Abstract: With analysing the non linear time series for correlative parameters prediction, a neural networks method to predict the multi parameters with multi steps using correlative parameters has been demonstrated and a reat time neural networks predicting model has also been built in this paper. The prediction results were dealt with comprehensively. The method not only can make up for the lack of original information, but also can give multiple alarms for a failure occurrence ahead of time. The problems which the traditional prediction methods often meet are solved and also the precision of prediction model is improved.

Key words: neural networks, parameters prediction　fault diagno sis, condition monito ring