Study on Mechanical Fault Diagnosis Based on Rough Set Theory and Neural Network
Shi Hui-feng · Journal of Kunming University of Science and Technology · 2011
The method of mechanical fault diagnosis based on rough set theory and neural network is put forward in this paper.Firstly,the discretization of continuous attributes by SOM and the reduction of condition attributes by discernibility matrix are studied.Then,the key problems about how to build a neural network are induced.Finally,the method is proved by an example.The results show that rough set theory can effectively get rid of redundant information,simplify the structure,decrease the training time of the network,and increase the efficiency of diagnosis.SOM can establish a mapping with perfect clustering outcome from the inputs of continuous attributes to discrete outputs.Furthermore,SOM can keep the topology between data from changing.The result of reduction by discernibility matrix is accurate and reliable.BP neural network has the powerful ability of functional approach,and it can establish a mapping from feature space to fault space quickly and correctly.