Rotating Machinery Fault Visualization And Prediction Methods

The prescient maintenance based on state monitoring of the rotary machine is paid a lot of attention by factories in and out of the country nowadays, in the meantime, visualization of faults, being characterized of its visibility, is pursued in many fields. This thesis makes some research on visualizing faults diagnosis and vibration forecast of rotary machine, integrates them and builds a visualized fault diagnosis and forecast system of the rotary machine.This t hesis b rings forward a m ethod o f v isualizing t he fault t o m ake the traditional fault diagnosis more visual. In virtue of self-organizing map, the network output layer characteristic flat of which reflects roughly the distribution of sample data with different kinds of sample data of t he faults are d imensionally r educed, c lustered and are s hown on a two-dimensional plane where different kind of sample data of the fault takes different area. In this way, new sample can be identified effectively and fault diagnosis is made visualized.Considering the characteristic of vibration of rotary machines, this thesis makes a thorough discussion of forecasting the trend of vibration by a means of time series model, puts forward means of processing the nonstationarity, nonnormality and singular value of the field data and distinguishing their models to build a appropriate model and gets precise mulstep forecast to the trend of vibration.Based on the above tow means, this thesis lays out a scheme of visualizing the fault diagnosis and its state forecast of rotary machines, analyses the key technologies and validates some of the theories by field data. It is shown by the results that this scheme is feasible as well as effective.By a concept of object-orientated, the system is programmed in the operating platform of Window9X and Visual C++6.0, the developing environment.