Anomaly prediction in mechanical systems using symbolic dynamics

This paper presents anomaly prediction in complex mechanical systems at an early stage where anomaly is defined as an observable deviation from the nominal dynamical response. The anomaly prediction algorithm is built upon two-time-scale analysis of time series data and relies on a combination of nonlinear systems theory and language theory. The algorithm has been validated for anomaly prediction on a rotorcraft gearbox testbed for two different types of anomalies.