Abstract: On-line identification the running state for kind of rotary machines, which are often running under complicated condition and have not clear evaluation criterion for states, is studied by adaptive digital filter arithmetic. The adaptive digital filter arithmetic and the identification principle are discussed, an application example is given, and the result indicates that the method can discriminate between normal states and abnormal states. By this method, computation cost is low, and on-line identification can be archived easily.

Key words: Adaptive digital filter arithmetic, ARIMA model, State identification, Time series