Data-Informed Structural Diagnostic Framework

Crack growth in airframe structures may be considered as a nonlinear dynamic process where the crack size at each time step is predicted by a physical damage model such as the Paris-Erdogan equation and updated by a sequence of noisy sensor measurements. The prediction-update implementation is a two stage recursive process that uses a sequential Monte Carlo sampling to obtain a posterior density of crack size at the desired inspection stage. The goal of this paper is to develop a datainformed diagnostic framework that combines the predictions of a physical damage model and evidence from ultrasonic sensor data to provide accurate estimation of crack-size in rotorcraft structures. The data-informed approach obtains the most probable crack-size (MPCS) at each inspection stage using a sequential Monte Carlo method known as Particle Filters. To validate the approach, two Al7075-T6 nested-angle plates were tested using rotorcraft spectrum loads. Results show that the prediction error of the data-informed method, as measured by root mean square deviation from the true value, is less than half of the error of both the Paris-Erdogan and NASGRO models.