Research On Function Optimization Based On Artificial Immune System And Its Applications In Complex System

A function optimization approach based on artificial immune system and its applications in intelligent control of complex system are researched. The dissertation consists of two parts.The first part: A function optimization algorithm based on the mechanism of clone and selection of B cells is presented. Functions optimization for multi-modal functions optimization and non-stationary are one of important research subjects of function optimization. Both of the problems are difficult to be solved by stochastic optimization algorithms in existence. Clone and selection mechanism of B cells is one of the important approaches to search antibodies, as well as one of important factors to generate antibodies diversity. It has strong optimization ability. In this paper, Clone Selection Algorithm (CSA) based on clone and selection operator for multi-modal functions and non-stationary functions was presented. Simulations were conducted. Results showed CSA had strong ability of function optimization and maintaining diversity. It was one of effective algorithms for multi-modal functions and non-stationary functions optimization.The second parts: Applications of AIS based on CSA in intelligent control. This part contained three subjects: designing and optimizing Fuzzy Logic Controllers (FLC) and Artificial Neural Networks (ANN), fuzzy neural network model identification of complex system. First, Fuzzy Logic Control and Neural Network Control are one of the main issues of the intelligent control. Usually, designing Fuzzy Logic Controller or Artificial Neural Networks depends on the designer's knowledge or experience. Their performance can't be assured. Through machine learning to design or optimize the two systems is an effective approach to solve such problems. In this paper, we discussed an approach using CSA to designing Fuzzy Logic Controller and Artificial Neural Networks. Simulation was conducted. Results showed: by this approach we could effectively realize extraction and optimization of Fuzzy Rules and its parameters, as well as designing and optimization structure and parameters of Artificial Neural Network. Next, Immune identification has strong adaptability and robustness. Referring to antigens-identification mechanism of immune system, an effective model identification approach for complex systems can be brought forward. Adaptive neural network model identification of uncertainty systems can be described as optimization of a non-stationary function. Because of the strong adaptability of CSA, in this paper, based on CSA, building blocks models constructing approach, mechanism of antibody's recognizing ball and fuzzy neural networks, a model identification approach for complex systems with uncertainty was presented. Simulations are conducted. Results showed: under the conditions of larger scale disturbance, fast online identification for fuzzy neural network model of complex systems could be realized by this approach. It has strong robustness and real time ability.