Unsupervised Time-series Fatigue Damage State Estimation of Complex Structure Using Ultrasound Based Narrowband and Broadband Active Sensing

This paper proposes unsupervised system identification based methods to estimate time-series fatigue damage states in real-time. Ultrasound broadband input is used for active damage interrogation. Novel damage index estimation techniques based on dual sensor signals are proposed. The dual sensor configuration is used to remove electrical noise, as well as to improve spatial resolution in damage state estimation. The scalar damage index at any particular damage condition is evaluated using nonparametric system identification techniques, which includes an empirical transfer function estimation approach and a correlation analysis approach. In addition, the effectiveness of two sensor configurations (configuration 1: sensors placed near the actuator and configuration 2: sensors placed away from the actuator) are evaluated. Furthermore, the time series 2s error bound is also evaluated to study the effect of measurement noise on damage state estimation. The time-series damage estimation approaches are validated on a complex Al-2024 cruciform specimen undergoing biaxial cyclic loading.