Sistemi di visione per misure di vibrazione dinamiche

Thanks to the technological developments related to CCD and CMOS sensors, non contact measurements have improved their importance in scientific and industrial fields. They allow to obtain an high automation level reducing measurements times also for very complex systems. Moreover using non contact techniques load effects due to the interactions between measurement device and measurand are not present (1). Very important applications fields are geometric measures (both 2D and 3D), quality control and reverse engineering problems. The main non contact measurement techniques are: Laser Doppler Vibrometer, GPS tracking systems and, vision algorithms, leiv motive of this thesis. In particular these algorithms are: pattern matching, blob analysis and edge detection. Nowadays, although their first application is static, all these techniques can be used for vibration study with good measurement results and low costs (2) (3). To analyze the AVI files, the pattern matching selection is better both for the results quality and because of the easy usage. If it is not possible to fix a target on the measured object, it offers higher probability to make correct vibration measurements, among edge detection and blob analysis (4). It happens because a non homogeneous an singular area is enough to apply pattern matching, instead particular targets are required by the other considered techniques. For these reasons pattern matching has been chosen as the processing algorithm for this work. The aim of this thesis is to understand which are the parameters that mostly influenced dynamic measures obtained with vision based techniques. Particular interest is placed on the research of quasi static test conditions that allow to perform measures also in non controlled systems. Another purpose is to describe how the deviation between real displacement and measured one changes when the experiments are carried out in non static conditions in order to be able to adjust the results.

E’ scopo di questa tesi capire quali siano i parametri che maggiormente influenzano le misure di sistemi dinamici controllati, elaborate attraverso la tecnica di pattern matching. Nello specifico sono stati eseguiti una serie di test combinando le principali variabili sperimentali al fine di osservare come queste influenzassero i risultati. Si vuole riuscire a definire le condizioni di prova quasi statiche che assicurino di poter eseguire le misure anche in caso di sistemi non controllati. A tale proposito in condizioni dinamiche, sono state eseguite parallelamente un’analisi della variazione della Modulated Transfer Function ed un’analisi della discrepanza assoluta tra i risultati ottenuti con tecniche di visione e con vibrometro laser. Ulteriore obiettivo è comprendere l’andamento della discrepanza in funzione alle generiche condizioni di prova al fine di poter operare una correzione dei risultati ottenuti in situazioni non ottimali.