Research at small university

1
Research at small university

• Ognjen Kuljaca
• Department of Advanced Technologies
• Systems Research Institute
• Alcorn State University

SWAN 06
2
Outline

• Problems
• Solutions
• Current Research Areas at SRI (where are we now)

SWAN 06
3
Problems

• Small number of students (about 3000) small
  pool of quality students to help with research
• Undergraduate institution no grad students
• Cant do what you want have to adapt to
  environment and find interesting projects
• Culture teaching institution that wants to go
  to research, but does not have infrastructure,
  culture and faculty evaluation based on research
  work

SWAN 06
4
Problems

• Small number of students (about 3000) small
  pool of quality students to help with research
• How to pay students (culture is to give them
  scholarships, so once they got it there is not
  much incentive to work outside regular student
  obligations)
• If you have to go through scholarships, you dont
  get to choose, financial aid office makes their
  picks
• If you have to go through limited number of
  positions available at Department level, the same
  problem jobs are given by financial aid office
  based on need, not merit

SWAN 06
5
Problems

• Undergraduate institution no grad students
• Undergrads can do something but not much
• Only way to get people to work is to get some
  people in post-doc or research positions
• The usual departmental structure does not support
  such positions, there are places for faculty and
  instructors only.

SWAN 06
6
Problems

• Cant do what you want have to adapt to
  environment and find interesting projects
• When you negotiate, there are relatively good
  existing labs, but need modernization
• Fund are promised for modernization
• When you talk with Vice-president for research
  about funds (lets do robots, vehicle control,
  neural networks and other sexy stuff he tells
  Look young man, we raise pigs here (actually
  they do more, but are interested in support for
  their strong agricultural program, not starting
  completely new activities with uncertain
  outcomes)

SWAN 06
7
Problems

• Culture teaching institution that wants to go
  to research, but does not have infrastructure,
  culture and faculty evaluation based on research
  work
• Infrastructure you want to write a grant
  proposal or work with local industry there is
  no financial lines in existing departmental
  budgets to support such activities if you get
  money, you cant spend it
• Culture we do teaching, so every expense from
  Department budget not connected with teaching
  work has to be explained to administration
  average time to do things varies between forever
  and never-ending
• Cant get money for research needs and trips
  (guys who only teaching complain that they are
  pushed back and they are right in many cases)

SWAN 06
8
Solutions

• Necessary conditions
• Have faculty interested in growth and research
  (Department chair and 3 faculty out of 10 total)
• Establish research entity so that research money
  is channeled through that entity and not troguh
  the Department
• Why new entity?
• With the new entity, there is a possibility to
  have money reserved for planned activity, employ
  students at will and have post-doc and research
  positions available
• Such entity provides the administration framework
  to solve all the problems, excerpt finding the
  research topics to secure support form
  institution

SWAN 06
9
Solutions

• Step 1
• 2004 early after a lot of beers, sweet tea and
  chicken wings a proposal for research entity is
  submitted to University new entity is called
  Systems Research Institute - SRI - (not precise
  name by plan nobody can tell you that something
  you want to do does not fall under definition of
  your institute helps with internal politics
  issues)
• Since we dont ask for any initial money,
  University and MS Institute of higher learning
  approve the new entity in early 2005
• SRI Director is Department Chair, associates are
  faculty interested in research work, SRI has its
  own budget line and can employ people almost at
  will

SWAN 06
10
Solutions

• Step 2 find the initial money
• Applying to federal grants not much success
  initially, lack of experience and institutional
  size
• However, look also in your backyard
• Backyard not so small, one of two Nissan
  factories in USA is based in MS, merely two hours
  drive from Alcorn (that is next door in MS, state
  is definitely not overpopulated)
• After some more beers, sweat tee and chicken
  wings, Nissan wants to look nice, so they give
  30,000 per year to pay students to work on
  projects (that should have been done through
  Alcorns Foundation office, but we keep them on
  need-to-know basis, they do not need to know)

SWAN 06
11
Solutions

• Step 2 find the initial money
• Look for regional needs
• Undeveloped area, high unemployment
• Shipyard in New Orleans is interested in having
  enhanced their ultrasound images of welds and
  have people trained to read them
• Write the proposal to DOL, will train people and
  generate jobs, DOL gives 1 mil grant in early
  2005
• So, SRI is here, it has budget line and it has
  money in the budget

SWAN 06
12
Solutions

• Step 3 find the people
• Look at the university university gets HBCU-UP
  grant to improve students performance in math,
  science and technology grant is originally done
  to support only math and chemistry students
  more beer, sweat tee and chicken wings and our
  department students get in the program more
  student workers we choose and pay from these
  funds
• Find faculty and researchers DOL grant pays
  image processing research a piece of luck
  there is an opening since some old faculty left
  so we get one image processing guy in faculty
  position and four more on short term contract
  from DOL funds (early 2005)
• However, there are no people to help start with
  control activity and setting up the labs go to
  old friends and ask for help ask Frank can he
  borrow a smart student for a semester or two to
  work as visiting instructor Frank helps and Mr.
  dang comes to Alcorn
• With image processing people, we work on new
  grant proposal to DOD TATRC for CAD system to
  process filmed mammograms and organize a database
  of breast cancer cases (Summer 2005)
• TATRC approves and grants 2.6 mil for 2.5 years
  (funds include six new full time research
  positions, funds for students and post-docs)

SWAN 06
13
Solutions

• Setbacks
• Foundation finds out about Nissan funds and takes
  the management less funds for us, but still we
  get some money for students
• Katrina hits shipyard sails to Gulf of Mexico
  and suffers from Titanic syndrome
• So, what to do with DOL grant.
• PIGS ARE BACK (AND COWS ALSO)
• well, those of us who are meat eaters know what
  is a good steak or schnitzel juicy, but not
  much fat what animal scientist says, you dont
  want outside fat but you need intramuscular fat
  (which also is not saturated, so not so
  unhealthy)
• so, the problem is how to grow pigs or cows that
  will have fat like that (and these can be sold
  for much more money)
• So, DOL grant officer is asked for changed of the
  scope of the grant and extension which is
  granted since the original beneficiary of the
  grant is at the bottom of the see more image
  processing people, money saved from TATRC grant
  more research and post-doc positions

SWAN 06
14
Solutions

• In meantime
• We all know that setting up new administrative
  units and having grant proposals funded takes
  time
• A lot of energy goes into these things, but,
  something had to be done in meantime
• So, do what you can
• Try to get pipeline form high school to get more
  quality students write small teaching grant to
  NASA-it goes through, 150.000 for one year use
  the grant to place students in Army ERDC center
  in Vicksburg new contacts, allow us to compete
  for DHS funds
• DHS for MS has particular interest in chemical
  sensors development adopt new chemical faculty
  into SRI, help her with grant proposal, granted
  in fall 2006, 950.000 for two years we have a
  small piece, but great working relationship and
  we are competing together for more

SWAN 06
15
Solutions

• In meantime
• Work with people you know and try to get some
  research topics and funds from them (in my case,
  go back home, work with my former company,
  research on power systems control not much
  money, but sustains activity)
• Work on cooperation with other departments and
  schools (Agriculture, pigs and cows
• Go to local industry (propose to Nissan to start
  working on Condition Based Maintenance, they
  refuse, but give you small project to network
  their robots, then to built small database for
  error codes)

SWAN 06
16
Current Research Areas at SRI (where are we now)

• Personnel for SRI
• 3 PhDs 50 percent of time in image processing
• 1 PhD 50 percent of time in control
• 1 PhD 50 percent of time in chemistry (coat
  sensors development)
• 1 full time software engineer
• 1 full time computer networking and busines
  administration person (MS in computer science and
  MBA)
• 6 new people hired in last 4 months
• Still looking for one faculty position in control
  systems, 6 research positions or post-docs in
  image processing, 3 research positions or
  post-docs in control and sensor development, 4
  research positions or post-docs in software
  engineering

SWAN 06
17
Current Research Areas at SRI (where are we now)

• Projects
• Development of a Knowledge Base to Support
  Detection and Diagnosis and Research in
  Mammography (funded by DOD TATRC just started)
• Goal This project focuses on development and
  implementation of a Computer Aided Detection
  (CAD) and Diagnosis (CADx) system with a
  searchable database (both by images as in content
  based searches as well as by metadata stored as
  text) to support mammography in clinical practice
  as well as in research.
• So, where is control or neural networks in there?

SWAN 06
18
Current Research Areas at SRI (where are we now)

• Projects
• Development of a Knowledge Base to Support
  Detection and Diagnosis and Research in
  Mammography (funded by DOD TATRC just started)
• Diagnosis or risk assessment depend on the
  results of mammogram and risk factors (age, race,
  weight, occupation, number of children,
  pregnancies carried out or aborted etc.)
• Many factors, so, how to find the correlations
• Statistics was helpful, but it really does not
  give the good models

SWAN 06
19
Current Research Areas at SRI (where are we now)

• Projects
• Development of a Knowledge Base to Support
  Detection and Diagnosis and Research in
  Mammography (funded by DOD TATRC just started)
• Possible solution neural network identification
  and modeling that will take into account all
  these factors together with mammogram results

SWAN 06
20
Current Research Areas at SRI (where are we now)

• Projects
• Development of a Knowledge Base to Support
  Detection and Diagnosis and Research in
  Mammography (funded by DOD TATRC just started)

SWAN 06
21
Current Research Areas at SRI (where are we now)

• Projects
• Image processing techniques to access
  intramuscular fat in cows and pigs (funded by
  DOL)
• Goal This project focuses on development and
  implementation of ultrasound measurement of
  intramuscular fat in cows and pigs.
• So, where is control or neural networks in there?

SWAN 06
22
Current Research Areas at SRI (where are we now)

• Projects
• Image processing techniques to access
  intramuscular fat in cows and pigs (funded by
  DOL)
• Control of ultrasound beam frequency and power
  (develop the sensor, medical ultrasound sensors
  are to expensive and are tuned to work for
  humans, not animals)

SWAN 06
23
Current Research Areas at SRI (where are we now)

• Projects
• Image processing techniques to access
  intramuscular fat in cows and pigs (funded by
  DOL)
• Amount of fat depends on breed, food, placement
  of the animals, weather, and many other factors.
  Animal scientists know what they want, but they
  cannot find the correlation between these
  factors. So, developing the neural network, or
  fuzzy logic or expert system to guide decision
  during the raising of the animals will help to
  achieve the quality and cut down cost. This is
  our next goal out of this project aimed to obtain
  funding form DOL and USDA

SWAN 06
24
Current Research Areas at SRI (where are we now)

• Projects
• Advanced control of power systems (funded by
  Croatian ministry of Science, cooperation with
  Faculty of Electrical Engineering and Computing,
  University of Zagreb, Croatia)

SWAN 06
25
Current Research Areas at SRI (where are we now)

• Projects
• Advanced control of power systems (funded by
  Croatian ministry of Science, cooperation with
  Faculty of Electrical Engineering and Computing,
  University of Zagreb, Croatia)
• All parameters can and do change. How to control
  efficiently?
• NN control scheme

SWAN 06
26
Current Research Areas at SRI (where are we now)

• Projects
• Advanced control of power systems (funded by
  Croatian Ministry of Science, cooperation with
  Faculty of Electrical Engineering and Computing,
  University of Zagreb, Croatia lead institution)

SWAN 06
27
Current Research Areas at SRI (where are we now)

• Projects
• Hydro turbine governors (funded by HEP and World
  Bank, cooperation with Brodarski Institute lead
  institution
• Simplified turbine governor scheme

SWAN 06
28
Current Research Areas at SRI (where are we now)

• Projects
• Hydro turbine governors (funded by HEP and World
  Bank, cooperation with Brodarski Institute lead
  institution
• Simplified turbine governor scheme in Laplace
  domain

SWAN 06
29
Current Research Areas at SRI (where are we now)

• Projects
• Hydro turbine governors (funded by HEP and World
  Bank, cooperation with Brodarski Institute lead
  institution
• Problem there are nonlinear jumps at certain
  frequencies
• Solution we used describing function to find the
  conditions when such nonlinear jumps can happen
  and then we used fuzzy controller to avoid this
  conditions (we dont have formal proof yet, but
  stuff works in practice)

SWAN 06
30
Current Research Areas at SRI (where are we now)

• Projects
• Nissan Condition Based Maintenance
• Problem Automobile assembly factory with several
  hundreds of robots.
• Maintenance Usual way. Close the line and do
  regular check-ups. 90 of things are fine. Waste
  of time and resources.
• Alternative network the robots, fill the data in
  database, design expert system to predict
  possible problems and do maintenance work only
  when needed.
• Proposed to Nissan, but rejected (we have world
  class engineers, what you peasants from MS can do
  better?)
• Are your robots networked? No. Well, let us try
  do that

SWAN 06
31
Current Research Areas at SRI (where are we now)

• Projects
• Nissan Condition Based Maintenance
• So, go slow. Use Nissan grant money for students
  to design and deploy wireless network for their
  robots.
• Nothing fancy, just basic networking, two
  undergrads did it during the summer

SWAN 06
32
Current Research Areas at SRI (where are we now)

• Projects
• Nissan Condition Based Maintenance
• Network is here, data can be read, so, they want
  to have a small database where technicians can
  read error codes and also write their own
  solutions to problems.
• No problem, pilot is done and checked with Nissan
  and it goes to test run in January
• Should it go through (and it will), the hardware
  and software base for condition based
  manufacturing is placed
• Late 2007, we start installing the software for
  data collecting and archiving .
• Where is the control? Well, to predict the place
  and time of fault, NN, or Fuzzy Logic, or expert
  system or some kind of combination of these will
  be needed.

SWAN 06
33
Thats all folksdont have a cool picture to
finish, but have a suggestionMAKE FRANK DO IT
AGAIN NEXT YEAR
SWAN 06