Ultrasonic measurement principles modeling, identification, and parameter estimation

Sammanfattning: This thesis presents contributions within the fields of ultrasonic modeling and measurement technology, with focus on solutions to difficult modeling and measurement problems. The work is divided into two categories: 1) processing of measurements obtained under non-ideal conditions, such as unsynchronized, distorted, and superimposed signals; 2) estimating acoustic models and parameters from materials, fluids, fluid mixtures, and thin-layered structures. The ultrasonic research field has traditionally been focused on either physical models to describe acoustic properties based on wave propagation experiments, or on statistical/empirical models to describe more complex systems. Physical models have the advantage that the parameters are directly connected to physical properties of the media, enabling an understanding of the underlying dynamics and simplifying the inverse problem. However, their disadvantage is that the derivations are often based on crude approximations and ideal conditions; limitations often leading to correlated residuals, biased parameter estimates, and the necessity of calibration measurements to solve the inverse problem. Conversely, statistical or empirical models often describe the measured data well with uncorrelated residuals, but have the disadvantage that the parameters (or models) are not directly connected to the physical properties of the material or fluid. In this case this connection is often retrieved through calibration. A key ingredient in the work presented in this thesis, is the use of a combination of physical and empirical models. This allows for a description of dynamic elements with both known and unknown structures, and the ability to have both uncorrelated residuals and unbiased parameter estimates related to the physical properties of the media. If sufficient prior knowledge exists of the physical structure and the location of possible non-ideal effects, calibration steps may be avoided or reduced significantly. This combination of hard physical structures with the variability of empirical models inherits advantages and disadvantages from both models. The benefits and limitations of the proposed solutions are analyzed and discussed, and the presented results are supported and validated with real experiments or with combinations of real experiments and simulations