Fault Diagnosis Method Based On Artificial Immune System

The biological immune system is a highly parallel adaptive information learningsystem, people has produced many new intelligent algorithm based on biological immunesystem by the study of the biological immune mechanism. In this context, Aiming at avoidingthe shortcomings of traditional fault diagnosis method, we through the theoretical analysis ofthe artificial immune system, and combinative with other methods, some research for theartificial immune system in the application of fault detection and fault diagnosis. This study ismainly divided into the following three parts:(1) Proposed a classification method of nonlinear fault based on kernel parameterimmune optimization. According to the characteristics of the process nonlinear couplingbetween the sampling data, extracting system nonlinear feature using the observation data ofresearch based on KLPP algorithm. Firstly according to the prior knowledge to determine abetter classification kernel function matrix, using the similarity measure between matrixes,then using immune optimistic algorithm to find approximate nuclear parameter of the matrixand determine the type of fault c data. Finally the TE process as an object to study based oncomputer simulation.(2) Through researched the recognition mechanism of Natural Killer cells, a new faultdetection method is proposed and its application to fault detection. In the recognition processanalysis of NK cells, Through introduced the "at least one" theory and "zero balance" model,NK cell activation mechanism is mediated by a balance between excitatory and inhibitoryreceptor on its surface to identify self and non-self detection of target cells through surfacereceptors and inhibitory receptors incentive change, Then introduce the detail artificialalgorithm model, and apply artificial NK cells algorithm on DAMADICS benchmarkplatform for fault detection.(3) Established a self-learning fault diagnosis model based on artificial immune algorithm. In view of the traditional fault diagnosis model is unable to track dynamic change,using artificial immune network algorithm training the initial fault classes of antibody library,at the same time, uses the incremental algorithm of single particle concentration sensors tosense the dynamic change of fault feature data, the failure characteristics of the new dataobtained new fault training antibodies to establish an automatic update model of fault typerecognition, on the DAMADICS benchmark platform to verify this method is effective.