Research of anomaly detection method based on improved artificial immunity

According to the abnormal state detection without enough priori knowledge and fault samples, a new detector generation method with step radiuses is proposed. The method separates abnormal state space into several regions, and then sets different detector radius for them according to actual situations, has great flexibility and higher coverage rate to space. Artificial immune real-valued negative selection algorithm based on this method is applied to find the model for abnormal state detection, and later on be tested and analyzed in two different cases to some bearing using the model founded above. One with the fault samples and the other not. A conclusion is reached from the test that the method mentioned above can detect the known and unknown fault efficiently, and the correct detection rate is satisfactory.