Neural Control System Design for Chemical Processes

• Main Authors: Chen Wen Chih, 陳文智
• Other Authors: Chen Cheng Liang
• Format: Others
• Language: zh-TW
• Published: 1993
• Online Access: http://ndltd.ncl.edu.tw/handle/39388652647025907486

博士 === 國立臺灣大學 === 化學工程研究所 === 81 === This dissertation studies the design and implementation problems of neural controllers and investigates their control performance on chemical process control. A self--organizing methodology is introduced to provide the training set for adjusting parameters of the controller. One important feature of the proposed mechanism is that, though its should lack extensive knowledge of the process dynamics at the outset of controller design, it will still be able to achieve its desired results by means of employing the subjective experience of control specialists as its training aids. A hierarchical, multilayered neural network to provide parameters for a nonlinear PI controller in response to local operating conditions. Since a multilayered neural net is used in aiding the controller''s adaptation mechanism, no specific a priori structural relations need be assumed between controller parameters and plant local conditions. A two-steps procedure is proposed for using neural networks with the Nonlinear Internal Model Control structure. The first step is to train a neural network to represent the plant response. The following step is to select the controllers as the plant inverse model (direct controller), or to solve the nonlinear operator equation (indirect controller). The error signal is integrated in the direct control system, and is then used to eliminate the steady- state offest. The design procedure for an optimized piecewise linear fuzzy controller (PLFC, a direct controller) by using neural network techniques is proposed. In the design procedure, the simplified PLFC is used as the initial controller structure, then a GPFNC, which gives the approximate control response to the initially given PLFC, is found for further optimization. The optimized GPFNC could finally be converted into its fuzzy counterpart. The resulting fuzzy controller could give a more structural interpretation of the intelligent control strategy.