New global optimization algorithms based on multi-loop distributed control systems with serial structure and ring structure for solving global optimization problems. (May 2021)

• Record Type: Journal Article
• Authors: Lu, Baiquan; Zheng, Zhongwei; Zhang, Zhijun; Yu, Ying; Liu, Tingzhang
• Is Part Of: Engineering applications of artificial intelligence. Volume 101(2021)
• Publication Date: 2021-05
• Subjects: Global optimization -- Neural networks -- Control system -- Transformation function -- Dimension reduction method -- Filled function method; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2020.104115
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: In this paper, new optimization algorithms using multi-loop distributed control systems with serial structure and ring structure are proposed for solving global optimization problems, where the control plant of the subsystem is replaced by the objective function of a given optimization problem. When the control plant of each sub-loop control system is known, some simplified methods of the control plant are first proposed. These approaches change a complex control plant into a simple function without changing global optimization solution to find the global optimization solution more easily by using multi-loop distributed control systems that has two kinds of serial structure and ring structure. When the control plant of each sub-loop control system is unknown, the objective function is identified by a neural network. In addition, a proposed special neural network as a local search rule is divided to m neural network subsystems by different sizes of the effective change interval of the transformation function. In other words, the smaller the index number of a subsystem is, the stronger the local search ability of the subsystem is, otherwise, the stronger the global search ability of the subsystem is. And the current best optimization solution between all neural network subsystems is used to guide each neural network subsystem. Simultaneously, a new filled function is proposed as a global search rule. It can jump out of a local minimum point and move to another local minimum point with smaller objective function value. Finally, 17 experimental examples show the effectiveness of the proposed method.