Signal processing within a real time AI architecture for condition monitoring and fault diagnosis

The authors present a method for diagnosing faults based on signal processing of signals from conventional sensors taking measurements from a dynamic system. This ensures that no disruption is applied to the system in fitting `conditioning sensors´. The method is based on the techniques of system identification and Kalman filtering in which the faults are diagnosed by checking the Kalman filter residuals against previously `learnt´ patterns