Research And Application Of Fault Diagnosis Based On Artificial Immune Algorithm

The fault diagnosis technology, which can early detect equipment failure, early maintain equipment and minimize the economic losses and personal injury from the failure, meets the reliability and security needs of equipment. The traditional fault diagnosis technology needs the samples of high quality and is lack of self-learning. This paper presents two methods of fault diagnosis based on artificial immune algorithm. The methods apply to fault diagnosis of gear. The concrete research content is as follows:(1) A fault diagnosis method based on improved artificial immune algorithm is put forward. The general artificial immune algorithm is optimized by genetic operator such as selection, crossover and mutation and by immune operator such as vaccination and immune selection. By analyzing the fault sample data, the standard deviation of the same attribute data is used as weight of Euclidean distance. The improved Euclidean distance is used as the affinity of artificial immune algorithm. The improved artificial immune algorithm can train the efficient super-ellipsoid detector. A fault diagnosis model based on improved artificial immune algorithm is build. The classification effect of this model has more obvious advantages than that based on no improved artificial immune algorithm by test of the Iris data set. The feasibility and accuracy of this method is verified by application of gear fault diagnosis.(2) A fault diagnosis method based on negative selection algorithm(NSA) is designed. The advantages and disadvantages of native negative selection algorithm(NNSA), real negative selection algorithm(RNSA) and real valued negative selection of variable radius detector(V- detector) are analyzed. Then V- detector algorithm was improved. At the start of the training, the detector of the radius as large as possible is found through the Monte Carlo method. Then cover threshold are introduced, so the detector number is effectively avoid increasing, at the same time, reducing the number of black holes, and increasing coverage of the non-self. The V-detector algorithm can only identify self and non-self, so the fault diagnosis model of immune clustering algorithm which can adaptively identify fault types is established. The fault diagnosis model based on based on the improved V-detector immune clustering has the better fault diagnosis effect than that based on the no improved V-detector immune clustering in fault diagnosis of gear.