Research On Immune Computational Intelligence And Its Application To Complex Systems Optimization

The biological immune system is an evolutionary system, which is highly adaptive, highly distributed and self-organizing, contains many mechanisms of information processing. The Artificial Immune System (AIS) is an intelligent information processing technology that is based on the mechanism of the biological immune system. As a novel branch of computational intelligence,AIS has strong capabilities of pattern recognition, learning and associative memory, it is a powerful information processing and problem-solving paradigm in both the scientific and engineering fields. In recent years, developing computational intelligence and devoting in the studyingof the theory and application is becoming one of the focuses.Under this background, an effective intelligent optimization strategies based on the biological immune principle are developed in this dissertation. Aim at the difficulties in the field of optimization, machine learning and complex system intelligent control, numerical simulations and algorithm comparison study are carried out. Simulation and application indicate all this methods obtained are feasible and effective, which have enriched the contents of the artificial immune system.The main work in the dissertation can be summed up as follows:(1) Based on the immune response principle in the biological immune system, an immune intelligent computing framework and its mathematic model are carried out. This model and its basic algorithm have the ability of self-learning and self-adaptation, which convergence reliability and convergence velocity is discussed based on the Markov probability theory, then the algorithm parameter and complication degree is analysized.(2) By introducing a mechanism of cluster and competition in the clonal selection process, a novel immune cluster and competition clonal selection algorithm is proposed. The cluster operation divides the population into subpopulations for the stage of selection and reproduction, thus improves the variety of antibodies and affinity maturation. Aim to quickly obtain the global optimum and local optimum, adopts a hybrid hyper-mutation operator. These mechanisms enhanced the variety of antibody and affinity maturation. The simulation results illustrated that the efficiency of the proposed algorithm for complicated function optimization, verified it has a remarkable quality of the global and local convergence reliability. (3) Combined the immune optimization and machine learning, aim at the problem of large scales of attribute reduction, a prior knowledge of attribute kernel as bacterins is introduced to antibody coding and the population is vaccinated in a stochastic way. The approximation quality as affinity evolution objection to optimized. This algorithm applied a niche sharing fitness method in clonal selection process and introduced the ideas of compete expansion, clonal elimination, antibody supplement. Through those operators, the diversity of antibody and affinity maturation rate is enhanced. Moreover, it solved the problem of quickly obtain minimum reductions and more reductions of the rough sets. Experimental results illustrate that the approach is an effective and quick way in solving attribute reduction.(4) Applying the immune optimum to intelligent control, the fusion study of immune system and intelligent control is proposed. Aim to the difficulties in the field of fuzzy control, a scheme of optimization design for fuzzy controller is proposed based the immune algorithm. Conduct the optimization on the membership function and the control rules of the fuzz controller respectively. Based on the immune fuzzy self-adaptive and self-learning capability, an immune fuzzy pid controller is designed. Finally, the dissertation applys immune optimization intelligent control method to complicated industry process control, a fuzzy and imuune pid hybrid controller is presented. The output of the hybrid controller synthesizes the ones of fuzzy and immune pid controller by a coefficient of weight which decides which controller can work at one time depends on the input error. Then a scheme based the integrated immune fuzzy control is bringed forward and applied to rotary kiln temperature control process. Experiment results show the feasibility and effectiveness of the control method and indicate the fusion study of immune computational intelligence and the control system is a beneficial complement to modern intelligent control method.