Friday, February 16, 2007

What will a PHD take?

So.. this PHD will require:

good supervisor
my supervisor is being pushy lately, but that could mean he's just concerned about me finishing my defense by May (as that seems to be the target of one of the other guys in my lab.)

-then again, I know we don't communicate very well. The thing is, I'm not sure if any of the professors at Windsor in EE would be "better" than X Chen. What I mean is, some are either likely more of a "hard ass" then his, retiring this year, not doing research I'm interested in, etc. However, I can think of the new EE prof ( Stephen O'Leary, who I'm TA'ing for) who I'd likely get along with well - I joke around with him a lot ... as a supervisor, who knows?.

- If I choose a different supervisor (and probably subject area) I lose a lot of time. I'm just getting used to literature in my subfield of control systems. Additional background survey work would take 4 months. I have a hardware platform to validate my ideas, which would lose me at least another 4 months time to build another for another subject area entirely. Approx 8 months.

- passing the comprehensive (which basically means you have to know the 8 core subject areas of UofW undergrad program in EE). I have an advantage here is that I have old "comprehensives" from which to study, know the curriculum well, and the professors who make the questions.

problem to focus on, proper motivation

Electric Power steering is interesting.. but not that interesting anymore to me anymore. The advanced motor drive control I've developed could potentially be applied to other types of electromechanical systems.


"Active Suspension" -- the idea has been around for a while, but no product yet available on a commercial vehicle. Bose is working on one.

"Brake by Wire" "Steer-by-Wire (no mechanical linkage, drive with a joystick!)".. I've did enough Steering stuff. I'd prefer to keep my mechanical linkage, thank you. It's the best "fault tolerance" you can design for a steering system. That is, if the electronics in your car fail, you don't die. ;)

Electronic braking (ABS) and EPS (electric power steering) are pretty much standard on vehicles now a days.

Hyrbid and electric vehicle technology are somewhat readily available, but again, could be improved by better controls I'm sure. Some of these systems get into heavy vehicle based modeling; I know the electric actuator part better.

Problem: these above subjects have been tackled even more then EPS.

So.. what then.....

Return to our Capstone 4th year project!

The fully varied, linear electric motor actuated, intake and exhaust valves to open and close IC engine valves.

Our system used "linear motors" (solenoids); the main limitation here was the size/power ratio of the solenoids could not meet the demands to drive the IC engine fast enough. However, I know hobbyists who have mastered the idea: it works, but doesn't work very well.

Our design used just a one-way solenoid to open the valve, and a spring to return it. Simple, and cheaper, is better here: more mechanical parts mean more wear. On a 6 cylinder 4 valve engine, that's 24 seperate actuators that would be needed. They need to have simple drivers, be heat efficient (being on the engine cylinder heads). However, changing the control design (assuming you don't have to add many sensors, which is the focus of my current research) just means adding some lines of code to your microcontroller.

Besides, there's many control objectives for the actuator:

- it needs to have a smooth motion (preferably "fit" the sinusoidal cam profile);
- it needs to deliver that power fast (as valves open, hold, and then close each time your engine makes a revolution... that goes past 7000 RPM!).
- there must be enough force available initially to ensure the actuator begins to move.
- Once the solenoid is done opening the valve, it needs to apply only the power required to hold the valve open, not more (as this wastes energy, and dissipates tons of heat in the solenoids.)
- Variable Valve timing required (this can be done with a few hall effect sensors mounted on an IC engine, as per our Capstone.)
- Variable valve lift required
- You need to dynamically control the position of valve (variable valve lift) - preferably without more hardware.

So, the controller here probably HAS to be complex. EPS only 3 basic objectives, in comparison, and dynamics are more linear. So, it's a good target for advanced control design. I suppose advanced aircraft engine or actuators would be interesting, too ( a fellow grad student is working on advanced design for a jet aircraft engine.)

All this implies you need to know information about your system to make better "control decisions" to make the actuator work better. You can certainly measure the force of the actuator (by measuring the current with a $0.01 resistor). You might might need speed/position information to do this, or know, or at least estimate, the "gas" force holding down the solenoid.

Indeed, my idea is perhaps more important now. Not having to add a $20 tachometer or $200 torque sensor to your steering assembly is still a big deal; having to add 24 high speed position sensors to the valve actuators would be expensive and a wiring/failure nightmare.

Most current industry research solutions use a dual acting valve -- that is, 2 solenoids, which allows you to use weaker springs and solenoids. No commercialized product, yet, though.

So, if my supervisor wants to get into more demanding, state of the art, vehicle/engine controls, this is the way to go.

Finally, I would have to:
-pass the comprehensive (which basically means you have to know the 8 core subject areas of UofW undergrad program in EE). I have some advantages: I have old "comprehensives" from which to study, know the curriculum well, and the professors who make the questions.

I figured out some things:

1) My supervisor and I were looking at my master's thesis problem much different: I found a solution to a similar problem recently published in literature in order to solve the problem he gave me: "Fault tolerant control for an EPS"; my idea was to combine other methods to improve said performance. That's what he has been saying for over a year now, as to what I need to do to get my master's (which implies a publishable(s) result).

2) I look at the research in what I can do to finish my master's -whereas my supervisor wants papers suitable for good publication.

3) I've learned a few things about "what a suitable paper is".
- it must be published, preferably, in a prestigious journal
- the problem must seem overly complex: trying to simplify a problem to a readable level gets your paper jettisoned quickly.
-this makes a PHD a nightmare, since your main job is to "defend" is to have a published paper(s) in "good" journals.
- in order to do this, your contribution "to the known scientific and technological knowledge" must be "significant".

The requirements of a PHD sound pretty vauge to me, heh. However, my "variable valve" idea has one advantage: there's little to no literature out there on the subject! Most info is a tightly held industry secret, as there are no commercial designs available.

4) Getting a PHD is a game of psychology. The hardest part is keeping yourself motivated enough to finish.

5) "Engineering" does not equal "research". Engineering is to implement/design based around an idea, based on scientific principles. Journal paper publishing research only overlaps engineering sometimes. :)

5) I learned truly "what I'm good at". Quite simply, I'm good at focusing on one thing at a time, and trying my best to finish it. Unfortunately, this makes me bad at most other things besides "nerdism" and "booksmartism" (like multitasking, or cooking a large multi-course meal).


Post a Comment

<< Home