Abstract: To deal with the problem of small-sample modeling in equipment condition on-line monitoring,an on-line monitoring method based on dynamic regression extreme learning machine(DR-ELM) is proposed.Condition data of mechanical equipment are used to train a prediction model based on DR-ELM.In an iterative manner,the latest condition data are adopted and the oldest condition data are abandoned,to achieve the DR-ELM prediction model training on-line.Thus, the current condition of mechanical equipment can be effectively predicted by the method.Simulation on chaotic time series prediction and fan condition monitoring based on time series prediction indicate that the method has better performance in training computational cost and prediction accuracy in comparison with conventional condition monitoring methods based on extreme learning machine(ELM) and on-line sequential extreme learning machine(OS-ELM).