State Tracking of Composite Delaminations with a Bayesian Filter

Abstract

We propose a method for tracking the condition of a composite part using Bayesian filtering of nondestructive evaluation (NDE) data over the lifetime of the part. NDE provides information about the state of a part or material without destroying or degrading the part. The Bayesian process builds on the lifetime history of NDE scans and can give better estimates of material condition compared to the most recent scan alone, which is the common practice in the aerospace industry. Bayesian inference provides probabilistic estimates of damage state that are updated as each new set of NDE data becomes available. The method is tested on simulated data and then on an experimental data set. Flash thermography NDE data was collected over the lifetime of a part to serve as a time history of that part. Computed tomography (CT) data was also collected after each damage event and provided a high resolution volume model of damage that acted as 'truth'. After each time point, the condition estimate was compared to 'ground truth' from CT to evaluate the performance of the thermography-based condition tracking.