Μη καταστροφικός εντοπισμός φαινομένων διάβρωσης σε δοχεία αποθήκευσης υγρών καυσίμων

Abstract

In non-destructive control, acoustic emission signals are used for reliable construction monitoring and damage recognition. In this thesis several methods for the acoustic emission (AE) source location are developed and evaluated. Automatic estimation of minimum number and optimal placement of sensors are derived at the minimum sum of localization errors at randomly positioning AE sources. A new method was proposed and evaluated for the estimation of optimum sensors position in problems of AE localization in spherically and cylindrical structures. The particular methodology can be easily adjusted in different structures, and is of paramount important in case where the sensors must be permanently placed in a structure. Six source location methods were developed using a parametric model for the AE signal, genetic algorithm and simulated annealing. The magnitude of the Fast Fourier Transform or the position of the maximum peak of cross correlation function are extracted from the AE signals acquired by multiple sensors positioning at arbitrary locations in a plain or a cylindrical structure. The AE source is estimated at the minimum of the error function between the signal or the features derived from the acoustic signal, and the signal or features estimated from the AE signal model. Moreover, a novel source location method based on radial basis function network is presented and evaluated. The problem of AE localization in plane surfaces and cylindrical surfaces are solved in a close-form using the arrival-time differences using three or more sensors. A close-form solution for Acoustic-Emission source location (AESL) and material constant G is presented and evaluated in simulation experiments using the Time-of-Arrival (TOA) of several events detected in arbitrary positioning sensors in 3d-space in dispersive media. The normalized distances and the constant G are derived from the TOA at four arbitrary selected sensors using the events propagation velocities in a reference material. The actual AE position is derived using the multidimensional scaling method using the complete set of sensors. In simulation experiments, the advantages of the proposed method are demonstrated. Overcoming the most important weakness of the proposed method, the use of only four sensors for the estimation of the parameter G, an algorithm for successive estimation of the AESL is developed using the complete set of TOAs. An extension of the AESL method is developed using a successive approximation algorithm assuming a minimum of two known propagation velocities for the recorded events. It is proved that the proposed algorithm converges to the local minimum of the optimization function. Under few restrictions the proposed algorithm can be used to estimate the AESL even in case where the propagation velocities for all events are unknown.

Πραγματοποιήθηκε μελέτη και ανάλυση σημάτων πηγών ακουστικής εκπομπής. Υλοποιήθηκαν συστήματα προσομοίωσης μεθόδων εύρεσης πηγής ακουστικής εκπομπής από υπάρχουσες και νέες μεθοδολογίες, βασισμένες σε συμβατικές αλλά και προχωρημένες τεχνικές επεξεργασίας και αναγνώρισης προτύπων. Οι µμεθοδολογίες που προτάθηκαν παρουσιάζουν δυνατότητα ανάλυσης περισσότερης πληροφορίας και αποδείχθηκαν ανθεκτικότερες στο θόρυβο. Επιπλέον η εφαρμογή τους επιτρέπει την χρήση μεγάλου αριθμού αισθητήρων σε διάφορα υλικά στο δισδιάστατο και τρισδιάστατο χώρο. Τέλος παρουσιάζεται μια προσέγγιση για την εκτίμηση της βέλτιστης τοποθέτησης αισθητήρων µε στόχο την ελαχιστοποίηση του σφάλματος εντοπισμού της θέσης της πηγής.