Multi-Class Classification of Lamb-Wave Based Sensor De-bond Identification Using Artificial Neural Network For Structural Health Monitoring

The ultrasonic Lamb wave-based Structural Health Monitoring is the current trend in diagnosing engineering structures. The dynamic response of sensors assesses the structural health; hence the interfacing of the sensor and structure should be integrated appropriately. Any fault in the sensor leads to misinterpretation of structural damages. Therefore, the present work aims to model sensor de-bond assessment by implementing the Artificial Neural Network (ANN) algorithm. The ANN model is trained by the statistical indicator extracted from raw dynamic ultrasonic Lamb waves to classify different types of sensor de-bond. The Lamb waves experiment was conducted on an isotropic planner plate structure, which represents the aircraft fuselage panel by surface bonding the piezoelectric emitter at the centre of the isotropic planner plate, and a circular PZT sensor network is mounted around the emitter. The experimental work was carried out using the National Instrument (NI) DAQ system, consisting of a NI signal generation card to generate Lamb waves and a NI digitizer card to capture the propagating Lamb wave signals. The captured raw signals were processed in the signal processing workstation, and statistical indicator was extracted. These statistical indicators serve as the features to the ANN model. The obtained results exhibited that, the proposed ANN architecture for sensor interface assessment has high classification accuracy in classifying different types of sensor de-bond.