| As a new generation natural computation method, artificial immune system (AIS) is a computing system for resolving all kinds of complicated problems, which is inspired by biological immune principle. AIS has been used in optimization, fault diagnosis, control and data processing and so on. It is becoming another research hotspot in the natural computation after artificial neural network, fuzzy system and evolutionary computation.AIS has two key problems, which are intrusion detection of system and optimization of internal learning principle of system. Performance of AIS can be improved if it has suitable computing model and algorithm.The research work of this dissertation belongs to the cross-frontier research field of life science, information science and computing science. Based on immune response principle, AIS is studied about resolving optimization problem and fault diagnosis problem in this dissertation. Research achievements of this dissertation can provide new ideas and methods for AIS. Furthermore, it plays important role in real optimization problem and mechanical fault diagnosis. The main work and innovative achievements of this dissertation can be concluded as follows: (1) Built a two-layer model framework of immune response on the basis of immune response process of biological immune system. According to mapping relation between biological immune system and artificial immune system, computation model framework of immune response was put forward. This model framework included two model frameworks. One is artificial immune response optimization model framework (AIROMF). The other is artificial immune response fault diagnosis model framework (AIRFDMF).(2) Proposed an artificial immune response optimization algorithm (AIROA) on the basis of AIROM. A series of operators and their calculation were defined. AIROA divided initial antibody population into memory antibody population and temporary antibody population with the affinity value. Different operators were executed to the two antibody populations as mentioned above for local searching and global searching with variable-sized domain. Searching precision in the later period of evolutionary was improved and accelerated. AIROA was used to optimize standard testing functions, and the results shew that it has faster convergence and higher precision and enhances the performance of CSA.(3) Put forward three path planning methods of mobile robots in the dissertation. Basic artificial potential field algorithm (APF) often gets into target and not reachable, so a new mobile robot path planning method based on APF and genetic algorithm (GA) was put forward. GA has slower convergence and gets into local minimum, so a mobile robot path planning method based on artificial immune potential field was put forward. Based on AIROA and APF, a path planning method of mobile robot has quicker convergence and higher searching precision. It can make mobile robot safely avoid obstacles and quickly arrive the target.(4) Presented a variable-sized detector based artificial immune clustering algorithm (VDICA) on the basis of AIRFDM in the dissertation. VDICA includes the production of initial detectors and clustering learning of detectors. VDICA uses Monte Carlo method to estimate detectors space cover, so it does not need determine the numbers of detectors. Furthermore, it sets the maximum radius of detectors in order that detectors have better distribution characteristic. At last, algorithm parameters are determined by experiment.(5) Combining fault type field marker method, a VDICA based fault diagnosis method was put forward and applied for machine gearbox. A VDICA-RBF neural network based fault diagnosis method was put forward and applied for diesel engine. The experiment results showed that VDICA has better ability of detectors space cover and the fault diagnosis method based on VDICA which can realize fault diagnosis in complicated fault mode and has higher accuracy and quicker convergence than that of RBF.
|
PDF Full Text Request