Fuzzy modeling and control: A characteristic-point approach
Abstract
This thesis proposes fuzzy inference systems (FISs) as a systematic method to model the input-output relationship of complex systems and their control. Characteristic points (CPs) which are a subset of given input-output data pairs can be used for establishing fuzzy rules of the system. With the CPs, a characteristic-point-based fuzzy inference system (CPFIS) is proposed to systematically construct FISs with two distinct features: maximum and minimum fuzzy rules which are related to the interpolation property of human reasoning, and their employment for the interpolation property, resulting in a small fuzzy-rule base for high-dimensional systems. Employing the CPFIS, a characteristic-point-based fuzzy control system (CPFCS) is proposed to control time-invariant, multi-input-multi-output plants. The CPFCS emulates the learning control of human operators in performing repetitive control operations with gradual improvement in the control performance. Besides the employment of CPFISs, the desired highest derivatives are synthesized to make the CPFCS perform as a generalized proportional-integral-derivative controller. If some conservative bounds can be established on the approximation errors of the constructed CPFIS, the controller can be further designed to have ultimate boundedness of the output tracking via a variable-structure-control method. A novel desired-highest-derivative-assignment controller, which is constructed on the forward-inverse-function-approximation error bounds, integrates fuzzy modeling and control to provide stability analysis of the overall system. Computer simulations were conducted on many examples to demonstrate the operations of the proposed CPFIS, CPFCS, and the integrated-fuzzy-modeling-and-control scheme.
Degree
Ph.D.
Advisors
Lee, Purdue University.
Subject Area
Electrical engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.