Steven R. Shaw - Courses, Teaching and Learning
EE321 - Control Systems I
EE321 is the first
in a two course series in control systems. EE321 begins with synthesis and analysis
of continuous time, single variable, feedback control systems using a transfer
function approach.
EE-521 - Nonlinear System Dynamics
EE-521 begins with an introduction to the history, mathematics, and notation
of dynamic systems. Some of the notation in the introduction will be different
from signals and systems courses. Autonomous state-space systems are the focus
rather than I/O transfer functions. The course will review some linear systems,
linear algebra, and introduce some ideas from dynamics, numerical simulation,
and linear algebra. The course finishes up with individual final projects.
The goal of the final project is to expose students to some part of the emerging
state of the art in nonlinear dynamics in engineering.
EE-525 -
System Identification
System identification can be approached from a number of directions. It
would be quite feasable to spend the semester discussing numerical methods useful
for system identification problems, classes of models, statistical implications,
experimental and measurement design, or the relationship between topics in system
identification and other areas in control. Instead, the course will adopt the
emphasis of the text, which is to concentrate on user choices and aspects of
system identification theory that are important for most practical scientific
and engineering problems.
EE-526 - Sequential State Estimation
EE-526 students will design optimal estimators to find asymptotically error-free
estimates of the internal state of a target system using one or more noisy sensors.
These estimators can be implemented with ease on any digital system, including
FPGA, DSP or general purpose processors. Students will consider three distinct
kinds of state estimation problems. In filtering applications the estimate uses
data received up to the current time. Smoothing prepares an estimate at time
k using some data after time k. Finally, prediction finds the state of a system
at a future time relative to the most recently received data. These techniques
are especially useful in real-time systems because they use data as it is received.
| Magnetic Suspension System for on-line control demonstration |
| The unconstrained suspension of a magnetic material in a field is an inherently
difficult problem. Open loop, the field can be designed to constrain the object
in two of three axes, for example, but the system is unstable in the third direction.
This instability can be overcome by the use of feedback control, in this case,
using a laser and detector. This single demonstration illustrates a number of
key properties of feedback control systems. The control modifies the dynamics
of the system, making an open loop unstable plant stable in closed loop. There
are numerous nonlinearities, but the control is designed and works well with
linearized plant model. The system is robust to changes in the plant, as illustrated
by the variety of objects that can be suspended, including spheres varying by
more than an order of magnitude in mass. We are constructing a new controller
with digitally programmable gain blocks so that students will be able to program
their control designs and test them on the real system via webcam. |
 |
Thermoset adhesive coated magnet wire was wound on an custom aluminum bobbin
held between centers on a lathe. The coil was baked on the bobbin to hold its
form. |
 |
Steel core, machined from a solid rod. A CNC mill was programmed to machine
the space for the wire by descending in a helical path. The coil was subsequently
potted in the core with epoxy to eliminate movement between core and coil under
control. |
 |
A 4-inch diameter ball bearing floats in the field. It is possible to attach
other, similar ball bearings to the bottom of this one. |
 |
An illustration of the robustness of the control. A proper classical control
design easily suspends a variety of objects, including this drill index. The
index has been taped shut by a sticker. Otherwise, the door would open, and
individual drills float out. Since the drills to not sufficiently occlude the
sensor, the system fails. |
| Induction coil accelerator |
 |
A group of undergraduates approached me with questions about a publically released
report from Sandia national labs, detailing simulations of an induction coil
accelerator for replacement of conventional artillary. An induction coil avoids
the detonation velocity limit of a conventional artillary piece. We formed a
special projects class and built two glass-fiber and linen composite induction
coil guns. The first was a three-stage unit, which accelerated a 1 inch diameter,
2.5 inch long cored aluminum slug to measured velocities in excess of 300 ft/s.
The larger version shown in the photo was never tested due to problems sourcing
the SCRs used for control. Peak coil current was about 9000A, the photo-flash
capacitors were charged to about 1kV using a transformer from a microwave oven,
and the control was open-loop. No students were permanently injured. |
| Tip-Tilt mirror control
|
 |
MEMS mirrors have a number of interesting control challenges that are a bit
different than the typical lab-scale, electromagnetic plants ordinarly encountered
in controls class. We are using this commericially available mirror to develop
controls lab exercises that give students hands-on experience with MEMS control. |
|
