An FS-TAR based method for vibration-response-based fault diagnosis in stochastic time-varying structures: Experimental application to a pick-and-place mechanism

The problem of vibration-response-based fault diagnosis, that is fault detection and identification, in stochastic time-varying structures is considered via a statistical time series method. The method is based on stochastic Functional Series Time-dependent AutoRegressive (FS-TAR) modelling of the structural dynamics, as well as on an appropriate statistical decision making scheme for fault diagnosis. It is an output-only method, capable of operating with a minimal number of random vibration response signals, even of limited time duration and frequency bandwidth, under normal operating conditions, and in a potentially automated way. The method is applied to the problem of fault diagnosis in a pick-and-place mechanism based on a single vibration response signal. Its performance characteristics are thus confirmed using various fault scenarios and a number of experimental test cases.

► A statistical vibration-based method for fault diagnosis in time-varying structures is introduced. ► Functional Series TAR models are used based on a multi-stage estimation algorithm. ► The fault diagnosis effectiveness is assessed via an application to a pick-and-place mechanism. ► Faults of various types and occurrence locations are successfully detected.