Reconstruction of Fouling Distribution from Aggregate Observations

Ultrasonic guided waves (UGWs) can propagate over long distances with little attenuation and are suitable for non-destructive evaluation of a large area. However, localizing fouling inside the closed systems (e.g., pipes, engines) remains to be challenging. Here we propose to use UGWs that propagate along the pipe circumference along multiple helical paths, thus, providing denser structural information without increasing the number of sensors. When the path crosses fouled area the UGWs propagation properties change, generating unique features into each helical wave. Such aggregate information together with the known structural geometry are then used to fit a latent variable Gaussian process to recover the fouling distribution map in a probabilistic manner.