Dr. Mark A. Hopkins, Professor

loading Hopkins' picture ...

Dr. Hopkins in 2014.

E-Mail: mark.hopkins@rit.edu

PZGui Toolbox for Matlab     (Pole/Zero Graphical user interface)

This Semester's schedule

This Semester's Classes and Office Hours

Office: Gleason Bldg (09), Room 3021

Phone Number: (585) 475 - 6640

Education and Interests

Publications, Patents, etc.

Education and Interests:

Degree:
Ph.D., 1988, Virginia Tech (Virginia Polytechnic Institute and State University), Blacksburg VA
Specialties:
System Identification
Control Systems
Research Interests:
-----
Model identification of flexible structures: Typically, the order of the model to be identified is at least 600, because the system it represents has considerably more than 300 (often closely-spaced) lightly-damped vibrational modes that are distributed over a frequency range of at least 4.5 orders-of-magnitude, with a dynamic range of about 120 dB. The identified model ideally has bode magnitude error less than 2 dB and bode phase error less than 10 degrees at every frequency over the full range of interest (when compared to the empirical data from the real system). This is a very tough problem, particularly the very wide bandwidth, not only in terms of system identification, but also in terms of numerical stability. I made very good progress on this modelling problem in the late 1990's in work at Eastman Kodak Company, Commercial & Government Systems Division (currently Exelis Inc.), and developed a technique, called the Pick And Place Algorithm (P.A.P.A.), to create models this good.
-----
Extending the P.A.P.A. modelling technique to MIMO systems: During a sabbatical (2001-02) at C&GS, I successfully extended P.A.P.A. to multiple-input, multiple-output (MIMO) systems. The crux of this problem is converting a matrix of transfer functions into a state-space model in modal-canonic form, by way of partial fraction expansions. Getting this to work involved creating a new algorithm, called Residue Rank Reduction (R-cubed), that iteratively reduces the rank of all the MIMO Cauchy-residue-arrays to rank one. That is necessary in order to avoid having any multiple poles at the same location in the state-space model. The feasibility proof of the method is a two-input, two-output, 850-pole state-space model of a large flexible structure. This model was created from four P.A.P.A. transfer functions, and has extremely good frequency response accuracy from 0.2 Hz to 10 kHz.
More recently (2009-12), I have solved the problem of creating very high-order discrete-time MIMO state-space models that agree quite well with empirical data over a very broad range of frequencies ( > 4.5 orders-of-magnitude). Over those years, and particularly during a 2011 sabbatical, I developed a method that successfully decouples individual modes, to the greatest extent possible, so that they can more easily be identified. This led to a very successful new modelling technique that I call FORCASTER (Frequency Observability Range Context And Subsystems To Estimate Residues). This new method relies on three fundamental tools, (1) The FORSE algorithm, published in 1996 by Liu, Jacques, and Miller in ASME JDSMC, (2) the mode-decoupling technique mentioned above, and (3) other linear model components that, collectively, comprise the "subsystems" that model the effects of non-modal poles (e.g., integration, delay) and modes above the Nyquist frequency. Some colleagues at Exelis Inc. have dubbed this method "FORSE-Fed P.A.P.A.", a name that has some merit. These results were published in 2013 at the SPIE International Symposium on Smart Structures and Materials, in San Diego, CA.
-----
Control of flexible structures: Apply advanced control techniques to high-order models generated by the model identification. This problem involves model-order reduction techniques, as there are significant numerical problems involved in designing a controller from models larger than 600-th order, and no possibility of implementing in hardware a real-time controller of that size at fast sample rates. In the past several years, I have been developing a method that tunes an initial LQR-based controller/observer design to achieve damping, isolation, and robustness goals, for the class of infinite-dimensional, reasonably linear, flexible structures that are my main interest. This method relies on having excellent high-order as well as reduced-order models, and uses the Nelder-Mead Simplex Method, and other similar algorithms, to tune the two feedback matrices. It also depends upon distributed computing to make it feasible to achieve good results in a reasonable amount of time. The initial work was funded by ITT Geospatial Systems (now Exelis Inc.).

Go back to the top

 

Dr. Hopkins will be off-campus during Summer 2014

 

Courses Dr. Hopkins will be teaching in Fall Semester, 2014-2015 (August 25, 2014 to December 10, 2014):

EEEE 414, Control Systems Design (8:00 - 9:15 am, Tu Th)
EEEE 661, Modern Control (5:00 - 6:15 pm, Tu Th)
EEEE 105 (Sec 5), Freshman Practicum (2:00 - 2:50 pm, Th)

Office Hours in Fall Semester, 2014-2015 (August 25, 2014 to December 10, 2014)

.

Go back to the top

Publications, Patents, etc.

Publications:

Patents:

Professional Affiliations:

Go back to the top

MATLAB  Toolbox Shareware to Download

PZGui (Pole/Zero Graphical user interface)

A MATLAB Add-On Tool for the Study of Linear Systems and Control

This is the ultimate interactive tool for studying SISO transfer-functions.

More than 50,000 lines of code developed over a period of 18 years.

Now in a major new release, it is Matlab 2014b Graphics Ready.

       Currently in Version 8.0.39   released September 15, 2014.

( Runs in   MATLAB Version 2008a or higher, Student version, or Professional version with Controls Toolbox
            THIS IS SHAREWARE -- FREE, IF USED FOR EDUCATIONAL PURPOSES ONLY !!
  

Go to download page for zipped PZGui M-files

    This highly interactive, GUI-based program helps students and engineers to comprehend the complicated relationships among the following:

    Completely updated comprehensive 55-page User's Manual.

    Includes special tools to study PID and Lead-Lag design.

    All features are available both for continuous-time and discrete-time systems, enhancing the ability to study relationships between them.

    Correctly handles models that have hundreds of poles and zeros.

    New Feature: Generate large random "flexible-structure-like" models.

The program consists of more than 65 MATLAB M-files,
and a 55-page PDF User's Guide.

            Click here for more info about PZGui (the User's Manual)
Go back to the top

e-mail Mark Hopkins: mark.hopkins@rit.edu

This page was updated 15-Sep-2014