dmitry yudovsky - uclayudovsky.bol.ucla.edu/projectportfolio.pdfproblem statement purpose: bubbles...
TRANSCRIPT
Dmitry Yudovsky, Project Portfolio, 2006
Dmitry Yudovsky
Project Portfolio,Fall 2002 through Spring 2006
2
Project Portfolio
� UCSD, MAE156B, Winter 2006� Z-Microsystems Server Cooling. Corporate Sponsored Project
� Lawrence Livermore National Laboratories, Summer 2005� Numerical Optimization of High Pressure Containers
� Applied Materials, Inc; Electro Copper Plating Group, Summer 2004� Bubble removal from Liquid Delivery System
� Applied Materials, Inc; Chemical Mechanical Planarization Group, Summer 2003
� Metrology Unit Validation � Improvement to Factory Automation Software Development Process
� San Diego Supercomputer Center, Spatial Information Systems Laboratory � GIS HTML Viewer
� UCSD Structural Engineering, Powell Lab, Professor Elgamal, Structural Health Monitoring Laboratory, Ongoing Projects
� Satellite Based Structural Healthmonitoring of the Vincent Thomas Bridge� Remote Structural Healthmonitoring of Voigt Bridge
Dmitry Yudovsky, Project Portfolio, 2006
Lawrence Livermore National Laboratories.
Manufacturing and Materials Engineering Division
KDP Crystal Mitigation Machine for the National Ignition Facility (NIF)Architecture and HMI Development
Summer, 2006
4
Lawrence Livermore, 2005. Page 1
Purpose
Background
Challenges
Method Choose
Problem Statement
5
data
HMI
interfacesensors
actuator
DXC-390
Pressure switches
amplifier
USER2: Stepper motorServo motor2 linear drives
MDIG Code
Lawrence Livermore, 2005. Page 1System Integration
6
Lawrence Livermore, 2005. Page 1Methods – Software architecture
Event driven architecture, object based.
World Event
Status Variable
Faulted?
Change Event
Fault Check
GUI
Key strokeMotor velocityInterlock breachSystem event…
Fault Event
Dmitry Yudovsky, Project Portfolio, 2006
UCSD, Mechanical and Aerospace Engineering Course
156B
Z-Microsystems Server Cooling ImprovementCorporate Sponsored Project
Winter 2006
8
UCSD MAE 156B Project. Page 1
SPONSOR:Z Microsystems • Leading The Evolution In “Field-Ready” Computing.• Design & Manufacturing of Rugged Computer Hardware.
PURPOSE:Sealed Computer Module (SCM) is overheating. Develop a cooling system that is passive and rugged. It must keep the internal temperature at 25ºC above ambient
9
Proposed Design
Air DamHeat Sink
Old Design
Thermal Pad
Top Cover
Thin Grease Layer
Aluminum Block
IMPROVEMENTS
• Eliminates About 50% of The Thermal Resistance Interfaces.
• Doubles The Effective Surface Area Of The Top Panel And Localizes It Above The CPU.
• Modifies Top Panel To Increase Airflow
ADVANTAGES
• Passive Cooling
• Does Not Require Refrigerants or Liquids.
• Has Zero Degrees Of Freedom
UCSD MAE 156B Project. Page 2HARDWARE CONFIGURATION
10
SIMULATION RESULTSHeat Sink Optimization for fin spacing (A) and thickness (T) . Coded in Matlab.
1 2 3 4 5 6 7 8 9 100.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
a - spacing (mm)
Res
ista
nce
C/W
Resistance vs. fin spacing for servarl fin widths
0.1mm
0.6mm1.1mm 1.6mm
Resistance Decreasing
Spacing Increasing
Sweet Spot
UCSD MAE 156B Project. Page 3
11
UCSD MAE 156B Project. Page 4
TEST RESULTS
At 100% system load and 23ºC ambient temperature:
• SCM internal temperature rises by 23ºC from ambient
• CPU core temperature rises 34ºC
Spec requirement met
PROTOTYPE -5 00 0 5 00 100 0 1500 200 020
25
30
35
40
45
50
55
60
Time (se c on ds)
Tem
pera
ture
(C
elci
us)
Temperatue vs. Time
Dmitry Yudovsky, Project Portfolio, 2006
Lawrence Livermore National Laboratories.
New Technologies Division
Numerical Optimization of High Pressure Containers
Summer, 2005
13
Lawrence Livermore, 2005. Page 1
PurposeOptimize cylindrical, multilayered high pressure vessels for ultra-high speed wind tunnel
BackgroundVarious techniques exist for raising the overall strength of vessels (such as cannons, or canisters). This includes increasing the strength (and cost) of materials, shrink fitting concentric shells, and over straining (autofrettaging) shells. An optimal assembly with shrink fitting and autofrettaging can be found.
ChallengesMultivariable optimization problem with many non-linear constraints. Numerically, negative radii might be great, but the solution must be kept realistic.
Method ChooseGradient projection method. Find steepest descent and project it on the set of active constraints.
GeometryMaterials
Thermal loadingPressure loading
Autofrettagetreatment
Gradient Projection
Stress Analysis
Optimal geometry
Problem Statement
14
Geometrical configuration of problem
… more shells
1 2 n N-1 N…
r2 rn rn+1 rN rN+1r1
d1 dn-1 dn dN dN-1
r = radius d = interference or radius mismatch
P1 PN+1
sym
met
ry
Po
Pi
r1
r2
unit length
axisymmetric
Lawrence Livermore, 2005. Page 2
15
Numerical Optimization Using Gradient Projection Method
� Definitions� Active set: the set of currently enforced constraints� Inactive set: set of constraints that, when violated, become active
� Process1. Find cost function gradient2. Select active constraints3. Project gradient onto active set4. Go down projection
gn(x)=0
Iterative downhill search
Breach step
Projected gradient
Ineq
uality
cons
traint
gn(x)<0
gn(x)>0
Lawrence Livermore, 2005. Page 3
16
Graphical User Interface Lawrence Livermore, 2005. Page 4
prop
rieta
ry
17
Calculation Results
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8-0.5
0
0.5
1
1.5
2
2.5
3
factor of safety
log(
cost
)
Relative log of cost vs. factor of safety
doubling FS about increase cost by ~10
Multilayered pressure vessel. Optimal cost varies with given require factor of safety. Obviously the cost increases, but here we see that in increases exponentially. The y axis is normalized cost on a log scale.
Adding a temperature gradient during operation imposes an added stress to the assembly. We see that by greatly heating the inside of the assembly, a total diameter must be made in order to sustain the thermal stress gradient.
Lawrence Livermore, 2005. Page 5
0
100
200
300
400
500
600
0
100
200
300
4000.5
1
1.5
T innerT outer
Log
Out
er r
adiu
s
Dmitry Yudovsky, Project Portfolio, 2006
Applied Materials, Electro Copper
Plating Group
Bubble removal from Liquid Delivery System
Summer 2004
19
Problem Statement
� Purpose: Bubbles form in a high pressure chem-delivery line. These are detrimental to the subsequent process.
� Background: Degassers are commercially available however these bubbles are caused by a chemical reaction, not dissolved gas. So a bubble trap is required.
� Challenges: Waste, startup time, and installation and maintenance costs mustbe minimized. This would be an upgrade to existing systems.
� Method: Utilize the disparity between liquid to surface interaction.
Applied Materials, 2004. Page 1
20
Bubble Trap Concept
2rpFpressure π=
delivery
clea
n liq
uid
waste
bubblesfloat up surface tension
pressure force
rFsurface πσ 2=
Find r to balance the two
Ideal bleed hole case
σ ≈ 0 for gas
Applied Materials, 2004. Page 2
21
Bubble Trap Implementation
L
pRQwaste
∆µ
π=8
4
µ very small for gasses
Ended up using long, very thin inner diameter tube. Friction (and some surface tension) effects counteracted the pressure force and allowed liquid to bleed at a prescribed rate, but gas to escape very quickly.
Hagen-Poiseuille:
Applied Materials, 2004. Page 3
Exit to atmospheric pressure
rLF zshear πτ 2=
2rpFpressure π=
Dmitry Yudovsky, Project Portfolio, 2006
Applied Materials, Chemical
Mechanical Planarization Group
Metrology Unit Validation
Summer 2003
23
� Purpose: Applied Materials CMP leads the world in wafer planarization. However, it only makes a crude, in process metrology unit to test the quality of the polishing. External metrology must be performed; this is done by third part units. CMP System Integration was validating two competitors. My function was to test to stability, functionality, and compatibility of the this unit with our CMP hardware and software.
� Challenges: � Two teams from different corporate cultures� Third party unit still in beta phase� Time constraints
� Accomplishments: � Prepared and implemented test scenarios� Found and helped fix many bugs
Problem StatementApplied Materials, CMP, 2004. Page 1
24
CMP ProcessApplied Materials, 2004. Page 2
wafers container
polishing process 1
polishing process 1
polishing process 1
polishing process 1
clea
ning
pro
cess
metrology validation
front end back end
back-end processes
pass
ed
failed
crude in-process metrology occurs during polishing
Unit of interest
Dmitry Yudovsky, Project Portfolio, 2006
Applied Materials, CMP Group
Improvement to Factory Automation Software Development Process
Summer 2003
26
� Purpose: Optimize the software development (SW) process dealing with factory automation (FA).
� Background: FA for a complex machine requires keeping track of 1000’s of events, status variables, errors, logging variables, and controllable variables in the operating system of the machine. Naming and numbering of these data variables requires careful planning; no two events can have the same logical address. At Applied Materials CMP SW, a software package simplifies the actual C++ coding by generating most of the header files and event handling functions. The input to this preprocessor is a syntax based language, GCD, where each datum is defined. The file was, in turn, generated from other files written by different members of the SW team in Excel, Perl, VB, or just text. This decentralized process was very cumbersome and slow.
� Solution: Create a one-source database (DB) that the entire SW team could share. The DB would be locked to one user at a time to prevent collisions, control number ranges, and allow for easy import from and export to any standard or user defined format.
Problem StatementApplied Materials, CMP, 2004. Page 1
27
Software development processApplied Materials, CMP, 2004. Page 2
Data (Variables, events, status reports, etc.) stored in different media (Excel, text, etc.). These require data specific converter to create a GCD file. GCD file in then compiled to C++ code.
Old Processdata converter
Events
GCD fileGCD
Preprocessor
Status Variables
Error variables
…
update C++ codeC++ Code
Data is stored in a database that sits on a network drive. Variables can be imported from GCD file. GCD file is generated with one click.
GCD fileGCD
PreprocessorC++ Code
New Process
MS Access Database on
networkEvents, Status variables, etc.
data, converter, importer
Import capability
Dmitry Yudovsky, Project Portfolio, 2006
San Diego Supercomputer Center,
Spatial Information Systems
Laboratory
GIS HTML Viewer Development
Spring 2002
29
� Purpose: Develop a web-based graphical user interface for a Geographical Information System (GIS) database.
� Background: GIS is a useful tool for decision making. It facilitates the graphical representation, superposition, and spatial querying of data.
� Method: Use XML, JavaScript, and DHTML to accomplish the said goal.
Problem StatementSDSC, 2002, Page 1
30
GIS database
Browser, JavaScrpit,
DHTML
Map generator
GIS = Geographic Information System
SQL
Vector data
XML
Images, data
User interacts with GIS database through an HTML viewer. This viewer must be portable over many platforms. And it must be pretty.
GIS OverviewSDSC, 2002, Page 2
31
Old Style ViewerSDSC, 2002, Page 3
static toolbar with limited featurs
layer list is unstructured
query results in pop-up window
32
New Style ViewerSDSC, 2002, Page 4
dynamics and hierarchal layer list
dynamic legend
minimize/move features added to windows
Dmitry Yudovsky, Project Portfolio, 2006
UCSD Structural Engineering
Department, Professor Ahmed
Elgamal
Satellite Based Structural Healthmonitoring of the Vincent Thomas Bridge
Ongoing
34
� Purpose: Build a a remote structural health monitoring system. Prove that off-the-shelf components can be used to do remote data gathering.
� Schematic:
RemotePC
sensors
DAQinternet
satellite dish
FTP LocalPC
Data processing
Data processing
Data processing
Project Overview – Vincent Thomas BridgeUCSD Structural Engineering, Page 1
Data flow schematic
On bridge In lab
35
� Sensors:� Temperature, surface, and air� Camera� Accelerometer
(red dot indicates approximate location)
� Extensive Collaboration� UCSD Structural Engineering� CalTrans� Hughes Network Systems� San Diego Super Computer Center
� Potential� New, more complex testbeds� Temperature to frequency response correlation
Project OverviewUCSD Structural Engineering, Page 2
http://healthmonitoring.ucsd.edu/
36
� Data for May 2005 through February 2006
Results – Thermal DataUCSD Structural Engineering, Page 3
37
Sunrise Single Frame
Results – Camera DataUCSD Structural Engineering, Page 5
http://healthmonitoring.ucsd.edu/vtb/remote/vtb_remote_imaging.jsp
Dmitry Yudovsky, Project Portfolio, 2006
UCSD Structural Engineering
Department, Professor Ahmed
Elgamal
Remote Structural Healthmonitoring of Voigt Bridge
Ongoing
39
UCSD Structural Engineering – Voigt Bridge
� Purpose: Monitor Voigt Bridge with accelerometers and a camera
� Methods: A camera is used to determine the location of a vehicle on a bridge. A network of accelerometers monitors the excitation cause by this vehicle along the bridge. Thus we know the location of the load and the excitation.
UCSD Structural Engineering, Page 1
http://healthmonitoring.ucsd.edu/voigt.jsp
40
Vehicle Tracking over Voigt Bridge UCSD Structural Engineering, Page 2
99 100 101 102 103 104 105 106 107 10850
100
150
200Location vs. f rame
frame number (seconds*.35)
loca
tion
on
brid
ge (
feet
)
28.0629MPH