Combination of discrete cosine transform with neural network infault diagnosis for rotating machinery

摘要

The orbits of shaft centerline are pieces of indispensableinformation for the rotating machinery fault diagnosis, in general, aspecial shape of orbits of shaft centerline corresponds to a specialfault type. The application of neural network is helpful to identify theorbits of shaft centerline, and the discrete cosine transform is helpfulto reduce the input dimension of neural network. The paper discusses themethod of combining the discrete cosine transform technique with theneural network, to compress the input data while the resolving power ofinput network is improved, so as to keep the input dimension of neuralnetwork invariant. The feasibility of improved back-propagationalgorithm which makes the convergence faster is proved. Learned with theexperimental simulated fault data, the neural network system can be usedto identify orbits of shaft centerline in a higher identification rateautomatically