Basic Research On Artificial Immune Algorithm And Its Application

Immune system, genetic system, as well as among the nervous system are provided with large-scale parallel information processing capability, and powerful learning ability, memory capacity, identifying capacity, adaptability and robustness, self-organization capabilities and the ability to maintain diversity.Premature convergence, together with easily stick to local optimum are fatal flaw of standard genetic algorithm. The main solution to this problem is to adjust selection pressure through fitness scale. But it is not versatile because the fitness scale depends on the problem.Through learning from the immune system of vertebrates, we obtain some enlightenment to the solution to the problem:Aiming at the demerit of Genetic algorithms, immune concentration adjustment mechanism is introduced to the genetic algorithm, thus immune genetic algorithm is formed. The Algorithm uses the two fitness and concentration indicators to evaluate the evolution of an individual (antibody), which effectively regulate the selection pressure to maintain the diversity of the groups and overcome the weakness of premature convergence and improve the quality of the solution. On the basis of existed research of immune genetic algorithm, a definition to the similarity of antibodies based on percentage is proposed, named improved immune genetic algorithms. The improved algorithm increases computing speed and overcomes premature convergence.Clonal selection theory is used to explain how the biological immune system to eliminate foreign antigens mechanism. Clonal selection algorithm uses the strategy of breeding and mutation in proportion to fitness. Through searching around the space, local optimization and global optimization are achieved. Aiming at the weak multi-optimization of clonal selection algorithm, on the basis of the study of classic clonal selection algorithm, an improved clonal selection algorithm is proposed. The simulation result shows the effectiveness of the improved algorithm. Based on "clonal selection, negative selection, immune network" and other immunological principle, artificial immune network simulates the stimulation process between the immune network and antigens. It is used to solve the problem of data clustering. Clustering problem will be regarded as a multi-optimization problem, thus a multi-function optimization algorithm is put forward. The algorithm, with excellent features such as automatic adjustment of the number of groups, real coding and others, can effectively extract the vast majority of the objective function local peak. But there will be premature sometimes. Aiming at the demerit of premature convergence, on the basis of the study of artificial immune network, an improved artificial immune network algorithm is proposed. The simulation result shows the effective solution to premature.Proportional-Integral-Derivative controller has been widely used in the industrial field for its simple algorithm, strong robustness and high reliability, which can provide higher performance/price ratio than other controller. The performance of traditional PID parameter tuning has greatly reduced when it was used in the nonlinear, time-varying, and the coupling parameters and structure of the complex process of uncertainty. By applying immune algorithm (immune genetic algorithm, clonal selection theory) to the parameter optimization of PID controller, the superior performance of the PID controller parameters off-line is acquired.