Vibration analysis can give an indication of the condition of rotating shaft highlighting potential
fault such as unbalance and rubbing. Faults may however occur intermittently and consequently to
detect these requires continuous monitoring with real time analysis. In this research, we describe
how to use Artificial Neural Networks (ANN's) for classification of machine conditions by using
two sensor techniques. In this technique, calculated moments from times series are used as input
features as they can be quickly computed from measured data. Orthogonal vibrations are
considered as two -dimension victor, the magnitude of which can be expressed as time series.
Some signal processing operations are applied to the data to enhance the differences between
signals. A fault signature data base is built which includes vibration signature of common failure
modes of critical components in rotating equipment. The database is used to train the neural
network to classify the different fault classes. Such expert system has some limitations because it
is tailored to a specific machine and specific faults under certain operating conditions. Comparison
is made with frequency domain analysis methods, which has some ambiguities when components
may, more or less overlap and certain faults may exhibit themselves in different ways in spectrum.
The results show that the success of the network is highly dependent on the deduced feature signal
which contain the symptoms of faults and healthy operation.