The increasing presence and applications of composite materials in aerospace structures precipitates the need for improved Nondestructive Evaluation (NDE) techniques to move from simple damage detection to damage diagnosis and structural prognosis. Structural Health Monitoring (SHM) with advanced ultrasonic (UT) inspection methods can potentially address these issues. Ultrasonic coda wave NDE is one of the advanced methods currently under investigation. Coda wave NDE has been applied to concrete and metallic specimens to assess damage with some success, but currently the method is not fully mature or ready to be applied for SHM. Additionally, the damage diagnosis capabilities and limitations of coda wave NDE applied to fibrous composite materials have not been widely addressed in literature. The central objective of this work, therefore, is to develop a quantitative foundation for the use of coda wave NDE for the inspection and evaluation of fibrous composite materials.

Coda waves are defined as the superposition of late arriving wave modes that have been scattered or reflected multiple times. This results in long, complex signals where individual wave modes cannot be discriminated. One method of interpreting the changes in such signals caused by the introduction or growth of damage is to isolate and quantify the difference between baseline and damage signals. Several differential signal features are used in this work to quantify changes in the coda waves which can then be correlated to damage size and growth. Experimental results show that coda wave differential features are effective in detecting drilled through-holes as small as 0.4 mm in a 50x100x6 mm plate and discriminating between increasing hole diameter and increasing number of holes. The differential features are also shown to have an underlying basis function that is dependent on the hole volume and can be scaled by a material dependent coefficient to estimate the feature amplitude and size holes.

The fundamental capabilities of the coda wave measurements, such as error, repeatability, and reproducibility, are also examined. Damage detection was found to be repeatable, reproducible, and relatively insensitive to noise. The measurements are found to be sensitive to thermal changes and absorbing boundaries. Several propagation models are also presented and discussed along with a brief analysis of coda wave signals and spectra.