resume

resume

RESUME / Research Summary

Ameet Deshpande

Email: adeshpan@ucsd.edu or ameet.deshpande@gmail.com

I am Third year doctoral student in Mechanical and Aerospace Engineering Department at UCSD

Started in fall 2004. Completed MS in 2006. Current GPA is 3.8

Major: Control Theory, Minor: Fluid Mechanics

Program Enrolled: MS/PhD in Mechanical Engineering

Expected Graduation Date: 2008

Research Interests

  • I am working on extending Max-Plus methods to solve Hamilton-Jacobi-Bellman partial differential equations which arise in almost every continuous time optimal control problem involving dynamic programming. Till now, such problems were subject to curse of dimensionality. Max-plus methods show promise to lead to better algorithms to solve these.

Salient Projects at UCSD

  • Wrote a from-scratch Direct Numerical Simulation (DNS) code in Fortran to simulate 3 D turbulence in channel flow in CFD course with Tom Beweley. This code used computationally very efficient multigrid techniques.
  • Used value iteration dynamic programming to find global optimal fuel efficient path on a very nonlinear Lunar surface. (Homework in Numerical methods course with Tom Beweley)
  • Implemented an adjoint based method of estimation and control of chaotic nonlinear system (e.g. Lorentz attractor) as a course project.
  • As a part of DARPA project, I worked on sensing and control of decentralized human networks of with emphasis on robustness to adversarial deception. This involved coding a state of art Bayesian net for belief propagation.
  • Worked on ASME, Human Powered Submarine team, from UCSD in 06, which won for us the 1st place in non-propeller category. I was responsible for design, CADing and assembly of control surfaces used to control pitch and yaw.
  • Currently involved in the developing a low cost DC-DC Maximum Power Tracking converter Axial flux Pedal power generator being developed at Mozda Collective (non profit) in Gujarat, India. Pedal Power, Wind power enables many economic activities in this off-grid village.

Professional Background

    • General Electric, Global research Center, Niskayuna. Automation & Controls Lab. (Summer Intern 05)
    • I worked on development of a novel tactile interface for a medical ventilation system. This involved Matlab modeling of a highly nonlinear system, control design, and realtime hardware in the loop simulation in Labview RT with human operator & ventilator in the loop. We also conducted extensive usability surveys on Doctors & Nurse anasthetics at AANA conference and improved the specs.
    • General Electric, Global Research Center, Bangalore. (July 2002 till June 2004)
    • I worked Mechanical Engineer in Edison Engineering Development Program (EEDP) in which we underwent rotations in 2 labs, namely AMT & EPST labs of GE global research. Also underwent a rigorous inhouse 32 week course (called A course) covering all GE's business areas, from gas turbines, electrical machines to nanotechnology & wind energy. Also completed my 'Six Sigma Green Belt' project in GE.

Patents:

An algorithm for intelligent power uprating of a turbine without reducing life in process of filing.

Awards:

HATS OFF award for work in wind energy program

Training:

IMPACT training covering leadership & finance at Crotonville, NY.

FOL training for personal leadership & communication skills.

Projects at GE GRC Bangalore:

  • Adhesion Test Rig setup (with GE Rail)
  • A 2 disk adhesion test rig was built sourcing from local vendors, which was basically a motor & a generator attached with a friction coupling, to replicate wheel & rail contact in a locomotive. This involved prototype building as well as interfacing it with a real time systems, like Opal-RT, thus serving as a control implementation platform. With this prototype, we were able to achieve stable creep tracking on the back side of adhesion curve.
  • Learnt: Working with vendors, stepper motors, DC motors, Real time OS, nonlinear control
  • Locomotive Adhesion/Traction Control (with GE Rail)
  • Tractive effort in a RailLoco depends upon the friction between wheel & Rail which is a nonlinear function of slip velocity/creep. This project aims at maximizing traction by a optimal control of wheel torque, leading to higher pulling power. This system being inherently unstable at backside of creep curve, poses a significant control challenge. Was able to study & model in Matlab the suspension dynamics of locomotive (6 wheels 30 states) and find reasons of instability at lower speeds. Also improved the wheel speed estimation by implementing an observer. We also tested the controller in GE Rail test track in Erie, PA, sitting next to the train operator!
  • Learnt: Observer design, Suspension dynamics, Locomotive Tractive system, leading to higher pulling power.
  • Improving Supervisory controller of a wind turbine (with GE Wind)
  • Due to tradeoff between operating power and life of a wind turbine, turbine downrates itself in high winds. As turbine fatigue, hence life depends on both mean and variance of wind speed, there is opportunity to uprate (operate at higher power) turbine in high mean, low variance winds. An intelligent power uprating algorithm was proposed for such turbine. And its benefits validated on Flex5, an inhouse load simulation package. Leading to significant increase in power capture without compromising life. Received a HATS off award from EPST, and submitted a patent disclosure on An algorithm for intelligent power uprating of a turbine without reducing life which is in process of filing.
  • Learnt: wind Turbine controller theory, fatigue damage simulation, rainflow counting, design of Experiments, matlab text/bin file i/o
  • Implementing spectral models for simulating atmospheric turbulence in MATLAB (with GE Wind)
  • To simulate loads on a wind turbine, one needs to simulate wind. To make such wind as realistic as possible, several models are proposed. Model developed by MANN has been the most recent & best to model wind turbulence. This model as well as others like Kaimaal & Von Karman model for statistical turbulence were coded in MATLAB, along with graphical user interface. As a benchmark case 3D wind at 100 grid points in rotor plane was generated. This wind satisfied all 3*100*100 coherence spectras which validated the method. This program will be used to generate wind files for loads simulation of a wind turbine using Flex5.
  • Learnt: Theory of Statistical turbulence, simulation of random processes, spectral analysis, Fast Fourier transforms, Signal processing toolbox Matlab.
  • Estimation of availability of gas turbines using neural networks, for use in warranty documents. (with GE Power systems)
  • Downtime of a turbine needs to be accurately estimated, as it enters warranty documents. But its estimation is difficult as it depends on too many noise variables. A backpropagation neural network was used to build a transfer function between about 32 Xs & downtime, which gave significant improvement in the goodness of fit with data than current method for all turbine frames. Also automated the constant training of neural network, by a visual basic macro, which also did pre & post processing and visualization.
  • Learnt: Neural networks, Neuralware software, writing Visual Basic macros, Gas turbine operation & maintainance
  • Flex5 user manual (with GE Wind)
  • Flex5, inhouse load simulation software used by GE Wind energy was complex pascal code, developed in stages by danish, german & english teams. But it was undocumented, leading to very steep learning curve. Attending to an urgent business need to document it, working with a team of 3 at feverish pace, we successfully delivered a 200 page user manual, in a short period of 2 months.
  • Learnt: Flex5, wind turbine simulation, project management, time management, all aspects of documentation.

Academic Background

  • Mathematics: I was ranked third in INMO (Indian National Mathematical Olympiad) 97 and was selected among 30 students across India for attending IMOTC (International Math Olympiad Training Camp) twice in 97 and 98.
  • I did 2 years of M.Sc. Maths postal program with ISI Calcutta (Indian Statistical Institute) as a part of a 'Nurture Program' for keeping IMOTC students active in mathematics.
  • IIT JEE: I was selected in IITJEE 98, with all India rank 347 out of about 3000 admitted
  • School: I secured a 15'th rank in Pune SSC board with 91.8% marks in 10'th std. and 14'th rank in pune HSC board with 98.9% marks in 12'th std. also was recipient of 4'th and 7'th std. scholarships.
  • NTS: I was a recipient of National Talent Search scholarship in 10'th std. was also selected in top 10% of partA exam of Indian national physics Olympiad (INPhO 98).
  • In IIT: I was ranked third in mechanical dept. in first year with CPI=9.03/10.0 My CPI after 4 years is 8.24 on 10 point scale which translates to 3.46 on a 4 point scale. Method used for this conversion is described at the end.
  • -> Projects in IIT:
  • 1. Mathematical Modeling and simulation of a yet unexplored phenomena of a pulse tube cooler, involved extensive use of MATHCAD and MATLAB. This was winter project in December 2000, with Prof. M.V. Rane
  • 2. Developing sensor fusion algorithms and Graphical User Interface for a sensor 'skin' developed at BARC (Bhabha Atomic Research Centre)showing an animated and clickable display of incoming data from 57 sensors in form of various views like time graphs, sensor gridview, histogramview. It also contained a module for calibration. Involved learning and using VC++ and MFC. This was during summer internship at BARC (may, june 2000) with under Dr. Debanik Roy, DRHR, BARC.
  • 3. Third Year Design Project( course conducted by H.M.Hirani)
  • developed a novel kind of rotary handle for cyclic brake which gave nearly 10 times reduction of hand force compared to 5 times in conventional cycle brakes.
  • 4. Developed a laser guided path follower robot using 8051 microcontroller for Open Hardware Competition in Techfest,2000 (annual festival in IIT) which could traverse any reflective path. Involved learning architecture and interfacing of 8051 and programming in its assembly language and elementary circuit design.
  • 5. NonTechnical: Through a group in IIT called GRA (Group for Rural Activities) we conducted an energy survey of 8 villages in Madhyapradesh in context of an alternative dam project on Boi river.
  • Senior Year Dissertation (B.Tech. Project)
  • Estimation of verticality of a walking robot using multisensor fusion.
  • Working on lines of Maybeck's 3 Books, I implemented an Extended Kalman Filter algorithm for a Tilt sensor fitted on the chassis of 6 legged robot being developed at IIT Bombay. (http://www.iitb.ernet.in/latest/natraj/)
  • Home Paper (seminar)
  • Map Building for Mobile Robots
  • Compiled a literature review covering many aspects of map building and sensor fusion like bayasian rule, dempster-shafer's rules, kalman filtering statistical sensor fusion and their applications to sonar maps, vision sensors etc
  • Fields of Interest
  • (Explored) Automatic control, Robotics, sensor integration (fusion) IEOR, stochastic models, nonlinear dynamics, Appropriate Technology, Artificial Intelligence. Statistical & classical Thermodynamics, Solid Mechanics, Game theory
  • (Unexplored) Macroeconomics, Statistical Turbulence, Chaos theory, Quantum physics
  • Software Exposure
  • C, C++, java, fortran, Windows programming in VC++, MFC (microsoft foundation classes)
  • Softwares -> MATHCAD, PROEngineer, ANSYS, AUTOCAD, Neuralware
  • MATLAB (Simulink,Image processing,Neural Net,Stat, Signal Processing toolbox)
  • Platforms: linux, windows, unix(solaris).
  • Test Scores
  • G.R.E. (general) 2320 (730-verbal 790-analytical 800-quantitative)
  • T.O.E.F.L. 270/300
  • Also selected for Management program in IIM-Calcutta in 2004, through Common Admisson Test (CAT), 2004.
  • Other Interests
  • Reading books on Mathematics, Thermodynamics & physics and reading Marathi novels. Amateur robotics & hobby electronics. Sketching portraits and landscapes. Trekking, Traveling & walking long distances.
  • References
  • Available on request