Research Interest
A list of stuff that has held my attention at one point or another
Last updated: May 2008
Home: : Research
Listed below are some of the things that keep
me in graduate school. For the most part I won't go into do much
detail, either becaues I havn't published or really I think it would
just bore everyone to tears. Instead, a catchy summery and if you are
interested you can email me for more information.
Particle Counting and Characterization in Diesel Fuel 
The main difficulty in increasing the efficiency of a diesel engine is
in decreasing the amount of time it takes to burn all the fuel after it
is injected into the cylinder. To do this engineers have relied on
creating smaller and more evenly dispered droplets using smaller fuel
injectors and higher pressure. Unfortunately these fuel injectors are
much more susceptible to damage due to small particles that are
difficult to filter out. End result is that for the expensive engines
on trucks and in boats it would be nice to know when the fuel is going
to tear your engine a new one. I am developing a sensor that will
count the number of particles in the fuel stream and tell you size
distribution and type of particle.
Flow System and Concentrator for Toxin Biodetector
I was a consultant for a research project in the Food Science
department here at Purdue. The project itself was interesting for
several reasons but my part had several chalanges. The first chalange
was that we had to cycle a material that collected the analyte from the
air between room temperature and roughly 200 F. That, by itself, is not
too difficult except for the fact that anything in the air stream had
to be made of a particularly inert plastic called PEEK. Plastics by
themselves are not exactly poor conductors of heat so it became quite
difficult to pour enough energy into the collection material to
overcome the conduction losses through the PEEK. On top of this the
PEEK itself got quite hot so thermal expansion and warping became a
significan issue with the valves we used to manipulate the air stream.
All in all it was a good experiance in actually producing a working
prototype, though I am not sure I would do it again!
The design mentioned above is of a type called a "Batch Process". This
means that it is not a continuous process. At some point you have to
stop collecting analyte from the air so that you can heat up the
collection material and drive the analyte to the sensor. Because of
this break there is a time when you are not sampling the
surrounding air. Additionally, heating up then cooling the device is
very energy intensive. An ideal concentrator would act continuously and
use a fraction of the energy. This is another project that I am working
on durring my "free time". Basically, I am trying to design a
microfluidic device that continuously concentrates without using
changes in temperature.
At the moment I am working on a proof of concept test to demonstrate
the basic mechanism behind the concentrator. A significant difficulty
that I am encountering is that the small gaps (10 um) necessary to the
function of the device are very difficult to produce by hand outside of
a clean room. Such are the trials of un-funded research.
Lab-on-a-Chip platform using Optoelectrowetting
While alot of work has gone into lab
on a chip platforms there is still no solution that is, in a sense,
universal in its applications. People have developed speciallized chips
with great success and many of these have started working their way
into comercial systems. While a truely flexible platform has still
illuded us I see alot of promise in digital microfluidics which is basically performing chemistry in droplets.
We believe we have a system that can move an arbitrary number of these
dropets around, split them, sort them based on conductivity, heat/cool,
all while allowing the droplets to be analyzed by external probes. We
will accomplish this through open optoelectrowetting
, a technology developed by my labmate Oswald Chuang. The advantage of
this approach is that the light used to actuate the droplets is fairly
dim and the voltages involved are fairly low so the impact on the
droplets and their contents is minimal.
3D uPIV
One
of my goals when I started working on my PhD was to develop a 3D uPIV
system which uses a conventional microscope and camera. Conventional
uPIV is used to determine the two dimensional velocity of a fluid in
the focal plane of a microscope. Three dimesional uPIV can determine
the three dimensional velocity of a fluid in a certain volume of
fluid. I started working on it largely because the idea of
determining the three dimensional location of particles using one
camera just seems cool.
Outside of how cool it is, there is a real need for a simple,
inexpensive uPIV system that can handle relatively dense particle
loadings and non-ideal images. Most approaches to this problem
currently put some kind of mask in the microscope's optical path or
relie on relatively subtle changes to the particle image geometry
caused by the particles location relative to the focal plane. My
approach is similar but differs in that the changes to the particle's
image are not subtle and, ideally, no light is lost. Because it is
highly tolerant of overlapping particle images and noise in general the
system should be able to handle higher particle loads in the fluid
being analyzed which means better visualization of the flow field.
I'm currently building a proof of concept prototype to test the idea.
The prototype uses a binary amplitude hologram as an optical element so
it's performance will be significantly lower then an ideal system but
the hologram is much cheaper.
Baysian Decision Making in Sensors
My experiance with the biodetector and the particle counter has raised
the issue of decision making in sensors. How, exactly, do sensors make
decisions? Sensors, almost by definition, turn one signal into a more
usable one. For instance a carbon monoxide sensor in your home turns
the concentration of carbon monoxide into a voltage that a computer
chip then reads. The computer chip then has to decide, based on this
voltage, when to make the loud and terribly obnoxious sound to get out
of the house. Or, at the vary least, remove its battery. The computer
chip in a carbon dioxide detector has a relatively easy job but more
sophisticated sensors are not so clear cut. When, exactly, do you raise
an alarm? What is noise and what is signal? For complex sensors these
are not straight forward questions to answer.
One
way to answer these questions is through Baysian Decision Making. The
Baysian approach deals with probability, as opposed to thresholds and
conditionals like most decision making methods. As a result there is a
built in measure of "confidence" in the result. For example, on the
picture on the left, the two gently curving data sets have a
probability of detection of over 99% while the wondering somewhat
horizontal line is much less then 1%. The detection shown on the left
is actually somewhat difficult to determine in a traditional fashion
since the baseline value is different for each sensor, the baseline
tends to move over time scales comparible to what is seen in a
detection, and finally these variations can be almost as large as the
real signal. What is different is the shape of a detection signal vs.
that of noise and distinguishing between noise and a purposeful shape
is something this method is very good at.
Another application I am looking at is particlulary complex signals for
which there is no simple analytic way to determine the actual
conditions. For example I once saw a presentation by a researcher who
is taking siesmic data all over the west coast and then feeding it into
a computer to try to determine where and when the next big earthquate
is comming. Unfortuantely the models that can predict this take a
really long time to run so the use a Baysian decision making scheme
that uses prior data from running these models for a range of
conditions. They literally spent several months of computer time
running the models for hypothetical situations to build enough results
to construct what I like to call the "probability field". In this way
they can get real time predictions for an event where even a few
minutes advance notice could save many lives.
Thermoelectric Measurements of Shark Gel and Polyelectrolytes in Salt Solutions: Masters Thesis
While the title is a classic example of a wordy an confusing thesis the
research itself was quite interesting if useless. In a nutshell I took
shark gel, which I will explain in a minute, applied a temperature
gradient across it and measured any voltages that developed. I also did
this to polystyrene beads in several different salt solutions in an
attempt to work out some sort of theory as to why the shark gel
produced such a high voltage. Oh right, shark gel... Well sharks, as
you might be familiar with, can sense the electrical signals of other
fish. To do this requires nerves, just like every other sensory organ.
These nerves though need to be in electrical contact with the sea water
but still protected from germs, brushing up against something, and in
general the harshness of the ocean. The shark accomplishes this by
puting the nerve at the base of a rather long pore (can be several cm
long!) and filling that pore with a gel that conducts electricity. This
last fact is not so special as the gel is filled with sea water and sea
water also conducts electricity. In anycase it is this gel that I
studied. It would be difficult to find a more obscure topic for a
mechanical engineer.
None the less I am quite happy I worked on it as it brought to my
attention the thermoelectricity of ions in solution. For an ionic
solution to produce a voltage in a temperature gradient the ions
themselves have to move. When I tried to find out exactly why they
moved I got a chemist's answer "Because the ions move to the area of
lower chemical potential". To a mechanical engineer that doesn't answer
the question. Ya, it tells you when they move but not how. As a result
for the last 4 years I have been working between other projects on my
own theory for why this happens.
While alot of work has gone into lab
on a chip platforms there is still no solution that is, in a sense,
universal in its applications. People have developed speciallized chips
with great success and many of these have started working their way
into comercial systems. While a truely flexible platform has still
illuded us I see alot of promise in digital microfluidics which is basically performing chemistry in droplets. 
One
of my goals when I started working on my PhD was to develop a 3D uPIV
system which uses a conventional microscope and camera. Conventional
uPIV is used to determine the two dimensional velocity of a fluid in
the focal plane of a microscope. Three dimesional uPIV can determine
the three dimensional velocity of a fluid in a certain volume of
fluid. I started working on it largely because the idea of
determining the three dimensional location of particles using one
camera just seems cool. 
One
way to answer these questions is through Baysian Decision Making. The
Baysian approach deals with probability, as opposed to thresholds and
conditionals like most decision making methods. As a result there is a
built in measure of "confidence" in the result. For example, on the
picture on the left, the two gently curving data sets have a
probability of detection of over 99% while the wondering somewhat
horizontal line is much less then 1%. The detection shown on the left
is actually somewhat difficult to determine in a traditional fashion
since the baseline value is different for each sensor, the baseline
tends to move over time scales comparible to what is seen in a
detection, and finally these variations can be almost as large as the
real signal. What is different is the shape of a detection signal vs.
that of noise and distinguishing between noise and a purposeful shape
is something this method is very good at.