Non-linear Control Via Automatic Differentiation
by
Quetzalcoatl Magana
Case Western Reserve University (CWRU) Cleveland, Ohio
Coauthors: Coleman B. Brosilow (Department of Chemical Engineering), Kennneth A. Loparo (Department of Systems Engineering), Marshall J. Leitman (Department of Mathematics at CASE WESTERN RESERVE UNIVERSITY)

ABSTRACT Chemical processes present special control problems and offer special opportunities due to their inherent non-linearity. The use of a linear controller often limits the range of operation to restricted feed-stocks (e.g. fluid catalytic cracking) and limits the speed of recovery from process disturbances.

Non-linear controllers which invert the process model require the computation of many Lie derivatives. These derivatives can be computed analytically and symbolically only for relatively simple non-linear systems. This work aims to substantially simplify the design and maintenance of non-linear controllers for highly non-linear processes through the use of automatic differentiation. The ADIFOR2.0 code (Automatic Differentiation of Fortran 77 programs) appends derivative code to computer codes for function evaluations, handle branches, loops, subroutines, and computes numerical values of derivatives within machine precision. We modified ADIFOR 2.0 software to compute the necessary Lie derivatives. Furthermore, we also developed a MATLAB5.3-ADIFOR2.0-FORTRAN6.0 package to calculate the feedback control laws for a class of non-linear systems.

Date received: January 27, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cads-50.