Applied Study Of Compound Dimensionless Indicator Immune Detector For Unit Fault Diagnosis Technology

Artificial immune system (AIS) is a learning technique inspired by the principle of biological immune system which has strong information processing capability, and AIS cannot only provides some evolution learning mechanisms, such as anti-noise, self-learning, self-organizing and memory, by learning natural defense mechanism that learns about for-eign substances, but also has advantages of classifier, artificial neural networks and machine reasoning. The negative-selection algorithm (NSA) implements a learning technique inspired by the immune system’s recog-nition mechanism "Self" and "Non-self", which provides a new idea and method to fault diagnosing.In this paper, a method of rotating machinery faults diagnosis based on NSA and dimensionless indicator is presented. The main work includes the following:(1)For the reason that the existing dimensionless indicators are only sensitive to some faults, causing that is difficult to achieve excellent effect to other faults. Moreover, with the rapid development of integration, pre- cision and complication of rotating machinery, the type of fault is increas-ing so that requires larger numbers of dimensionless indicators, but the number of dimensionless indicators is limited. In order to solve the short-age of dimensionless parameter and the inadequate problem in ability of fault diagnosis, in the paper genetic programming(GP) is used to con-struct the new compound dimensionless parameters aimed at the common fault of shafting and bearing by recombining and optimizing the five ex-isting dimensionless parameters (waveform indicator, kurtosis indicator, impulse indicator, margin indicator, peak indicator). Experimental result shows that the recognition ability of optimum dimensionless parameter is better than that of existing ones, and accurate classification of shafting fault and bearing fault can be achieved.(2) Aim at huge computation and strong blindness in generating detectors of classical negative-selection algorithm at random, Method of Mutation Search to generate detectors is proposed. Experimental result shows that it could efficiently generate detectors.(3) As to the problem of useful fault information loss in the process of reduction and clustering, a simple, efficient and quick integrated diag-nosis algorithm is presented in order to improve the accuracy of diagnosis. Experimental result shows that it is effective in improving the precision of fault diagnosis.