Data-driven online prediction of remaining fatigue life of a steel plate based on nonlinear ultrasonic monitoring.

ABSTRACT

Online monitoring fatigue damage and remaining fatigue life (RFL) prediction of engineering structures are essential to ensure safety and reliability. A data-driven online prediction method based on nonlinear ultrasonic monitoring was developed to predict the RFL of the structures in real-time. Nonlinear ultrasonic parameters were obtained to monitoring the fatigue degradation. A Bayesian framework was employed to continuously compute and update the RFL distributions of the structures. Nonlinear ultrasonic experiments were performed on the fatigue damaged Q460 steel to validate the developed prediction methodology. The result indicates that the developed method has high prediction accuracy and can provide effective information for subsequent decision-making.