ELECTRICAL
& COMPUTER ENGINEERING
o
The thesis investigates and compares methods for
modeling and controlling non-linear systems in which the friction in the system
cannot be sufficiently modeled linearly and classical control methods cannot
provide adequate performance especially when load variations occur. The plant
was a small robot arm system from Quanser Consulting.
A real-time non-linear model was developed based on Stribeck
friction. System performance was measured and compared for conventional PID
control, static and adaptive feed-forward controllers, and an optimal control
method. The research was supported by a Caterpillar Fellowship Award. Mike
completed the MSEE degree in August 2007. He is a research engineer with the
Systems and Controls Research Department at Caterpillar in
o
PID tuning methods for closed-loop position control of a small robot arm
were compared with model-based approaches using root locus and frequency domain
design methods. Model-based methods included Internal Model Control (IMC), gain
scheduling control, and feed-forward (FF) compensation. An ADALINE was used for
plant system identification. The converged ADALINE was used as a plant model
and inverse plant in the IMC and FF methods.
o
The Atmel AT90CAN128 8-bit microcontroller with a built-in CAN
communication bus was selected as the platform for the low-cost system. Sensor interfacing was a primary objective as
well as a communication interface to MATLAB. Stephen completed the MSEE degree
in May 2007. He is an electrical design engineer with Harris Corporation in
o
The objective of the project was to implement multiple 980 Medium Wheel
Loader Implement Features within the Simulink/Stateflow/Matlab environment for
the Catalyst Program within Caterpillar Inc. Mike completed the MSEE degree in
December 2005. Mike is a lead engineer in the Core Electro-hydraulic Program at
Caterpillar Inc. in
o
The research objectives were: (1) develop force
control capability for a backhoe loader test bed, (2) investigate accuracy
using cylinder pressure and displacement sensors, (3) develop forward and inverse
force calculations, (4) develop a linkage and hydraulic model in SIMULINK, and
(5) develop and tune force control algorithms. The research was fully supported
by Caterpillar ($75,000). Joe completed the MSEE degree in July 2005. He is
employed with the Systems and Controls Research Department at Caterpillar in
o
The objective of this project was to design and
implement a proportional and proportional plus integral controller for a
Pittman DC motor in a velocity control application. The project involved analysis of theoretical
data obtained from MATLAB/Simulink and experimental
data obtained from the implementation of the code developed. The controllers
used a linear power amplifier to drive the motor and a frequency to voltage
converter as the feedback mechanism. Brett is employed with the Signature and
Antenna Technology Group with Boeing Corporation in
o
The research objectives were: (1) develop a
neural network algorithm for voice recognition in a home automation
application, (2) implement the best neural method in real-time on a personal
computer using C++, (3) implement an interface from the neural network to the
X-10 home automation data standard, and (4) use the X-10 interface to control
electrical appliances. Kevin completed the MSEE degree in August 2002. He is
employed with Kass Electronics (home automation) in
o
The research objectives were: (1) compare linear
controller algorithms using frequency domain design methods, (2) developed
static nonlinear controller (software lookup table) for hydraulic application,
and (3) apply control methods to an electromechanical system and to an hydraulic transformer application. The research was fully
supported by Caterpillar ($27,256), and the Graduate School/College of
Engineering ($9400). Megan completed the MSEE degree in May 2002. She is
employed with the Systems and Controls Research Department at Caterpillar in
o
The research objectives were: (1) develop and
compare standard LMS algorithm with (2) normalized LMS, (3) correlation LMS, and (4) search-then-converge LMS for noise cancellation
architectures. Tom completed the MSEE degree in December 1998. Tom is employed
with the Engine Systems Department at Caterpillar in
o
The research objectives were: (1) experimental
and theoretical system identification of robot arm mechanism, and (2)
conventional controller design. The research was partially supported by a
Caterpillar Fellowship Award. Todd completed the MSEE degree in August 1998 and
is currently employed with the Engine Systems Department at Caterpillar in
o
Jeff started this research for his undergraduate
senior project and completed a hardware implementation of an analog neural
phase-locked loop (PLL). In addition, his research has included: (1)
theoretical and experimental comparisons between neural and conventional PLLs, (2) investigation of neural and conventional feedforward compensation methods, (3) investigation of
anti-windup compensation methods, and (4) assistance in supervising an
undergraduate senior project on a microcontroller-based neural PLL system. The
research was partially supported by Bradley research awards (
o Jeff and Simulink Design Package
o
Ashley is an electrical engineer with The John
Deere Company in
o Sophia graduated with a MSEE in August 1996 and is currently an electrical engineer with Motorola. Sophia started this research as a project in my artificial neural network class (EE 691). The research objective was to compare a linear digital signal processing (DSP) noise cancellation system with a neural network method. The approaches were implemented on a Texas Instruments DSP chip. Sophia's results showed that a single nonlinear neuron yielded approximately the same results as a 100 tap FIR filter.
o Ramireddy graduated with a MSEE in May 1995. The research objective was to design a neural network to detect circuit faults. PSPICE data was used to train the neural network. An analog and digital circuit were used for neural network development and training. The results showed that neural methods are better suited for subsystem circuit testing rather than individual circuits components.
