Abstract: Aim To give the description and the way of judgement on trend analysis for machinery working condition; to study the building of prediction models and the selecteing of prediction parameters for trend analysis and probe an improvement on the prediction model.Methods A combined prediction model(GMAR) was adepted; a practical method of the trend prediction with frequency spectrums of six sections was used.Results and Conclusion The technology of trend analysis for machinery working condition can offer valid way for predicting the faults of machinery system on line.