Characterization of thin multi-layered materials using ultrasound

Sammanfattning: This thesis contributes to the field of ultrasonic modeling and measurement technologies. The objective is to address the advancement of methods for ultrasonic characterization of thin multi-layered materials. The main focus of the work is on the development of a parametric model for multi-layered structures with an arbitrary number of layers, for normal incidence pulse-echo and through-transmission ultrasonic measurements using unfocused transducers. The main advantages with the model are that it can handle measured waveforms with overlapping echoes, its complexity is connected to the number of layers rather than the number of received echoes, and the estimated parameters are directly connected to the properties of the investigated media. The layer model is also valid for any kind of signal waveform and does not depend on the excited signal's appearance. The possible limitations with the model regarding the use of unfocused transducers, the assumption of normal incidence planar waveforms and parallel surfaces, and the decreasing layer thicknesses are addressed in the thesis. Estimating the parameters of the model is a highly nonlinear process and robust optimization methods are required. The connection between the measured ultrasonic data and the physical multi-layered model by a robust estimation and optimization algorithm is a main contribution of the thesis. To contribute to the industry with accurate and cost-effective diagnostic methods, all further analysis are performed on the model parameters rather than the entire ultrasonic waveforms. The model and the estimated model parameters are used for investigation of thin embedded layers, where flaw detection, estimation of properties, layer imaging, material classification, and separation of overlapping echoes are utilized. Results show that the parametric model is able to reconstruct the significant dynamics of waveforms consisting of multiple overlapping echoes by using the estimated parameter vector. This has been verified with both pulse-echo and through-transmission measurements with normal incidence waveforms. The main conclusion in this thesis is that a robust and accurate optimization algorithm, estimating the parameters in a model describing a multi-layered structure with reasonably few parameters, is found. Once the model parameters are estimated, analyzing them in the post-processing stage will enable several applications in process control.