53_Krys Bangert.pdf

8/10/2019 53_Krys Bangert.pdf

1/16

Mini-project report

Radiator heat transfer augmentation bychanges to wall surface roughness and

emissivity

Mr Krys [email protected]

June August 2010


2/16

Radiator heat transfer augmentation by changes to wall roughness and emissivity

1) Abstract:

This mini-project is a part of ongoing work carried out at Sheffield University by Dr Stephen Beck andmultiple PhD and MEng students looking into the effects of combined radiation and convection on householdradiators. This study looks at how changes to a walls surface finish can affect a radiators heat transfer. A

series of tests and computer simulations were run comparing the heat loss from a radiator when the wallroughness and emissivity directly behind it was changed. To gather the empirical data needed for the studyan existing radiator test rig built to approximate European Standard EN 442-2 was used. Thermocoupleswere mounted on the radiator, wall, inlet/outlet pipes and in the air gap behind the radiator, to gathertemperature readings for levels of conduction, radiation and convection in the system. Three tests werecarried out under steady state conditions, with a 10C drop across the radiator to comply with the BritishStandard. The first test used a plain wall as a control, the second and third tests had sandpaper sheetsmatching the profile of a radiator attached to the wall behind the radiator. The sheets were sprayed in glossblack and silver paint respectively, to modify the surface emissivity. A computer simulation of the setup andtests was also created for comparison using the CFD (computational fluid dynamics) program Fluent.Unfortunately the results from the tests and simulations were inconclusive due to high levels of experimentalerror and convergence issues in the CFD model. However, the data did show that the overall trendsdiscovered in previous work relating to emissivity are valid. It is hoped that in future work more accurateresults can be obtained based on the recommendations of this study.

Keywords: E-Futures, CFD, Heating, Radiator, Convection, Radiation, Heat transfer.

2) Introduction:

Domestic energy consumption is one of the key areas currently been looked at by environmentalresearchers. It has been estimated that the domestic sector accounts for over 30% of the current UK nationalenergy demand, the second largest usage behind the transport sector

1.

If the greenhouse gas emission targets set by the UK government are going to be met in the comingdecades, the way we use energy in the home is going to have to become more efficient. One of the bestways to achieve this, in the short to medium term is to upgrade the existing housing infrastructure to make itas energy efficient as possible. This will help to offset the transition as new zero carbon housing graduallyreplaces the ageing housing stock.

The biggest domestic efficiency gains are to be made in space and water heating, which account for 58%

and 24% of the energy use respectively2. Since the 1970s the rates of overall domestic heat loss have fallenprogressively due to improvements made in the levels of insulation

3. However, the single largest contributor

to this heat loss; the heat lost through walls, is still a common problem today3.The focus of this mini-project is to continue the work carried out in Sheffield Universitys department ofmechanical engineering, looking into new methods of increasing household thermal efficiency. This studylooks into the effect that wall surface roughness and emissivity have on domestic radiator heat output. Thiswork will be building on the findings gathered by former MEng student S.G. Blakey and current PhD studentA.K.A Shati. In their previously published work

4-6, it was found that if surfaces with a higher emissivity were

placed behind a radiator at an optimal distance (s/L = 1/12. Where s is the separation and L is the verticalheight of the radiator), a larger heat output could be obtained from the radiator

4. It was then subsequently

discovered that if large scale roughness was also added to the surface finish, the heat flow could beincreased even more (upto 26% using a high emissivity saw-tooth surface)6. With further analysis it wasfound that these effects were caused by the wall heating up due to its higher emissivity (more infraredradiation was absorbed than a plain/reflective wall). This heating in turn created a convecting surface behind

the radiator which increased the air flow. The addition of roughness to the wall further amplified this effect byincreasing the surface area for heat transfer and creating more turbulence, which improved the masstransfer and heat flux in the air gap behind the radiator.

This mini-project will look into how the heat transfer is affected by using a different geometry on the wallsurface with dissimilar emissivities. The experiments and simulations will be carried out in a similar fashion tothe previous studies, but the surface finish will be based upon coarse grain sandpaper with different emissivecoatings.


3/16

3) Experiments/methodology

Test setup:

To perform the experiments an existing domestic central heating test rig from the previous studies wasreused (see appendix 1.0-1.1). This setup was designed and built to approximate European standard EN

442-2

7

, this enabled pervious test results to be compared with the performance figures for commerciallyproduced radiators.

The setup consisted of the following apparatus:

Amount: Description:1 Standard 600mm high by 600mm wide single plate radiator with 50mm expanding foam

insulation covering one side1 3kW immersion heater regulated by a PID temperature controller (accuracy: 0.1C).1 Hot water cylinder with insulation.1 Water cistern.4 Control valves to regulate water flow: two control valves on the radiator, one bypass valve to

the water tank and one output valve.1 Class F mains powered pump operating at 60W 2000 RPM.

1 Rotormeter to measure the water flow rate.1 Pico technology TC-08 Serial thermocouple interface.1 Pico technology TC-08 USB thermocouple interface.10 Type T thermocouples (accuracy: 1C)6 Type K thermocouples (accuracy: 1C)1 Digital thermistor type anemometer (accuracy: 0.3C and 0.015m/s)- 15mm copper piping with foam lagging (10mm thick).- 20mm thick High density particle board.

For each test the radiator was mounted 150mm above the floor and 50mm away from the wall, to match theprevious test setups. These measurements were used again because it was discovered that the maximumvalues of heat transfer occur at this distance away from the wall in Blakelys initial study

4. The thermocouples

were arranged on the radiator, front and back of the wall, inlet and outlet pipe and in the air gap behind theradiator (see appendix 1.2-1.3). A thermocouple was also used to take the ambient air temperature readings

and the thermocouple linked to the PID controller was attached to the radiator inlet pipe for temperaturefeedback; it was assumed that the external pipe temperature was the same as the water. The rest of thethermocouples were linked to a PC for data logging using the PicoLog software, the sample rate set to a 1minute interval for each interface. The anemometer was also linked to a computer for data logging using themanufacturers software. The sensor was mounted using a clamp and stand to give readings at the top of theradiator in the middle of the air gap (see appendix 1.1).

To comply with the European standards the temperature gradients across the radiator must have a 10Cdrop and be maintained with 0.1C. To calibrate the setup the heater was set to 80C using the PID and leftover night to achieve steady state conditions. The valves were then adjusted to get a temperature droprequired. However, it was subsequently found that this could not be achieved without stopping the water flowcompletely. So a compromise value of approximately 8C was used for the tests with a flow rate of 0.2 L/min.

Each test surface was constructed of 9 sheets of 40 grit coarse grain sandpaper held together using fabric

tape. The surface was then cut to a 600mm x 600mm size to match the profile of the radiator. One of thesurfaces was then spray coated with black gloss multi-surface enamel paint and the other with silver radiatorand appliance spray paint (see appendix 1.4). The surfaces were attached to the test wall using fabric tapefor each of the tests. Holes were made for the air temperature thermocouples to protrude and the wallthermocouple was attached to the surface using a small piece of fabric tape sprayed in the same finish asthe relevant test surface.


4/16

Test method:

Before the tests were carried out the PID controller was set to 80C and the system was left for a minimumof 6 hours to reach steady state conditions with an appropriate temperature drop. Once this was achievedeach test was run for a period of one hour with the thermocouple and anemometer readings takenautomatically by the appropriate software and the initial rotometer flow reading taken manually (previoustests had shown it to be consistent in steady state conditions). The first test was run with a plain wall, the

second with the black surface and the third with the silver wall. The test data was then collected from thesoftware and complied in Microsoft Excel for analysis (see digital appendix).

CFD modelling:

To find the full effects of the heat transfer and air flow a series of CFD simulations were carried out usingsimilar geometry to the physical test setup. The software used for this was Gambit version 2.2.30 and Fluentversion 6.3.26.

A 2D mesh was created in Gambit which featured a 4 by 3 meter room with 0.1m walls (to comply with theEuropean standard), a 0.6m by 0.02m radiator, 0.050 by 0.6m of thermal insulation (which was attached tothe radiator) and a 0.6m separate surface attached to the left hand wall (see appendix 1.5). A quadrilateralmesh was used with the highest density elements around the radiator. The model contained 22058 cells,45440 faces, 23370 nodes and 1 partition.

The 2D mesh was imported into fluent and the following properties and boundary conditions were applied:

Entity: Property:Room wall material Brick Density: 1850 (kg/m

3), Cp: 800 (j/kg-k), Thermal conductivity: 0.595 (w/m-k),

Absorption coefficient: 0.9 (1/m), Scattering coefficient: 0 (1/m), Refractive index: 1.

Room fluid Air Density (Boussinesq): 1.225 (kg/m3), Cp: 1006.43 (j/kg-k), Thermal conductivity: 0.0242

(w/m-k), Viscosity (kg/m-s): 1.7894x10-5

, Absorption coefficient: 0 (1/m), Scatteringcoefficient: 0 (1/m), Thermal expansion coefficient (1/k): 0.00343, Refractive index: 1.

Radiator insulation material Foam Density: 320 (kg/m3), Cp: 1455 (j/kg-k), Thermal conductivity: 0.0485 (w/m-k),


Radiator wall material Steel Density: 8030 (kg/m3), Cp: 502.48 (j/kg-k), Thermal conductivity: 16.27 (w/m-k),


Rough surface material Test surface (black sandpaper) Density: 0.69 (kg/m3), Cp: 830 (j/kg-k), Thermal

conductivity: 0.2 (w/m-k), Absorption coefficient: 0.97 (1/m), Scattering coefficient: 0 (1/m),Refractive index: 1.Test surface (Silver sandpaper) Density: 0.69 (kg/m

3), Cp: 830 (j/kg-k), Thermal

conductivity: 0.2 (w/m-k), Absorption coefficient: 0.47 (1/m), Scattering coefficient: 0 (1/m),Refractive index: 1.Wall Roughness (Both surfaces) - Roughness height: 0.000425(m), Roughness constant:0.5.

Radiator heat source Fixed value - Temperature: 353 (k) [80C]

External wall boundary Temperature: 293 (k) [20C], Wall thickness: 0(m), Heat generation rate: 0 (w/m ).

Five simulations were run emulating the physical tests. The first with ordinary plain wall, the second and thirdwith high emissivity rough and smooth surfaces and the fourth and fifth with a low emissivity rough andsmooth surfaces (see table above). The emissivity of all the walls except the test surface and radiator wasset to 0.93, with the exception of in the plain wall test in which the left wall and the test surface had the sameemissivity. The two tests with rough surfaces used the standard wall function with additional wall roughnessattributes added to the simulation. The other simulations used the enhanced near wall functions. Due to thetest surfaces have a relatively uniform roughness, it was assumed that the roughness height correspondedto the average sand particle diameter (P40 grit sandpaper = 425 Average particle diameter (m) = 0.000425

m). The roughness constant was left at the default value of 0.5, which is taken from empirical resistance datafor pipes roughened with tightly-packed uniform sand-grains (ref: Fluent 6.3 Users Guide). The standard k- turbulent with full buoyancy and thermal effects (when using enhanced wall function) was used; along withthe Discrete transfer (DTRM) radiation models. The gravity was set at -9.81(m/s2), an operating temperatureat 303 (k) with 1 atmosphere pressure. The rough and smooth high and low emissivity surfaces weremodelled for comparison purposes due to uncertainty with the simulation of near wall effects.


5/16

4) Results/Discussion:

To calculate the heat transfer of the radiator the thermodynamics of the system needed to be analysed. Inorder for the first law of thermodynamics to be obeyed, the energy input to the system must equal the output.Because the radiator itself does not perform work, this energy (in the form of heat) must come from the waterpassing through it. As the radiator heats up due to the conduction of the water, the heat is then transferred tothe air by the process of convection and radiation. This causes the water flowing through the radiator to lose

heat energy

5

. Because the water is f lowing within a closed loop system, if its temperature is measured on theinput and output of the radiator, the net heat loss demonstrates the heat transfer into the surroundings.However, this calculation is only applicable under steady state conditions within the system (hence the setuptests mentioned in the previous section).

To calculate what proportion of the heat loss is down to convection, conduction and radiation, the energyflows in and out of the system must be considered (see appendix 1.6). The heat flows can be grouped intothe following main categories; convection into the air, radiation to the room and wall (with some reflectionback into the radiator) and conduction through the wall. The amount radiated into the room is so small it wasdisregarded in this study. Taking these flows into consideration the following equations have be derived:

Heat transferred from the wall surface to the air:

Convection heat transfer from the radiator to the air:

Total heat transfer to the air: By combining equations 1 and 2 into equation 3 the total heat transfer to the air can be calculated. This isdone using the equation for water heat flow (eq:5) and wall conductive heat transfer (eq:6):

Where and

These calculations were performed on the averaged data from all three of the tests and compiled in aMicrosoft excel spreadsheet (see digital appendix 1). The experimental error due to equipment data loggingaccuracy (see table 1) was not incorporated into the calculations.

Test data analysis:

The graphs showing the various heat transfers from the radiator and through the wall are included inappendix 1.9 and the digital appendix. It can be seen that the black sandpaper wall has the highest heatoutput to air (86.5w), followed by the plain surface (84.6w) and the silver surface with a much lower output(37w). This result seems to validate the prediction that a black surface should aid convection and producethe largest heat output. However, when the heat transfer by conduction and total heat transfer are alsocompared, some major abnormalities are present. It was predicted that the silver wall covering have the leastheat loss via conduction (due to its lower emissivity), followed by the plain wall and then the black. But theresults demonstrate the exact opposite, with Silver, Plain and Black having 67.5, 43.5 and 37.3 Wattsrespectively. The radiator air temperature profiles also show a different correlation of results, aftercompensating for ambient temperature. The results show that both rough surfaces give higher outputs than

just the plain wall alone, but the silver produces the highest temperature overall. This result as also opposed

by the total heat transfer reading which shows that during the silver sandpaper experiment the radiatorproduced a lower heat output compared to the others.

It is difficult to see a clear trend with the results because the data sets appear to be contradictory, but after acareful review of the data and inspection of the experimental apparatus there is a possible explanation.Following the silver sandpaper experiment it was discovered that the thermocouple at the back of the wallhad become unstuck from the wall surface. This caused an ambient temperature reading which made theheat loss through the wall calculations artificially high for the silver sandpaper experiment.It has also been shown that the temperature drop across the radiator was not uniform for each experiment,especially the silver again. The plain, black and silver experiments had drops of 8.8, 8.5 and 7.2C


6/16

respectively). These main factors along with other smaller experimental errors (discussed later the in thereport) are likely to have produced such an unexpected set of correlations.

CFD data analysis:

Each of the CFD models converged with all residuals to the power of x10 -3 (see digital appendix). The twosimulations with roughness added to the test surface however did show minor oscillations still present in the

turbulence residuals. Each of the simulations shows a heat flux error in the region of -25 watts, whichindicates the models solutions have a reasonably large error in the predictions (approx 6%). Each of themodels also shows similar velocity and thermal flow characteristics (see appendix 1.7-1.8) with aconvectional current present around the walls of the room and radiation causing the heating of the wallbehind the radiator as expected. However, there is a slightly anomaly in the convection current because itappears to split into two flows on the floor of the model rather than forming a single enclosed flow circuit asexpected. Many different meshes were tried to see if this artefact was grid dependant, but each time it wasfound consistently. The source of this abnormality was not found, but the results varied by a small enoughdegree to be classed as been converged for the purposes of the mini project.

The total energy output of the radiator was shown to be highest for the black smooth wall, followed by theplain wall, rough black wall, plain silver wall and the rough silver wall (see appendix 2.0). This trend seems tosupport the hypothesis that a higher emissivity surface increases the radiator output, but it does not showthat the turbulence generated from the rough finish helps. In fact, the roughness seems to reduce the outputof the radiator (especially in the case of the black surface), making it less efficient than the plain wall alone.However, this result could be artificially low due to problems in convergence with the two rough models and itis also uncertain whether the near wall boundary layers are properly represented using this technique.Further simulations would need to be carried to validate these results, but due to time constraintsunfortunately this was outside the scope of this project.

Comparison:

The results from the experiment and model were compared in another spreadsheet (see digital appendix).The total power output of the radiator and the thermocouple readings for the top, middle and bottom of theair gap behind the radiator were compared with the CFD analysis (see appendix 2.0). There is a largedifference between the power outputs from the model and the experimental data. It is not known what hascaused this discrepancy; it could be due the oversimplification of the model, unconvergence, experimentalerror and/or many other possible factors. The trends do not match between the two data sets for the reasonsmentioned previously. The air gap temperature readings show a good correlation for each of the wallsurfaces, will most of the readings within the two error bars. The CFD readings again support the wallemissivity hypothesis and not the roughness, but the difference in temperature increase is very small (lessthan 1C between the plain and black walls). The biggest difference in results between the experimental andsimulations is the radiator bottom reading. The simulation trend shows a temperature similar to the radiatormiddle value but the experimental shows a much more linear drop off. This difference could be due to theradiator mounting bracket and weight (see appendix 1.1) disrupting airflow under the radiator.

5) Conclusions and further work.

The findings of this mini-project are on the whole inconclusive, with the exception of some simulation resultsshowing that higher emissivity coatings behind a radiator can improve its heat transfer. There have beenmany avoidable experimental errors made and problems with the method highlighted during the course ofthe study. If these issues can be addressed by the next group of students following on from this work,significant progress can be made. In particular I think the experimental test setup can be improved by

changing the test surface mounting mechanism, securing thermocouples differently, using a more accuratedata logger and altering the radiator mounting from to reduce air drag. Other forms of data logging such asusing thermal imaging cameras and surface velocity tests could also be beneficial, along with making thesetup closer to British standard and making the test samples more uniform. Many other improvements couldbe made to the test procedure by eliminating further experimental error (see below). In addition to thephysical testing the CFD simulations also need to be improved in accuracy. The roughness effects need tobe looked into and possibly new equations developed, the mesh also needs to be properly tested to see itsaffect on the simulation (invariance testing), a 3d model could potential produce more accurate results. Ifsome of these issues can be addressed in the future work, more useful results can be obtained giving a


7/16

valuable insight into how radiator and wall design can be improved to make domestic heating more energyefficient in the future.

Further sources of experimental error:Thermocouple error +- 5C, timing between data sets +- 10 secs, flow readings too variable, Existingthermocouple layout not central, CFD materials based on estimates, Lagging not continuous on apparatuspipe work, temperature taken at surface of pipe not in fluid, backing samples non uniform in surface finish,

alignment of backing samples not correct, air flow from ventilation system affected air flow readings,thermocouple layout not aligned properly with drilled holes, different to computer model, sand paper backingnot fully flush with wall, mounting tape emissivity and test surface is different material.Acknowledgment:

Work reported here is supported by the Department of Mechanical Engineering. I would like to thank mysupervisor, Dr Stephen Beck for putting this mini-project together and helping to guide me along the way.PhD students Abdulmaged Shati and Richard Collins; for their help explaining various processes and usingthe Fluent and Gambit software. My friend and fellow DTC classmate Robert Richards, for his help andcollaboration in the initial stages for the project. Technician Malcolm Nettleship for his help setting up the testrig and Fluent Europe of the use of their software.

References:

1 Change, D. o. E. a. C. (2010).2 Change., D. o. E. a. C. (2010).3 Utley, J. I. & Shorrock, L. D. (BRE Housing, 2008).4 Beck, S. B. M., Blakey, S. G. & Chung, M. C. The effect of wall emissivity on radiator heat output.

Building Services Engineering Research and Technology 22, 185-194,doi:10.1191/014362401701524217 (2001).

5 Beck, S. M. B., Grinsted, S. C., Blakey, S. G. & Worden, K. A novel design for panel radiators.Applied Thermal Engineering24, 1291-1300, doi:10.1016/j.applthermaleng.2003.11.026 (2004).

6 Shati, A. K. A., Beck, S. B. M. & Blakey, S. G. The effect of surface roughness and emissivity onradiator output (awaiting publishing). (2010).

7 Institute, B. S. in Part 2: Test methods and rating (1997).8 Badr, H. M., Habib, M. A., Anwar, S., Ben-Mansour, R. & Said, S. A. M. Turbulent natural convection

in vertical parallel-plate channels. Heat and Mass Transfer43, 73-84, doi:10.1007/s00231-006-0084-z (2006).

Nomenclature:Symbol: Description: Units:

Heat transferring surface area (m ) Specific heat capacity (Jkg1K1)

Gravitational acceleration (ms2)h Heat transfer coefficient (Wm

2K

1)

Thermal conductivity of the fluid (Wm1K1) Enclosure wall length (m)

Water mass flow rate (kgs1) Convection heat transfer from the wall to the air (W) Net radiation heat exchange between the wall the radiator (W) Heat loss by Conduction through the wall (W)

Convection heat transfer from the radiator to the air (W) Total heat transfer from the radiator (W) Total heat transfer to the air (W)

Water inlet temperature (K)

Water outlet temperature (K) Wall surface temperature facing radiator (K) Wall surface temperature facing outside (K)

Thermal conductivity of the wall (Wm1K1) Thermal diffusivity (m2s-1) Kinematic Viscosity (m s

-1)

Thermal expansion coefficient (K-1

)

Temperature difference (K) Nusselt number h (-) Rayleigh number

(-)


8/16

Appendix 1.0 Schematic of apparatus

Cistern

Tank withimmersion

heaterFlowmeter

Pump

Flow ratecontrol pipe

Radiator

ValveT1 T2


9/16

Appendix 1.1 Test right setup

Fabric tape

Sandpaperbacking layer

Foaminsulation

layer

Inlet pipe andinsulation

Outlet pipe andinsulation

Serialthermocouple

interface

Thermocouple (x15)

Airflow monitorRadiator

Mounting frameand weights

50mmair gap

Clampand

stand

Regulator

valves (x2)

Chipboard wall


10/16

Temperaturecontroller/regulator

Hot water tankw/immersion

heater

Cistern

Water pump

Control valve (x2)

Rotormeter

USBThermocouple

interface


11/16

Appendix 1.2 - USB interface thermocouple layout:

Thermocouple Type Name

TC 1 T Ambient air temperatureTC 2 T Water temp inTC 3 K Radiator air temp bottomTC 4 K Radiator air temp middleTC 5 K Radiator air temp topTC 6 K Wall temp frontTC 7 K Wall temp backTC 8 K Water temp out

TC 8

TC 2

TC 1

300mm

TC 5

Left to right:TC 7, TC6, TC4

TC 3

150mm

300m

m

750mm

20mm

25mm

50mm


12/16

Appendix 1.3 - Serial interface thermocouple layout (un-insulated side):

Appendix 1.4 Test surfaces

Silver sprayed surface (left), Black sprayed surface (right)

Thermocouple TypeTC 1 TTC 2 TTC 3 T

TC 4 TTC 5 TTC 6 TTC 7 T

600mm

550mm

300mm

50mm

50mm

300mm

550mm

600mm

TC 6

TC 4

TC 2

TC 7

TC 1

TC 3

TC 5


13/16


14/16

Appendix 1.7 CFD Vector flows (plain wall).

Appendix 1.8 CFD Temperature contours (plain wall).


15/16

Appendix 1.9 Experimental results.

Appendix 2.0 CFD results comparison.


16/16

Documents

53_Krys Bangert.pdf