Mining engineers are tasked with ensuring that underground mining operations be both safe and efficiently productive. Induced stress in deep mines has a significant role in the stability of the underground mines and hence the safety of the mining workplace because the behavior of the rock mass associated with mining-induced seismicity is poorly-understood. Passive seismic tomography is a tool with which the performance of a rock mass can be monitored in a timely manner. Using the tool of passive seismic tomography, the advance rate of operation and mining designs can be updated considering the induced stress level in the abutting rock. Most of our current understanding of rock mass behavior associated with mining-induced seismicity comes from numerical modeling and a limited set of case studies. Therefore, it is critical to continuously monitor the rock mass performance under induced stress. Underground stress changes directly influence the seismic wave velocity of the rock mass, which can be measured by passive seismic tomography. The precise rock mass seismicity can be modeled based on the data recorded by seismic sensors such as geophones of an in-mine microseismic system. The seismic velocity of rock mass, which refers to the propagated P-wave velocity, varies associated with the occurrence of major seismic events (defined as having a local moment magnitude between 2 to 4). Seismic velocity changes in affected areas can be measured before and after a major seismic event in order to determine the highly stressed zones. This study evaluates the seismic velocity trends associated with five major seismic events with moment magnitude of 1.4 at a deep narrow-vein mine in order to recognize reasonable patterns correlated to induced stress redistribution. This pattern may allow recognizing areas and times which are prone to occurrence of a major seismic event and helpful in taking appropriate actions in order to mitigate the risk such as evacuation of the area in abrupt cases and changing the aggressive mine plans in gradual cases. In other words, the high stress zones can be distinguished at their early stage and correspondingly optimizing the mining practices to prevent progression of high stress zones which can be ended to a rock failure. For this purpose a block cave mine was synthetically modeled and numerically analyzed in order to evaluate the capability of the passive seismic tomography in determining the induced stress changes through seismic velocity measurement in block cave mines. Next the same method is used for a narrow vein mine as a case study to determine the velocity patterns corresponding to each major seismic event. / Doctor of Philosophy / Mining activities unbalance the stress distribution underground, which is called mining induced stress. The stability of the underground mines is jeopardized due to accumulation of induced stress thus it is critical for the safety of the miners to prevent excessive induced stress accumulation. Hence it is important to continuously monitor the rock mass performance under the induced stress which can form cracks or slide along the existing discontinuities in rock mass. Cracking or sliding releases energy as the source of the seismic wave propagation in underground rocks, known as a seismic event. The velocity of seismic wave propagation can be recorded and monitored by installing seismic sensors such as geophones underground. The seismic events are similar to earthquakes but on a much smaller scale. The strength of seismic events is measured on a scale of moment magnitude. The strongest earthquakes in the world are around magnitude 9, most destructive earthquakes are magnitude 7 or higher, and earthquakes below magnitude 5 generally do not cause significant damage. The moment magnitude of mining induced seismic events is typically less than 3. In order to monitor mining induced stress variations, the propagated seismic wave velocity in rock mass is measured by a series of mathematical computations on recorded seismic waves called passive seismic tomography, which is similar to the medical CT-scan machine. Seismic wave velocity is like the velocity of the vibrating particles of rock due to the released energy from a seismic event. This study proposes to investigate trends of seismic velocity variations before and after each seismic event. The areas which are highly stressed have higher seismic velocities compared to the average seismic velocity of the entire area. Therefore, early recognition of highly stressed zones, based on the seismic velocity amount prior the occurrence of major seismic events, will be helpful to apply optimization of mining practices to prevent progression of high stress zones which can be ended to rock failures. For this purpose, time-dependent seismic velocity of a synthetic mine was compared to its stress numerically. Then, the seismic data of a narrow vein mine is evaluated to determine the seismic velocity trends prior to the occurrence of at least five major seismic events as the case study.

Αντικείμενο της παρούσας διδακτορικής διατριβής αποτελεί η μελέτη και ανάπτυξη νέων μεθοδολογιών αυτόματης επεξεργασίας σεισμολογικών δεδομένων, µε σκοπό την επίλυση σημαντικών προβλημάτων που συναντώνται στα πεδία των επιστημών της σεισμολογίας και της γεωφυσικής όπως: 1) η ανίχνευση μικροσεισμικών γεγονότων από µία καταγραφή, µε άλλα λόγια ο διαχωρισμός της καταγραφής σε τμήματα που αποτελούνται από εδαφικό θόρυβο και σε τμήματα που περιέχουν την χρήσιμη πληροφορία (σεισμικά γεγονότα) για τους γεωεπιστήμονες και 2) η εκτίμηση των χρόνων άφιξης των διαμήκων (P-) καθώς και των εγκαρσίων (S-) σεισμικών φάσεων. Πιο αναλυτικά, η διατριβή είναι δομημένη ως εξής: Το πρώτο κεφάλαιο αποτελεί την εισαγωγή της διατριβής. Στο δεύτερο κεφάλαιο συγκεντρώνονται και κατηγοριοποιούνται όλες οι υπάρχουσες τεχνικές που έχουν αναπτυχθεί για την επίλυση του προβλήματος της αυτόματης ανίχνευσης σεισμικών γεγονότων καθώς και τον αυτόματο προσδιορισμό του χρόνου άφιξης των P και S σεισμικών φάσεων. Συγκεκριμένα γίνεται κατηγοριοποίηση αυτών σε τεχνικές που στηρίζονται στην ανάλυση και επεξεργασία των σεισμικών καταγραφών στα πεδία του χρόνου και της συχνότητας, στη χρήση νευρωνικών δικτύων, στην ανάλυση χρονικών σειρών και αυτοπαλινδρόμησης, στην ανάλυση της πόλωσης των κυμάτων, στις στατιστικές υψηλότερης τάξης, μεθόδους ασαφούς λογικής, κυματιδιακές μεθόδους κτλ. Στο τρίτο κεφάλαιο, αναπτύσσεται νέα τεχνική για την επίλυση του προβλήματος της αυτόματης ανίχνευσης σεισμικών γεγονότων από μία καταγραφή, η οποία βασίζεται σε μία μη αυστηρή διαδικασία ελέγχου υποθέσεων. Η προτεινόμενη τεχνική πραγματοποιείται σε δύο στάδια. Κατά το πρώτο στάδιο εκτιμώνται οι εμπειρικές συναρτήσεις πυκνότητας πιθανότητας που προκύπτουν τόσο από τον εδαφικό θόρυβο όσο και από τα υπόλοιπα που προέκυψαν από την λεύκανση αυτού. Κατά το δεύτερο στάδιο προτείνεται στατιστικό τεστ τύπου κατωφλίωσης για την αυτόματη ανίχνευση μικροσεισμικών γεγονότων. Η προτεινόμενη τεχνική εφαρμόζεται σε συνθετικά και πραγματικά δεδομένα και συγκρίνεται με τον γνωστό αλγόριθμο του λόγου βραχυπρόθεσμου προς μακροπρόθεσμο μέσο (STA/LTA). Στο τέταρτο κεφάλαιο παρουσιάζεται μέθοδος για την επίλυση του προβλήματος του αυτόματου προσδιορισμό του χρόνου άφιξης της P φάσης κάνοντας χρήση στατιστικών ανώτερης τάξης. Συγκεκριμένα, γίνεται χρήση των ποσοτήτων της λοξότητας, της κύρτωσης και μίας εκτίμησης της αντιεντροπίας ως γραμμικός συνδυασμός των παραπάνω. Επιπλέον παρουσιάζονται τα αποτελέσματα από την εφαρμογή της συγκεκριμένης τεχνικής σε συνθετικά αλλά και πραγματικά δεδομένα μικροσεισμικού δικτύου, κατάλληλα για χρήση στην παθητική σεισμική τομογραφία υψηλής ευκρίνειας. Τα αποτελέσματα αυτά συγκρίνονται με γνωστές ενεργειακές μεθόδους. Στο πέμπτο κεφάλαιο, αναπτύσσεται νέα τεχνική για την επίλυση του προβλήματος της αυτόματης εκτίμησης του χρόνου άφιξης της S φάσης. Η προτεινόμενη τεχνική βασίζεται στην στατιστική επεξεργασία συγκεκριμένης χαρακτηριστικής συνάρτησης, η οποία προκύπτει από τις ιδιότητες πόλωσης των σεισμικών κυμάτων που έχουν καταγραφεί. Επιπλέον, για να ελαττωθεί η εξάρτηση του προτεινόμενου αλγορίθμου από το χρησιμοποιούμενο παράθυρο, ακολουθείται μια πολυ-παραθυρική προσέγγιση του προβλήματος σε συνδυασμό με χρήση συναρτήσεων βαρών οι οποίες εκτιμώνται αυτόματα και βασίζονται στις μεταβολές της ενέργειας του σήματος κατά τη S άφιξη. Τέλος, παρουσιάζονται τα αποτελέσματα της εφαρμογής της μεθόδου σε πραγματικά δεδομένα καθώς και η αξιολόγησή τους σε περιβάλλον θορύβου. Στο έκτο κεφάλαιο, παρουσιάζονται τα αποτελέσματα της εφαρμογής των προτεινόμενων τεχνικών σε δεδομένα μικροσεισμικού δικτύου και συγκεκριμένα σε δεδομένα που προέκυψαν από πειράματα παθητικής σεισμικής τομογραφίας και τεχνητής υδραυλικής διάρρηξης που έλαβαν χώρα στην περιοχή Δέλβινο της ΝΔ Αλβανίας. Επιπλέον, γίνεται ανάλυση των αποτελεσμάτων βάσεις των δεικτών αβεβαιότητας που επέλεξαν οι αναλυτές στις εκτιμήσεις τους, καθώς και βάσει των λόγων σήματος θορύβου των καταγραφών. Στο έβδομο κεφάλαιο παρατίθενται τα συμπεράσματα της παρούσας διδακτορικής διατριβής, καθώς και πιθανές μελλοντικές προεκτάσεις. / The problems of seismic event detection and P- and S-phase arrival time estimation constitute important and vital tasks for the geoscientists. The solution of the aforementioned problems provides with important geophysical and seismological information, that can be used in a number of problems such as the structure of the earth’s interior, geotectonic settings, hypocentric and epicentric coordinates of an earthquake, the seismicity of an area and seismic hazard assessment. Traditionally, human experts have carried out this task. Nevertheless, during the last three decades due to the progress in computer technology, several methods have been developed for the automatic seismic event detection and P- and S- phase identification. After the introduction of the first chapter, in the second chapter the majority of the existing methods that have been developed and applied up to now, are gathered and categorized. These methods involve energy criteria, the seismic wave polarity assumption, artificial neural networks, higher order statistics, maximum likelihood methods, fuzzy logic methods etc. In the third chapter, a new thresholding type technique is proposed, tailored to fit real world situations where our knowledge on the statistical characteristics of the background noise process are unknown and a strict hypothesis testing framework can not be followed. In such cases the replacement of the unknown probability density function under the null hypothesis by its empirical counterpart, constitutes a possibility. In this work, a two stage procedure is proposed. The first one concerns the estimation of the empirical functions of the noise process itself as well as its whitened counterpart. In the second stage, using the above empirical functions, a thresholding scheme is proposed in order to solve the problem of the detection of seismic events in a non strict hypothesis testing framework. The performance of the proposed technique is confirmed by its application in a series of experiments both in synthetic and real seismic datasets. In the fourth chapter, the problem of automatic P-phase identification is solved using higher order statistics. The first- and second-order statistics (such as mean value, variance, autocorrelation, and power spectrum) are extensively used in signal processing to describe linear and Gaussian processes. In practice, many processes deviate from linearity and Gaussianity. Higher order statistics can be used for the study of such processes. The P-phase arrival time is estimated using these HOS parameters and additionally, an estimation of the negentropy defined as a linear combination of skewness and kurtosis. According to the implemented algorithm a moving window “slides” on the recorded signal, estimating skewness, kurtosis, and negentropy. Skewness can be considered as a measure of symmetry of the distribution, while kurtosis is a measure of heaviness of the tails, so they are suitable for detecting parts of the signal that do not follow the amplitude distribution of ambient noise. Seismic events have higher amplitudes in comparison to the seismic noise, and these higher values occupy the tails of the distribution (high degree of asymmetry of distribution). In the case of seismic events, skewness and kurtosis obtain high values, presenting maxima in the transition from ambient noise to the seismic events (P-arrival). The proposed algorithms are applied on synthetic as well as real seismic data and compared to well known energy based methods. Algorithms that deal with the automatic S-onset time identification problem, is a topic of ongoing research. Modern dense seismic networks used for earthquake location, seismic tomography investigations, source studies, early warning etc., demand accurate automatic S-wave picking. Most of the techniques that have been proposed up to now are mainly based on the polarization features of the seismic waves. In the fifth chapter, a new time domain method for the automatic determination of the S-phase arrival onsets is proposed and its implementation on local earthquake data is presented. Eigevalue analysis is taking place over small time intervals, and the maximum eigenvalue which is obtained on each step is retained for further processing. In this way a time series of maximum eigenvalues is formed, which serves as a characteristic function. A first S-phase arrival time estimation is obtained by applying the kurtosis criterion on the derived characteristic function. Furthermore, a multi-window approach combined with an energy-based weighting scheme is also applied, in order to reduce the algorithm’s dependence on the moving window’s length and provide a weighted S phase onset. Automatic picks are compared against manual reference picks and moreover the proposed technique is subjected to a noise robustness test. In the sixth chapter, the results of the implementation of the proposed techniques on microseismic data are presented. Specifically, the proposed methods are applied on two real sets of data. One dataset was been recorded during a Passive Seismic Tomography (PST) experiment, while the second one during the seismic monitoring of fracking operations. Both experiments took place in a hydrocarbon field in Delvina, SW Albania. These results are also analyzed, based on the arrival times and their uncertainty as they were evaluated by human analysts as well as the corresponding signal to noise ratio of the seismic records. Finally, the seventh chapter concludes this work and possible future extensions are discussed.