Intelligent multivariable control system
Abstract
The objective of this study is to develop an intelligent controller for general nonlinear multivariable systems and implement it in complex grinding processes. Conventional control techniques require a precise analytical model. Fuzzy controllers offer a promising alternative as the system dynamics is captured by human knowledge and experimental data rather than by an accurate analytical model. In this work, a Multi-Level Fuzzy Controller (MLFC) is developed for a general Single-Input Single-Output (SISO) nonlinear dynamic system. The system uncertainties and time-varying parameters are compensated by the embedded adaptation mechanism. The closed loop system is guaranteed to be stable with input-output passivity analysis. Simulation examples are carried out for three uncertain nonlinear systems and it is shown that much better system performances are achieved compared with the conventional fuzzy logic controllers (FLC), even in the presence of unknown system uncertainty and disturbances. And then this controller is implemented for constant cutting force control of a creep-feed grinding process and an end-milling process. Next the fuzzy controller is extended to general Multi-Input Single-Output (MISO) nonlinear systems. The plant dynamics is modeled in fuzzy domain with triangular input membership functions, singleton output membership functions and fuzzy-mean defuzzification. Next, the fuzzy inverse model is automatically constructed from the obtained fuzzy model and then the intelligent controller is designed based on the fuzzy inverse model. The controller has a hierarchical structure for each control variable to compensate for the system uncertainties and time-varying parameters. The stability of the MLFC-MISO controller is proved and simulation examples are presented to illustrate the effectiveness of the proposed controller for multivariable systems under various system conditions. At the end, an intelligent fuzzy control system is developed for a general Multi-Input Multi-Output (MIMO) system, which could eliminate the multivariable interaction effect and achieve desired system performance for multiple outputs simultaneously. A novel multivariable input-output interaction analysis method is proposed based on a set of Fuzzy Basis Function Network (FBFN) models. The Relative Gain Array (RGA) is constructed to represent the multivariable interaction degree based on system steady-state gain and then integrated into the multivariable fuzzy control design.
Degree
Ph.D.
Advisors
Shin, Purdue University.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.