Encyclopaedia Index
 Task Completion Report and Validation Report of   SCONES – Power-station condensers MICA deliverable D7.6

### Contents

1 Introduction to the industrial problem(s)

The heat exchanger application considers a condenser of a Swedish nuclear power plant. A picture of one half of the condenser is shown in Figure 1. In the condenser, steam from the low-pressure turbines is condensed and heat is exchanged to a flow rate of 20 m³/s of sea water. The total heat exchanged amounts to approximately 1000 MW.

Figure 1. The condenser model.

One purpose for making CFD calculations of a condenser is to find measures to lower the pressure at the inlet to the condenser and thereby increase the power output from the low-pressure turbines.

A serious problem in the condenser is that water droplets collide with the tubes and cause fatigue on the tube material. CFD calculations of droplet motion in the condenser can explain why damages are more severe in certain parts of the tube bundles. It can also give ideas on geometrical modifications of the condenser that can stop high-velocity droplets from hitting the tubes.

3D calculations for the condenser shown in Figure 1, including droplet motion, were previously done with PHOENICS (Jal E.N. & Tinoco H. "Mathematical modelling of industrial power condensers with an application to droplet erosion". J. Mech. Eng. C461/029 1993).

2 Description of the simulated case(s)

The size of the condenser is typically 10x20x10m. A mixture of steam, water and non-condensable gases coming from three low-pressure turbines enters the condenser from the top. The total mass flow rate is around 400 kg/s, the volume fraction of water being around 0.003%. This gives an average inlet velocity of around 90 m/s for an inlet cooling water temperature of 5C. The mean pressure throughout the condenser is around 0.045 bar and the tem perature around 35C for the same cooling water temperature.

A cooling water flow of 20 m³/s flows through approximately 25000 horizontal and straight tubes. The tubes are arranged in eight tube bundles in a so-called "church window" design. Each tube bundle is shown as a light-grey object in Figure 1. The steam is condensed at the surfaces of the tubes and falls down through the tube bundles and is finally collected in a water tank at the bottom of the condenser. Non-condensable gases are sucked out of the condenser through an exhaust system insid e the tube bundles.

A complete CFD model of a condenser requires sub-models that are seldom found in commercial CFD codes. These models are:

• a transport equation for cooling water temperature, which is linked to the steam model only by heat exchange through the tube material.
• empirical models for momentum loss (including two-phase effects) and turbulent viscosity in the tube region.
• a model for condensation mass flux at the tubes, including inundation effects.
• a model for heat transfer between steam and cooling water.

3 Preprocessing of the case

3.1 Availability of all the essential objects. Limitations

The pressure at the outlet from the exhaust system has to be changed manually until a zero mass flow of steam is achieved at the outlet. An automatic handling of the outlet pressure condition would be beneficial.

The original plan was to be able to study droplet motion and the effects that a non-uniform inlet velocity with swirl and the k-e model of turbulence has on the droplet trajectories. Because of difficulties in transferring the condenser model to the SCONES model resources were insufficient to include these features in the model.

3.2 Availability of all the essential mathematical models. Limitations

All models required for calculation of momentum loss, condensation and heat exchange in the tube bundles are included in the SCONES software.

Langrangian equation for calculation of isothermal droplet motion, including an expression for the drag coefficient, has not been implemented.

3.3 Speed of the VR editor

For the applications studied, the lack of speed with which the model can be moved is not a major annoyance. For other applications, using more advanced clip-art, slow graphics could be a problem.

3.4 Ease of use of the VR editor

When you enter the PHOENICS Commander and create a new case you do not get the opportunity to immediately specify project name and case name. When you exit the VR-EDITOR you get the question "do you want to save data", but you do not get a question if you want to save the case to give it a project name and a case name. You have to do that by choosing save in the file menu. It is not until you start a new case that you get a question if you want to save the old case. By then you might have forgot if you should store the old case or not.

"Scale" should change places with "probe position". As it is now "scale" is associated with "probe position" instead of "domain size". "Snap size" should be changed to "probe size".

"X-dimension" should be changed to "Number of cells in X-direction".

"Rinner" is not understood by non-PHOENICS users. It should be changed to something like "inner diameter"

"alfa", "beta" and "theta" should be changed to "positive rotation around x-, y- or z-axis".

It should be possible to store views.

3.5 Visual appearance of the VR world

Sometimes objects which should be visible are obscured by other objects.

The grey colour surrounding the icons is too dark.

When you have minimized the VR-Editor window and then maximize it again it appears as almost completely blue. The VR screen can be regained with a mouse click but you have to be careful where to click, you could for example accidentally create a new object. This problem is the same for VR-Viewer.

3.6 Need to supplement VR settings with PIL

Relaxation and other solution controls must be changed with PIL. However, as solution control finally will be automatic, user control of solution is not required.

3.7 Errors encountered

POLIS and View RESULT can not be started from the Phoenics Commander.

4 Submission of the case, remote calculation and data retrieval

Hardware

The MICA software has been installed and run on a HP Vectra VL5 P-200 MMX with 64 Mb of RAM running on Windows NT, version 4.0.

The network

The local network is a 10 Mb Ethernet network. TCP/IP is used for data transfer. A firewall from SUN is used, which is configured to work on an IP level. The maximum possible data transfer capacity is 830 Kbits/s.

Job submission.

Most job submissions were successful. Failures to submit jobs have been experienced for the following reasons:

• "unable to create file user space." Possibly caused by lack of disk space on the remote machine. A run is started anyway, producing erroneous results.
• "no contact centre available".

The estimated run time presented when transmitting a job has always been underpredicting

The ability to choose centre through using "Preferred centre" has worked without problems.

Job status

It has been possible to delete and interrupt jobs.

When a contact centre is not available you do not get the message that it is not. You just get no information of the jobs running on that centre. You could get the impression that there are no jobs running.

It has not been possible to read account status.

Retrieval of results

Problems were initially experienced with receiving large result files (around 2Mb) during busy traffic conditions on the Internet. The error message given was "server did not answer". This problem was later resolved.

Retrieval times for result files of 2 Mb was usually around 2 minutes.

Zipped result files attached to emails can be translated. Zipped result files included in emails can not be translated.

5 Post-processing of the case

5.1 Availability of all the essential tools for dataset interrogation. Limitations

It would be useful to be able to determine the average values of variables in regions that the user can specify in VR Viewer.

5.2 Speed of the VR viewer

For the applications studied, the lack of speed with which the model can be moved is not a major annoyance. For other applications and if higher resolution is allowed for contours, slow graphics can be a problem.

5.3 Ease of use of the VR viewer

It would be useful to be able to store series of commands, which would make it easier to repeat commands.

It should be possible to change the scale and number of vectors in the vector and to explicitly set the value of the variable for an isosurface.

5.4 Visual appearance of the results in VR viewer

It should be possible to improve the resolution for contour plots and the smoothness of isosurfaces.

5.5 Errors encountered

6 The CFD simulation

6.1 Analysis of results

The grid used for the calculations is a 3D grid consisting of 33280 cells, 20 cells in the z-direction (along the tubes), 52 cells in the x-direction (in the other horisontal direction) and 32 cells in the y-direction (vertical direction).

The CPU time required per iteration is 36 seconds on the PC specified in Chapter 4. The convergence is quite rapid for about 150 iterations, after which convergence is very slow. Satisfactory convergence, based on the average pressure at the steam inlet, is reached after about 1500 iterations, or 15 CPU hours. The size of the phi file produced is 14 Mb (zip-file 1.9 Mb).

The first case calculated was for a cooling water temperature of 5C and a cooling water velocity of 2.014 m/s. Vectors on a vertical plane perpendicular to the axis of the condensor tubes and in the middle of the condenser is shown in Figure 2. One can see that the flow accelerates towards the tube bundles. The maximum velocity calculated is 243 m/s in the region between the tube bundles. The velocities near the top of the tube bundles are higher than the critical velocity for droplet ero sion.

Figure 2. Vectors at a plane in the middle of the condenser.

The pressure loss in the condenser, from the steam inlet to the outlet is 1022 Pa. The difference between the minimum and maximum pressure at the steam inlet plane is around 200 Pa, with the highest pressure above the sloping wall.

Figure 3 shows the condensation rate (kg/s) on a vertical plane through the middle of the first tube bundle. The highest condensation rates are blue. One can see that the condensation rate is higher in the colder part of the condenser (on the side where the cooling water enters). A consequence of this is a net flow of steam towards the colder part of the condenser.

Figure 4 shows the cooling water temperature at the same plane. One can observe the cooling water temperature increasing on its way from the inlet to the outlet. The average cooling water temperature at the outlet is 15.8C or 10.8C higher than that at the inlet.

Figure 3. Condensation rate on a vertical plane in the middle of the first tube bundle.

Figure 4. Cooling water temperature in the middle of the first tube bundle.

6.2 Validation data

For validation purposes, pressure and temperature have been measured in the condenser. Pressure has been measured at two positions and temperature has been measured at one position. All measurement points are positioned at a short distance above the tube bundles.

6.3 Validation of results

Figure 5 shows a comparison between the measured and calculated pressures as a function of inlet cooling water temperature. The calculated values are plotted with green symbols, the measured with black. The calculations give very close agreement with measurements.

The comparison between the measured and calculated temperatures is shown in Figure 6. The calculated temperatures are in satisfactory agreement with measured ones.

Figure 5. Comparison between measured and calculated pressures.

Figure 6. Comparison between measured and calculated temperatures.

6.4 Parametric analysis

A computation was performed with a refined grid to check grid independence. The refined grid had 87472 cells, or 2.6 times as many cells as for the grid used for the calculation presented in chapter 6.1. The number of cells in each direction was increased with about 40% to 71, 44 and 28 cells.

The pressure at the inlet increased from 5522 Pa for the coarse grid to 5552 Pa for the finer grid, an increase of 2.9% relative to the pressure drop of 1022 Pa in the condenser. The cooling water temperature rise was increased from 10.77C to 10.81C (+0.4%). These changes are not considered to be big. The coarse grid was therefore used for the parameter tests presented below.

The following variations in operating conditions were studied:

• Cooling water temperature (1C, 11C and 16C).
• Cooling water flow rate (+10% and +20%).

The variations in cooling water temperature reflect the yearly variations in cooling water temperature. These calculations were done to verify the ability of the model to predict the flow field for a wide range of cooling water temperatures. The comparison between calculated and measured pressure and temperature presented in Figure 5 and 6 shows that this is the case.

By increasing the cooling water flow rate one can get a lower temperature rise of the cooling water and thereby get a lower pressure in the condenser. With a 10% increase in flow rate the pressure at the condenser inlet decreased with 0.8%. With a 20% increase in flow rate the pressure at the condenser inlet decreased with 1.6%.

The table below lists all computations for which results are presented in chapter 6.

 CASE Tcool (C) ucool (m/s) Number of cells 1 5 2.014 33280 2 1 2.014 33280 3 11 2.014 33280 4 5 2.417 33280 5 5 2.215 33280 6 16 2.014 33280 7 5 2.014 87472

Tcool = Cooling water temperature (C)

ucool = Cooling water velocity (m/s)

7 MICANET in the context of an industrial environment

It is believed that the MICA concept can be beneficial, especially for smaller companies that have a intermittent need for CFD and who therefore do not find it worth the cost to invest in advanced hardware, a locally installed CFD code and CFD knowledge. If remote execution times will become much shorter than what can be achieved at local workstations the concept can be useful also for bigger companies that use CFD more regularly.

8 Final recommendations

To make the product more attractive, more work has to be done to increase user-friendliness, especially concerning the visual appearance of VR-Editor and VR-Viewer, reliable access to the remote machines, automatic grid refinement and automatic setting of parameters that control convergence.