Digital Analysis of Damped Vibrations

In physical models of measurement processes, material and geometrical parameters are often characterized by a specific level of uncertainties due to measurements, random noise, or undesired variations in industrial manufacturing procedures. For those reasons, especially with gradient-based methods, a solution to the inverse problem might be affected by both the selection of initial points and involved uncertainties. Those critical factors were considered in comparing the Levenberg-Marquardt (LM) and the Gauss-Newton (GN) optimization algorithms with varied regularization methods regarding execution speed, the amount of allocated memory, and code security. Finally, the most robust and efficient algorithm was implemented in the end-devices with limited resources.