PowerPoint Presentation
Performance Evaluation of Plate-fin and Pin-fin Heat Sinks and
Design Optimization of Dynamic Cold Plate (DCP) ByParth Jayeshkumar
SoniMS Mechanical Engineering Graduate StudentThe University of
Texas at ArlingtonDate:03/21/2016Thesis Advisor: Dr. Dereje
AgonaferCommittee Members: Dr. A. Haji-Sheikh Dr. Miguel A
AmayaParth Soni Advisor: Dr. Dereje Agonafer1
3/22/2016Parth Soni1
Why Liquid CoolingPower Trends
https://www.cisl.ucar.edu/nar/2006/1.0.sc.jspParth Soni Advisor:
Dr. Dereje Agonafer2
23/22/2016Parth Soni
Focus Study 1: Design Validation of Dynamic Cold Plate (DCP)
Study 2: Parametric Study and Performance Comparison of Pin-fin
and Plate-fin Heat Sinks for the Application of Oil Immersed
Cooling
Parth Soni Advisor: Dr. Dereje Agonafer3
3/22/2016Parth Soni3
Focus Study 1: Design Validation of Dynamic Cold Plate (DCP)
Study 2: Parametric Study and Performance Comparison of Pin-fin
and Plate-fin Heat Sinks for the Application of Oil Immersed
Cooling
Parth Soni Advisor: Dr. Dereje Agonafer4
3/22/2016Parth Soni4
Original Cold Plate(OCP) vs Dynamic Cold Plate(DCP)One inlet one
outletSame water flow for all heat generating componentsParth Soni
Advisor: Dr. Dereje Agonafer5
Separate inlet and outlet for different compartment inside cold
plateDifferent flow for different compartment
3/22/2016Parth Soni5
What is DCP?Dynamic cold plate is extended version of original
cold plate
Parth Soni Advisor: Dr. Dereje Agonafer6
CAD modelFin placementReferance heat generating platform
Parth Soni Advisor: Dr. Dereje Agonafer7Dimensions (in mm)ASICs
14.71 13.31 0.8FPGA 10.5 12.7 0.8Reference Platform
Component
QuantityPower(W)Base1-ASIC1(B1)40ASIC115FPGA15LICA1370
ASIC:-Application-Specific Integrated
CircuitFPGA:-Field-Programmable Gate ArrayMCM serves as basis for
design of solution- Power dissipation of 485W over 78mm 78mm
3/22/2016Parth Soni7
CFD Modeling of DCPParth Soni Advisor: Dr. Dereje Agonafer8
Methodology
Modeling :SolidworksMashing:ICEM CFDAnalysis:Ansys Fluent Parth
Soni Advisor: Dr. Dereje Agonafer9
Model of Dynamic Cold Plate
Parth Soni Advisor: Dr. Dereje Agonafer10Dimensions cold
lateFoot print : 90*90 mmBase height: 5mm
Fin dimensionsThickness:0.5 mmHeight:2mmLength:29mmPitch:1mm
Dimensions of coverFootprint :90*90 mmHeight: 15mm of inlet
outlet:7mm Dimensions from Experimental setup
Purpose of StudyTo visualize the flow in side the cold plateTo
validate of the CFD modelFor, Optimization of flow inside cold
plateFor, parametric study of DCP
Parth Soni Advisor: Dr. Dereje Agonafer11
MeshUnstructured meshNo of elements:11,15,321Shape of elements:
Tetra elementsMesh algorithm:Robust (octree)top-down meshing
approachtop-down meshing approach
Parth Soni Advisor: Dr. Dereje Agonafer12
Mesh DensityFin surface:0.001 mFin thickness:0.005mOther
surfaces:1mParth Soni Advisor: Dr. Dereje Agonafer13
AnalysisModels in fluentEnergy equationTurbulent model:
K-epsilon turbulence modelMaterials used in the modeling
Parth Soni Advisor: Dr. Dereje Agonafer14
K-epsilon Turbulence ModelParth Soni Advisor: Dr. Dereje
Agonafer15
Turbulent kinetic energy kEddy Dissipation 10% turbulent
intensity from length scale model (Fluent user guide)
Boundary Conditions 4 inlets: Velocity inletsVelocity of water:
4 lpm (Reference experimental data)Temperature: Ambient Pressure :
3000 pa4 outlets: Pressure outletsNo back pressureParth Soni
Advisor: Dr. Dereje Agonafer16
3/22/2016Parth Soni16
ResultsParth Soni Advisor: Dr. Dereje Agonafer17
Particle Tracking
Parth Soni Advisor: Dr. Dereje Agonafer18
Temperature Distribution
Parth Soni Advisor: Dr. Dereje Agonafer19This figure shows the
heat spread on the copper plateAs temperature of B1 increases
neighboring chips also shows some higher temperature
Pressure Drop
Parth Soni Advisor: Dr. Dereje Agonafer20
Pumping power=Pressure drop * Flow ratePumping power= 2.2 W(CFD
result)Pumping power=2.76 w(Experimental result)
Validation Between ResultsParth Soni Advisor: Dr. Dereje
Agonafer21
Experimental Data: Ruturaj kokates Thesis
3/22/2016Parth Soni21
ConclusionThus, from the CFD results this model is in good
agreement with experimental resultsHence, this model can be used
for the further study of optimization and parametric study Parth
Soni Advisor: Dr. Dereje Agonafer22
Future WorkStudy 1:This model can be used to optimize the design
of the DCPCover design optimization for better flow rateParametric
study and documentation of different fins and different cover
design Parth Soni Advisor: Dr. Dereje Agonafer23
Focus Study 1:Design Validation of Dynamic Cold Plate (DCP)
Study 2: Parametric study and performance comparison of pin fin
and plate fin heat sinks for the application of oil immersed
cooling Parth Soni Advisor: Dr. Dereje Agonafer24
Study 1Parametric study and performance comparison of pin fin
and plate fin heat sinks for the application of oil immersed
cooling for open compute generation one serverParth Soni Advisor:
Dr. Dereje Agonafer25
Previous StudyValidation of the model using the identical
boundary condition generated in experimentDocumentation of
parametric study and the performance of the parallel plate heat
sink for the oil immersed applicationOptimization of the parallel
plate heat sink design
Parth Soni Advisor: Dr. Dereje Agonafer26
Parth Soni Advisor: Dr. Dereje Agonafer27Evaluate the
performance of pin fin and plate fin heat sink
Compare the thermal resistance of both the heat sinks with
current heat sink in use for open compute server
Motivation
MethodologyPlace plate fin and pin fin heat sink in place of the
parallel plate heat sink on the validated modelParametric study of
heat sinks for same condition as experimental set upBase height Pin
thickness and radius Flow rate Documentation of the performance of
both heat sinks
Parth Soni Advisor: Dr. Dereje Agonafer28
3/22/2016Parth Soni28
Types of Heat Sinks
Parth Soni Advisor: Dr. Dereje Agonafer29
http://img.diytrade.com/cdimg/720673/6272157/0/1214899819/pin_fin_heat_sink.jpg
http://www.enertron-inc.com/images/resources/forged/forged_2.JPG
http://www.newegg.ca/Product/Product.aspx?Item=N82E16816101827Parallel
plate heat sinkPin fin heat sinkPlate fin heat sink
Parth Soni Advisor: Dr. Dereje Agonafer30Material PropertiesPCB,
Ram- FR-4 epoxy resinDensity- 1900 kg/m3Thermal conductivity 0.17
W/m KSpecific heat 749 J/kg KHeat sink - AluminumDensity- 2700
kg/m3Thermal conductivity-218 W/m KSpecific heat- 900 J/Kg K
Mineral Oil Properties31STE Oil Company data sheets and
MSDS:http://www.steoil.com/msds-tech-dataDensity 1680 Kg/m3Thermal
conductivity 0.13 J/kg KSpecific heat 1680 Re: 4.6
Parth Soni Advisor: Dr. Dereje Agonafer
32 Flow ConditionsInlet temperature:30CVolume rate :1
lpmVelocity : 0.00115 m/sPressure : 6 psiRe : 13.6 Parth Soni
Advisor: Dr. Dereje Agonafer
3/22/2016Parth Soni32
CFD Set Up
Parth Soni Advisor: Dr. Dereje Agonafer33Heat sink data:Foot
print- 10 cm 7 cmBase height 0.6 cmTotal height 3.1 cmNumber of
fins 25Fin radius 0.6 cmModel dataCabinetFootprint: 35*35
cmRamDimension: 14*3*8 cm
Heat sink data:Foot print- 10 cm 7 cmBase height 0.6 cmTotal
height 3.1 cmNumber of fins 25Fin thickness 0.4 cm
Why Fixed Flow RatePressure drop in air cooling application is
around 4 PaFor parallel plate heat sink in in oil cooling
application pressure drop around 0.9 PaWhere as, in oil cooling
application pressure drop is around 0.032 Pa
Parth Soni Advisor: Dr. Dereje Agonafer34Pumping power=Pressure
drop * Flow rateAdvantage of fix Flow rate conceptLess time
consumingEasy to use
3/22/2016Parth Soni34
ResultsParth Soni Advisor: Dr. Dereje Agonafer35
Grid independent StudyParth Soni Advisor: Dr. Dereje
Agonafer36Inlet Temperature - 30 CVolume rate 1 lpm
Flow Rate vs Thermal ResistanceParth Soni Advisor: Dr. Dereje
Agonafer37Inlet Temperature - 30 C
Base Height vs Thermal ResistanceParth Soni Advisor: Dr. Dereje
Agonafer38Inlet Temperature - 30 CVolume rate 1 lpm
Include pic of base3/22/2016Parth Soni38
Fin Height and Fin Radius vs Thermal Resistance for Pin Fin Heat
SinkParth Soni Advisor: Dr. Dereje Agonafer39
Highlight optimize zone3/22/2016Parth Soni39
Fin Height And Fin Thickness vs Thermal Resistance for Plate fin
heat sinkParth Soni Advisor: Dr. Dereje Agonafer40
ConclusionBase height: Optimize base height can be 0.6cm as
after that the thermal resistance is not varying much.Fin
height:Fin height shows the optimum result at 2.5 to 3 cm as after
that flow passes through lease resistance pathFin radius or
thickness:recommended thickness of the fin can be 0.4mm to 0.6 mm
(Plate fin) as after that resistance increases and performance
decreasesRecommended radius of the fin can be 0.5 to 0.6 mm as
after that flow resistance increases and performance decreases
Parth Soni Advisor: Dr. Dereje Agonafer41
Future WorkPerformance of the other available heat sinks can be
documented for the same model as well as higher generation
serversThis same can be studied for the fix pumping power
method
Parth Soni Advisor: Dr. Dereje Agonafer42
Parth Soni Advisor: Dr. Dereje Agonafer43Thank
YouQuestions?...