Research And Application Of Fault Diagnosis System Based On Rough Sets-Neural Network

At present, fault diagnosis has been developed to the stage of the smart. The research focus for intelligent fault diagnosis technology has been gradually shifting from traditional artificial intelligence to computational intelligence. Some theoretical in computational intelligence fields, such as artificial neural network, Rough Set theory has been widely used in fault diagnosis. Artificial Neural Network is an adaptive nonlinear dynamical system with learning and parallel computation capability, can achieve some function such as classification, self-organizing, associative memory function. RS is a mathematical tool to characterization integrity and uncertainty. It can effectively analyze and process inaccurate, incomplete, inconsistent and other incomplete information, find that the implicit knowledge and the potential law. In this paper combining RS and ANN, using both the advantages , giving a fault diagnosis system model based on the rough set neural network.Firstly, studying the attributes reduction methods use of rough set theory ,and on this basis giving a methed based on the binary matrix identifiable attribute reduction, reduce the amount of computation, can quickly be given to the request of attribute reduction.Secondly, researching the discrete method. Rough set theory is based on discrete data processing methods, continuous data discretization direct impact on its effect, in this paper giving a method use Particle Swarm Optimization(PSO), transformation the find of segmentation points into indicators optimization problem , the results of example demonstrate the effectiveness of this method.Then, concerning the shortcomings of ANN in a limited data, such as low studying precision and poor generalization, adopted the PSO to improve ANN ,then combine rough sets and neural networks, mostly exerts the advantage about both rough sets and neural network.Finally, using the fault diagnosis system by rough-neural network on Tennessee-Eastman process(TEP) fault diagnosis, achieved good diagnosis result.