Research Of Fault Diagnosis Based On Artificial Immune Algorithm

In this paper, two fault diagnosis methods based on Artificial Immune Algorithm are researched, and applied to fault diagnosis of the wind turbine. The following is the contents of this work:A Variable Sized Detectors Based Immune Clustering Arithmetic (VDBICA) is presented in this paper. The main innovations of this new algorithm are as follows:First, utilizes the Monte Carlo methods for the estimates of the coverage of the detectors set and use this coverage as a control parameter. Second, set the maximal size of radius for the detectors in order to make the algorithm has better clustering results than before. Third, define the system mode eigenvectors obtained under the normal working condition of the equipment as the 1st kind of antigen, and define the eigenvectors get from the abnormal working condition as the 2nd kind of antigen. And this algorithm developed a new detector clustering algorithm which takes the 2nd antigen as the clustering center, and gives the ranges of some key parameters and the gives the analyzing of feasibility. In the state space of the system, we use the result of immune clustering and faults information knowledge to mark the zones corresponding different faults. Finally, apply the marked results to the fault diagnosis of gear cases in wind turbines.A fault diagnosis method based on the Immune response mechanism is proposed. Clone selection and hyper mutation are introduced besides genetic operator to keep the population’s diversity and overcome the prematurity; meanwhile the diversity of memory antibodies is ensured. This paper propose the idea that using multi-antigens to generate memory antibodies and proposes the strategy that stimulate and suppress the antibodies which are generated by different antigens belong to the same fault pattern based on the concentration. This can avoid the phenomenon of immaturity convergence and can make the memory antibodies express structures and characters for more antigens. Besides, k nearest neighbor method based on threshold is designed in this paper. Finally, apply this fault diagnosis research to the fault diagnosis of wind turbines, the result show that this method has higher accuracy and quicker convergence.