|
Current ultrasonic acoustic NDE methods for long distance inspection in cylindrical structures are primarily focused on axisymmetric guided waves excitation. However, there are many occasions where the physical limitations imposed by the system to be inspected restrain the ability to utilize equipment capable of exciting those waves. This study explores the excitation of the flexural guided wave modes by a limited number of piezoelectric transducers for damage detection in hollow cylinders with limited surface access and large diameter. In addition, the use of distributed optical fiber system as the guided wave receptor is investigated as an alternative to piezoelectric transducers (PZT), as their capability to acquire spatial-temporal data synergizes with the complexity in a signal containing several flexural GW modes. More specifically, the study is conducted based on a numerical analysis of the guided waves excited by a 2 PZT configuration in a pipe available for experimental testing. The resulting flexural modes and its interaction with welds and local loss of material are analyzed in terms of the time series data of a local sensor in the surface, and the angular profile differences from a healthy case. A method based on the analytical solution of an infinite cylinder is introduced in preliminary stage to extract the behavior of the dominant modes from simulation and experimental results and used as a simulation-experiment similarity comparison. Finally, a simplified convolutional neural network (CNN) is trained to demonstrate feasibility of using flexural modes excited by limited actuators for damage detection. Overall, this study contributes to the development of a damage detection method applicable to cylindrical structures with dimensional and access limitations, by enhancing the understanding of how simultaneous several flexural modes interact with mechanical features, presenting an early-stage interpretable method to compare simulation and experimental fiber optic sensor data, and demonstrating the feasibility of using DAS like data for analyzing the structure.
|
