Danger Model Immune Algorithm And Its Application Research

In recent years, the development of artificial intelligence is so fast, and artificial immune algorithm is one of the hot technologies. Danger model theory is a new theory of immunology, and it breaks a new path for immunology research. At the same time, the theory has a profound influence to the artificial immune algorithm. To solve complex engineering problems, the artificial immune algorithm has been constructed based on danger model theory by some experts. But there still exit some theoretical and application problems such as the basic framework of danger model immune algorithm design, and the convergence analysis of the algorithm. Therefor, we do some researches on these hot technologies in this thesis. It is of great value for the further studies of danger model immune algorithm. We use genetic algorithm and conventional artificial immune algorithm for reference, and construct an artificial immune algorithm based on danger model theory. We have done some researches in theory, and combined the practice of ship safety, and studied the application of danger model immune algorithm in ship collision avoidance optimization. In the end, we do some simulations to testify the algorithm.Firstly, this thesis starts with the medical basis of immunology, and elucidates the medical mechanism of danger model immune algorithm. The danger signal, danger area and danger operator are defined according to the characteristic of danger model theory. The danger signal and danger area are the most important and basic conception in danger model theory. Antigen-antigen affinity is the basis of danger signal generation. The super-sphere of the best antibody in current iteration is defined as danger area. Use the designing method of genetic algorithm and artificial immune algorithm for reference, the basic framework of danger model immune algorithm is proposed through the cooperation of danger signal, danger area, danger operator, mutation operator and selection operator. In the simulation studies, taking some complex functions optimization for example, and the simulation results demonstrate that the danger model immune algorithm is valid.The antibody population of danger model immune algorithm is a markov chain, so that the convergence of algorithm is proved strictly according to the mathematical properties and theorems of markov random process. It settles a theoretical basis for algorithm design and application.The basic framework algorithm is the simplest realization form of artificial immune algorithm based on danger model theory. Some flaws in the basic framework algorithm still need to be improved. The size of danger area is fixed. So the selection of the danger area radius is experiential or decided by trial-and-error method. To solve the danger area selection problem, the danger area adaptive danger model immune algorithm is proposed. In the amelioration algorithm, the danger area radius is adaptive adjustment with the increase of iteration. At the beginning of the iteration, the danger area radius is commonly larger. In order to search the global optimum value better, the danger area will cover the whole variant field of the problem. With the increase of iteration, the danger area is reduced continuously in order to realize local searching for the important region. The simulation results demonstrate that the optimum value searching ability of danger area adaptive adjustment algorithm is better than the basic framework algorithm in convergent speed and the precision of optimization result. Chaos system is a deterministic system like a random system. Chaos variable has the characteristic of randomicity and ergodcity, which guarantees a good searching ability in large scale and doesn't limit by local extremum. And the chaos variable is effective to avoid premature phenomena and enhances the convergent speed. Besides it doesn't need the optimization problem with the continuity and differentiability. So combining danger model immune algorithm with the chaos theory, a chaos danger model immune algorithm is proposed. In the iteration process of the algorithm, the danger antibody set and safe antibody set are operated by small chaos perturbation and chaos re-generation technology, respectively. This will ensure the local searching ability for danger area and global searching ability of the algorithm. The simulation results demonstrate the validity and superiority of the algorithm.In order to apply the danger model immune algorithm to ship safety field, this thesis studies the mechanism of ship collision avoidance, ship domain, and the calculation methods of ship collision risk and ship movement parameter which settles a basis for ship collision avoidance simulation. For ship collision avoidance strategy optimization problem of ship safety field, a ship collision avoidance strategy optimization algorithm based on danger model immune algorithm is proposed. According to the principle of ARPA trial test, the concerned parameters of ship collision avoidance operation are encoding as the antibody string of danger model immune algorithm; the antibody is evaluated by the affinity of antibody; and establishes an iteration process to search the optimal collision avoidance strategy by updating the population. Finally, the simulation results demonstrate that the algorithm is valid.In summary, a framework of danger model immune algorithm is designed in this thesis according to the new idea of immune response in danger model theory, some improved algorithms are discussed, an application of ship collision avoidance strategy optimization is studied, and the convergence of the algorithm is proved in theory at the same time.