#### Date of Award

Spring 2014

#### Document Type

Dissertation

#### Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)

#### Department

Electrical and Computer Engineering

#### First Advisor

MengChu Zhou

#### Second Advisor

John D. Carpinelli

#### Third Advisor

Edwin Hou

#### Fourth Advisor

Abdallah Khreishah

#### Fifth Advisor

Zhiming Ji

#### Abstract

There has been no known research that applies nonlinear transfer function to a nonlinear control problem. The belief is that nonlinear systems have no transfer functions. The Laplace transformation required to define transfer functions is not tractable mathematically when the coefficients of the differential equation are functions of state, output and control variables. In other words, it is not defined for systems that do not obey principles of superposition. Only linear systems obey this principle. Therefore, this dissertation work represents the very first research to demonstrate how transfer functions can be used to represent and design feedback control for nonlinear systems.

Real systems are inherently nonlinear. A few important examples include an aerospace vehicle whose mass parameter is variable because of fuel consumption, artificial pancreas and HIV drug delivery systems in the bio-medical field, robot arm and magnetic levitation systems in the mechanical engineering field and phase-locked-loop in the electrical engineering field. The subject of nonlinear system control, however, is more of an art than science. There is no unified framework for analysis and design. Success of a design usually depends on a designer’s experience. All the theory and design tools available, e.g., the whole subject of linear algebra, are based on systems described with linear models, which obey the principle of superposition. Control system design by linearization, which is based on approximated linear time invariant (LTI) system design model, is the closest to a general design framework available for nonlinear systems.

The most important problem in a control system designed by linearization is the problem of design model parameter variation during its operation. Obviously, this problem is the result of assuming a constant parameter or LTI design model for a real system that is actually nonlinear or has variable parameter model. In other words, a real system does not have constant parameters as approximated by its LTI design model. This problem is important enough to have specific design methods such as robust control and Horowitz quantitative feedback theory developed to address it. As the system is operated further and further out of the approximate linear range this problem gets worst. Furthermore, the controller based on design by linearization is not a tracking controller. It is a regulator that usually cannot track a varying reference input.

Investigated in the research presented in this dissertation is a nonlinear transfer function-based control method, i.e., one based on a model represented with varying parameters therefore a natural solution to the model parameter variation problem of design by linearization. The class of applicable nonlinear and time-varying systems are those that are affine in their control input such that they can be described by the central concept of this scheme, a state-dependent transfer function (SDTF). The introduction of this concept of nonlinear transfer function design model and the feedback control scheme based on it are the contributions of the research presented in this dissertation.

#### Recommended Citation

Kwadzogah, Roger Kobla, "Modeling, control and simulation of control-affine nonlinear systems with state-dependent transfer functions" (2014). *Dissertations*. 164.

https://digitalcommons.njit.edu/dissertations/164