Nonlinear model predictive control of a cutting process

作     者：Potocnik, P Grabec, I 

作者机构：Univ Ljubljana Fac Mech Engn SI-1000 Ljubljana Slovenia 

出 版 物：《NEUROCOMPUTING》 (神经计算)

年 卷 期：2002年第43卷第1-4期

页      面：107-126页

核心收录：

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：nonlinear model predictive control neural networks genetic algorithms cutting process 

摘      要：Nonlinear model predictive control (MPC) of a simulated chaotic cutting process is presented. The nonlinear MPC combines a neural-network model and a genetic-algorithm-based optimizer. The control scheme comprises a process, a model, an optimizer, a controller and a corrector. Neural networks are used to build a nonlinear experimental model of the process which is applied to recursive prediction in MPC, A robust genetic-algorithm-based optimizer is used for the optimization of control trajectories, A neural-network-based controller is included in the control scheme for enhanced optimizer initialization and for autonomous control after the learning period. The nonlinear MPC is applied to control the simulated chaotic cutting process, The dynamics of a cutting process are very complex due to the nonlinear effects of high order involved. The control objective is to construct an on-line control system capable of improving the quality of the manufactured surface by preventing tool oscillations which result in the rough surface of the workpiece. A feedforward network is applied as an experimental model of the cutting process, and MPC strategy with tool support manipulation as a control variable is investigated. The results show considerable improvement of the manufacturing quality obtained by the proposed nonlinear model predictive control. (C) 2002 Elsevier Science B.V. All rights reserved.