PROST-seminaret denne fredagen er ved Dan Rivera fra Arizona State
University som oppholder seg på friår i Linkøping.
-------------------------------------------------------------------
Tid: Fredag 08.nov. kl. 13.00-14.30
Sted: Lunsjrom, institutt for kjemiteknikk (Kjemiblokk 5, 2. etg)
Prof. Daniel E. Rivera, Arizona State University:
A Control-Relevant Methodology for Multivariable System Identification with
Application to High-Purity Distillation
-------------------------------------------------------------------
Pizza e.l. serveres kl. 13.00.
Foredrag begynner kl. 13.30 (45 min pluss 15 min diskusjon).
Alle interesserte er velkomne.
PROSTere oppfordres på det sterkeste til å møte!
-Sigurd Skogestad
--------------
SUMMARY:
A Control-Relevant Methodology for Multivariable System Identification with
Application to High-Purity Distillation
Daniel E. Rivera
The presentation will describe a control-relevant methodology for
identifying MIMO systems which has been developed at Arizona State
University. The methodology is characterized by three main stages:
open-loop excitation, consistent nonparametric estimation, and frequency
weighted parameter estimation. Identification data can be generated either
from a single open-loop MIMO experiment (all input channels excited
simultaneously) or from open-loop SIMO experiments (one input channel
excited at a time) using random inputs or periodic, deterministic inputs,
such as PRBS and Schroeder-phased signals. Consistent estimation of
frequency responses from the data is obtained either through DFT analysis
or from high-order ARX models.
The frequency-weighted parameter estimation problem is solved via an
iterative least-squares algorithm, resulting in parsimonious transfer
function descriptions that can be realized into low-order state-space
models. The models are thus suitable for modern predictive control schemes
that compute predictions directly from state-space models. Since the
weights in the parameter estimation problem explicitly incorporate
multivariable closed-loop requirements, the method is applicable to
demanding control problems such as those presented by highly interactive,
ill-conditioned systems. Using the Jacobsen and Skogestad and Weischedel
and McAvoy high purity distillation columns as examples, we demonstrate
that adequate models for control can be identified using the proposed
methodology without requiring a priori knowledge of gain directionality.
These results shed new insights into the modeling of highly interactive
systems using "black-box" identification methods.
----------------
BIOSKETCH:
Daniel E. Rivera
Associate Professor
Dept. of Chemical, Bio, and Materials Engineering
Arizona State University
daniel.rivera@asu.edu
Daniel E. Rivera is an Associate Professor in the Department of Chemical,
Bio, and Materials Engineering at Arizona State University and Program
Director for the Honeywell-ASU Control Systems Engineering Laboratory.
Prior to joining ASU he was an Associate Research Engineer in the Control
Systems Section of Shell Development Company. He received his Ph.D. in
chemical engineering from the California Institute of Technology in 1987,
and holds B.S. and M.S. degrees from the University of Rochester and the
University of Wisconsin-Madison, respectively. His research interestes
include the topics of system identification, robust process control,
control implementation, and the interaction between process design and
control.