The fault diagnosis of the history, present situation and main trend is reviewed. Based on a comparison of the traditional remote monitoring and diagnosis system, provides the architecture of a remote monitoring and diagnosis system. The remote monitoring and fault diagnosis system architecture for two dust removal fan remote monitoring system, the related research and development of the function and fault diagnosis research of the monitoring system, through the remote monitoring and fault diagnosis system can understand two dust removal fan running status, ensure the wind machine running status in the design constraints. It is proposed based on a nonlinear autoregressive moving average model, tested the sunspot data, then the model is applied to fault diagnosis of fan, the experimental results show that, the linear model and the traditional prediction model, has the high prediction accuracy, a new method is proposed for the Metallurgic Fan Machinery Fault diagnosis.