Cepstrum - based Deconvolution Techniques for Ultrasonic Pulse - echo Imaging of Flaws in Composite Laminates

Ultrasonic imaging is a powerful nondestructive evaluation (NDE) tool for flaw characterization. This thesis discusses the signal processing techniques developed for ultrasonic imaging of flaws in composite laminates. Commonly used time-domain signal processing techniques have problems including poor resolution, dependence on operator's experiences and subjectivity. Some transform-domain processing techniques, such as homomorphic deconvolution, eliminate the problems associated with traditional time-domain signal processing techniques. However, these techniques are often sensitive to environmental noise and difficult to apply under certain conditions. A modified deconvolution technique, based on homomorphic deconvolution, is proposed, which allows for processing real-life signals with noise. Results from inspecting composite laminates show clear improvement of this technique over traditional methods. Thesis Supervisor: Shi-Chang Wooh Title: Assistant Professor of Civil and Environmental Engineering