Improving accuracy of structural health monitoring system by bayesian approach

Abstract

Structural Health Monitoring (SHM) is a procedure of assessing structural integrity during its service period to create an improvement in maintenance procedure in terms of cost and reliability. In aerospace SHM applications, crack identification using ultrasonic Lamb wave propagation and reflection is one of the most active research areas of interest. Usually, complete information about the severity of a crack includes location, size, orientation and shape of the crack. Especially, current researches about size estimation procedure are still elusive in most two-dimensional applications. In this thesis, we focus on methods of improving crack size estimation especially by using Bayesian inference. In the experiment and simulation, a crack propagating horizontally from a small hole is considered. An array of piezoelectric sensors that are also acting as actuators is used for wave generation and sensing. We start with a small hole of 3mm diameter, and machine a crack 3 mm in each increment up to 50mm, measuring the signal amplitude at sensors and using the migration techniques to image the crack at each crack increment. From the signal amplitude variation obtained, we find a good agreement that can describe how an actual crack size changes in both simulation and experiment. Up to about 27mm, there is almost a linear increase in signal amplitude with an increase of crack size. The simulation results also agree well with the experimental results at this range. When the crack is larger, no linearity is found. We have discussed how we can obtain more detailed information about crack size and shape from the images obtained by the migration technique. Since the main challenge for SHM is the accuracy with relatively small cracks, the results that we can obtain information from signal amplitude scattered from relatively small cracks is encouraging. Once we start detecting the change of crack sizes, it is possible to obtain a more accurate diagnostic result by a...