Rotating machines fault identification using back-propagation artificial neural network

The authors describe a newly developed technique and system for real-time monitoring and identification of machine condition. The machine health identification process is mainly based on recognition and comparison of the real-time captured vibrational signature to its standard signature. The features extraction of the vibrational signature uses the technique of higher order spectra analysis. These signature features will then input to an artificial neural network (ANN) for recognition and identification. The output of the neural network was trained to generate a healthy index that indicates the machine health condition. A DSP56001 based digital signal processor is employed to implement the signal processing algorithms together with the artificial neural networks for real-time operation. The authors briefly describe the methodology, system and vibrational signature recognition. Very encouraging and successful results have been obtained and are presented and discussed