeth depth). All of the previous properties helps define the actual scroll ring which provided the performance curve deduced from the experimental setup in Fig. 4. The geometry has been drawn using Solid Works 2015 software as shown in Fig. 2 utilizing the actual pumping ring as a reference. Furthermore, the non– API pumping scroll radial clearance is 0.5 mm. Therefore, when drawing the fluid volume as a part of the pre-processing stage the scroll cavities as well as the 0.5 mm radial clearance is the fluid volume representing the rotating scroll ring volume which increased the complexity of the three dimensional model drawn. This is considered to be the primary reason for author’s choice to use tetrahedral mesh cell type for representing such complex volume. In meshing, the geometry’s complexity is taken into consideration as well as the necessity to represent in a valid way the change of fluid volume alongside the assembly. Soft behaviors as well as the proximity and curvature properties are being implemented in the CFD model in order for the mesher to adapt against complex geometry conditions. Figures 5 and 7 illustrate the Meshing geometry in non API pumping scroll in both tangential and radial outlet port utilizing the proximity and curvature function to deduce a high quality mesh. Table 1 states some of the statistics of the Mesh for both outlet port configurations for the fine mesh. It is noticed that due to the
H.A. Warda et al.
Fig. 6 A screenshot of 3-D model of the tangential inlet port volume using Design Modeler software in ANSYS 14.0 software.
different outlet port geometry, the mesher deduces a higher number of elements to represent the radial outlet geometry than that of the tangential outlet configuration while using the same body sizing value as an input. This data has been deduced after performing the grid independence test successfully which be described in details in the next section. As shown in Figures 5 and 7, unstructured meshing technique was used for adapt meshing and to allow mass distribution on complex geometry of the scroll part in the CFD domain. Furthermore, similar to the previous work of Warda
Fig. 5 (a) Grid generation view of the radial outlet port assembly indicating the fluid volume meshing around the pumping ring, (b) Frontal view of the assembly revealing the grid generation around the radial port, (c) sectional view indicating the grid generation around three interfaces of three different subdomains in the radial outlet assembly.
Experimental and numerical investigation
2725
Fig. 7 (a) Grid generation view of the tangential outlet port assembly indicating the fluid volume meshing around the pumping ring, (b) Grid generation view of the tangential inlet port while suppressing the rest of the assembly (c) Frontal view of the assembly revealing the grid generation around the tangential port, (d) sectional view indicating the grid generation around three interfaces of three different subdomains in the tangential outlet assembly.
Table 1 Summary of the significant statistics of the fine mesh used to represent both tangential and outlet port configuration.
rate) is added as an input and pressure outlet where the pressure output value is deduced numerically.
Outlet configuration
Tangential
Radial
4.3. Grid independence test
Body sizing Number of Nodes Number of elements Max. Aspect ratio Min. Element quality Max. Skewness
5e-004 m 2,132,625 11,176,869 10.484 0.204 0.8214
2,353,325 12,333,240 11.257 0.2035 0.8240
In the following section, the validity of the meshing grid in representing both tangential and radial outlet will be tested by using grid independence test. This test is a measure of the discretization accuracy of the numerical model. This is achieved by running the numerical model on several different grid specifications: coarse, medium, fine meshes and comparing the numerical model outcome results. Once the results are constant among the meshes, this indicates that the grid is not considered a variable factor that affects the discretization accuracy of the numerical solution. In the current study, the grid independence test is performed on both tangential and radial outlet port numerical models. The results are shown in Table 2 and 3 where the meshes results are being deduced utilizing the maximum flow-
[6] in the radial integrated pumping ring meshing representation, the multiple reference frame (MRF) technique has been used in this study to represent the zone domains of the CFD model. The rotating zone or frame has been selected for the scroll ring zone. The distinct difference in that approach is that since the non-API radial clearance from the scroll teeth to the wall of the outer gland is too small (0.5 mm), MRF approach was used in such a way that the whole zone above the ring is in fact the rotating zone while the wall outer gland is the stationary zone as well as stationary wall which will be considered as a boundary condition. The rest of the ring geometry which has no teeth in it is considered to be a rotating wall as an additional factor in the boundary condition input. The rest of the zone starting from the inlet and the outlet piping, fittings as well as ports are to be stationary frame. The approach used in the numerical study is velocity inlet where velocity (i.e. flow
Table 2 Details of grid used in grid independence test in tangential outlet numerical model. Grid
Number of nodes
Number of elements
Pressure output (bar)
Coarse Medium Fine
1,025,845 1,492,879 2,132,625
5,234,189 7,601,932 11,176,869
0.230957 0.211589 0.206957
2726
H.A. Warda et al.
Table 3 Details of grid used in grid independence test in radial outlet numerical model. Grid
Number of nodes
Number of elements
Pressure output (bar)
Coarse Medium Fine
1,123,562 1,578,462 2,353,325
5,565,754 8,024,186 12,333,240
0.084596 0.089112 0.087807
rate value of the tangential (5.2 lit/min) and radial outlet (3.8 lit/min) as an input. It is deduced that as the number of cells increases, from coarse to fine mesh, the steady state value of the pressure output almost remains constant for the same flow rate input value which indicates the fine mesh size stated in Table 1 converges on the most accurate results. Therefore, the fine mesh approach will be used in the current study to deduce numerically the best turbulence model that represents the flow field inside the outboard seal. The numerical investigation is performed on a computer with specifications of 32 GB RAM and processor IntelÒ XeonÒ Processor E5-2630 v4 with run time reaching 9–12 days till convergence. 5. Results and discussions In order to extensively test the compatibility of the turbulence model to the CFD model, three flow points were investigated.
These three points are the zero flow, mid-flow and maximum flow. Furthermore, the CFD model is to be subjected to two different rotational speeds; 1500 rpm and 3600 rpm considering that these are the common used rotational speeds used in the industrial applications. Tangential outlet configuration CFD numerical model results shown in Fig. 8 shows that k-epsilon RNG and kepsilon standard modeling results are significantly close but not accurate while the k-epsilon realizable results are relatively the least error deviation. It is noticeable as well that by decreasing the rotation speed, the numerical models are more successful in producing relatively more accurate results in relation to the experimental performance curve with the k-epsilon realizable being the most accurate turbulence model of the three k-epsilon equations as well. As for the radial outlet configuration, the numerical results in Fig. 9 reveal as well the same conclusion about the k-epsilon realizable turbulence equation being relatively the highest in the accuracy of the three equation with the remaining two being close but not accurate. The three turbulence model is more successful in the deduction the performance curve in three flow points with lower rotational speeds. It is noticeable that radial outlet CFD model has relatively higher error deviation from the experimental performance curve for the same turbulence model equation that the tangential outlet numerical model which was the same deduction in previous work of Warda [6] for the radial integral pumping ring.
Fig. 8 Graphical representation of the tangential outlet performance curve against all three k-epsilon equations in speeds 3600 rpm (left) and 1500 rpm (right).
Fig. 9 Graphical representation of the radial outlet performance curve against all three k-epsilon equations in speeds 3600 rpm (left) and 1500 rpm (right).
Experimental and numerical investigation
Fig. 10
2727
Pressure distribution on the scroll ring wall.
Fig. 11 (a) radial pressure distribution in the non API clearance around the pumping ring, (b) pressure distribution in one of the helical scrolls in the pumping ring, (c) Frontal view of the axial pumping ring exit indicating the barrier fluid’s maximum pressure value on each scroll.
Utilizing CFD-Post software, the numerical investigation is extended to comprehend the pressure distribution along the single axial pumping scroll assembly. One of the main observations on the post processing results, the symmetrical replication of all the flow field properties as it is considered nearly the same for each tooth alongside the circumference of the
ring. Fig. 10 is considered as one of the crucial pressure contours representations to help understand the flow field inside the outer seal. It is noticeable that the pressure is almost constant outside the scroll teeth region on the scroll ring except for certain low pressure spots alongside the circumference of the ring. The number of these spots matches exactly the number
2728 of scroll teeth in the circumference which indicates that due to the fluid flow inside the scroll teeth which rotates with the assembly acting as a rotor pump, these spots are considered to be suction region for each of the scroll cavity. These spots have a concentric nature in pressure distribution decreasing gradually as shown in Fig. 10 reaching its minimum value in its center. As the flow field advances toward the exit side of the pumping scroll, the pressure varies in an exponential increase reaching its maximum at the tip of the teeth midway the scroll lead. The maximum pressure area increases alongside the tooth until it overwhelms the whole side of the tooth at the end of each tooth. As predicted from the results of Fig. 10, the barrier fluid pressure increases exponentially as fluid progresses towards the exit port. This is revealed in Fig. 11a which shows the pressure contour distribution axially as the fluid moves along the ring assembly as well as in side view position revealing the circumferential pressure distribution of the barrier fluid on the ring as it exits. Fig. 11b also illustrates the pressure contour plane captured in the mid area of one of the scroll cavity which indicates an increase of the pressure value as it progresses through the cavity reaching its maximum at the exit end which is shown in Fig. 11c. It is noticed that at the beginning of the scroll’s helical tooth, the radial plane in CFD-Post captures the pressure of its minimum value which matches the previous conclusion in Fig. 10 being the lowest possible pressure value of the assembly. As the fluid progresses axially alongside the radial plane, the barrier fluid pressure value increases normally until it reaches its maximum value on the exit side. It is also noticeable that the pressure at the topside of the radial clearance is higher than the pressure near the wall which is due to the centrifugal pressure building up from the pumping scroll rotation. Furthermore, when observing the side view indication, the pressure distribution of all the exit barrier fluid pressure in each of the scroll cavities in the exit, few conclusions are made: First of all, the pressure in each of scroll cavity is relatively the highest on the left wall side and decreases slightly when progressing to the cavity’s right wall. This occurs due to the stagnation pressure of the barrier fluid as the ring rotates in a clockwise direction. Last but not least, when observing each of the pressure values of each of the scroll cavities, it is observed that the pressure distribution profile is symmetrical on each of the scroll tooth along the circumference. 6. Conclusions The experimental results vary according to the shaft speed as well as the outlet port configuration. The experimental performance curve results have been used as a validation method while undergoing the numerical investigation for both tangential and radial outlet model. Moreover, the numerical results have shown acceptable performance curve using the kepsilon realizable model which has higher accuracy in repre-
H.A. Warda et al. senting the experimental performance curve than the rest of the k-epsilon family. Finally, the post processing data reveals that the barrier fluid pressure increase as it progresses through the scroll teeth until it reaches the maximum at the ring exit. Acknowledgments The authors would like to acknowledge E.A.Selim and M.W Gamal Aldin for their assistance and support in the experimental and the numerical investigations respectively and John crane UK for providing the mechanical seal assembly. References [1] API, Pumps — Shaft Sealing Systems for Centrifugal and Rotary Pumps ANSI / API Standard 682 ISO 21049: 2004, Pumps — Shaft Sealing, 2005. [2] R. Clark, H. Azibert, visualizing fluid flow and heat transfer in rotating shaft seals., In: 15th Int. Conf. Fluid Seal., BHR group conference series publication, 1997: pp. 353–378. [3] C. Carmody, A. Roddis, J. Amaral Teixeira, D. Schurch, Integral pumping devices that improve mechanical seal longevity., In: 19th Int. Conf. Fluid Seal., BHR group conference series publication, UK, 2007: pp. 235–247. [4] C. Carmody, A. Roddis, Saving Energy, Saving Water and Saving the Planet through the use of Affordable, Premium BiDirectional Pumping Rings, In: Des. Optim. through Value Eng. Fluid Mach., IMECHE, 2008: pp. 1–15. [5] R. Smith, Contradicting requirements in API 682 dual seal design configurations, Rotherham, England, 2008. http://www. aesseal.com/en/resources/whitepapers/contradictingrequirements-api-682-dual-seal-design-configurations [6] H.A. Warda, E.M. Wahba, E.A. Selim, Integral pumping devices for dual mechanical seals: experiments and numerical simulations, J. Eng. Gas Turbines Power. 137 (2014) 22504, https://doi.org/10.1115/1.4028384. [7] H. A.Warda, I.G. Adam, A.B. Rashad, M.W. Gamal Aldin, Effect of Kinematic Viscosity of Barrier Fluids on The Performance of a Bi-Directional Integrated Pumping Ring for Dual Mechanical Seals, In: ASME Proc. | 17th Symp. Turbomach. Flow Simul. Optim., Washington, DC, USA, 2016: pp. 1–36. http://doi.org/10.1115/FEDSM2016-7763 [8] ANSYS, ANSYS FLUENT Theory Guide, Release 14.0., ANSYS Inc., Canonsburg, PA, 2011. [9] W.P. Jones, B.E. Launder, The prediction of laminarization with a two-equation model of turbulence, Int. J. Heat Mass Transf. 15 (1972) 301–314. [10] T.-H. Shih, W.W. Liou, A. Shabbir, J. Zhu, A new k-e eddyviscosity model for high reynolds number turbulent flows model development and validation, Computer. Fluids 24 (1995) 227–238. [11] S. Patankar, Numerical Heat Transfer and Fluid Flow, first ed., Hemisphere, Washington, DC, USA, 1980. [12] Z. Luan, M.M. Khonsari, Numerical simulations of the flow field around the rings of mechanical seals, J. Tribol. 128 (2006) 559–565, https://doi.org/10.1115/1.2197845. [13] M. Chmielewski, M. Gieras, Three-zonal wall function for k-e turbulence models, Cmst 19 (2013) 107–114, https://doi.org/ 10.12921/cmst.2013.19.02.107-114.