ebook img

Temporal and spectral pattern recognition for detection and combined network and array waveform PDF

263 Pages·2017·48.58 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Temporal and spectral pattern recognition for detection and combined network and array waveform

Temporal and spectral pattern recognition for detection and combined network and array waveform coherence analysis for location of seismic events Von der Fakultät für Bau- und Umweltingenieurwissenschaften der Universität Stuttgart zur Erlangung der Würde eines Doktor-Ingenieurs (Dr.-Ing.) genehmigte Abhandlung Vorgelegt von ick Benjamin S aus Heilbronn oswig Hauptberichter: Prof. Dr. Manfred J okelmann Mitberichter: Prof. Dr. Götz B 12 2017 Tag der mündlichen Prüfung: . April Institut für Geophysik der Universität Stuttgart 2017 The story so far: In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move. — Douglas Adams, The Restaurant at the End of the Universe Simplicity is prerequisite for reliability — Edsger W. Dijkstra ACKNOWLEDGEMENTS My special thanks goes to the following people, organizations and everybody who I forgot but would deserve to be thanked: My sincere gratitude goes to my advisor, Prof. Manfred Joswig who gave me the opportunity for this PhD thesis. Thank you for your trust and the freedom you gave me to be inde- pendent during the whole time of the thesis. I am thankful to Prof. Götz Bokelmann who was so kind to be the second reviewer. Thanks to my colleague Rudolf Widmer-Schnidrig for proof reading the thesis and giving me detailed feedback. Special thanks also to the colleagues of the institute for a nice atmosphere and open ears for any problems. Especially: Patrick Blascheck (also for proof reading parts of the thesis), Marco Walter, Sabrina Rothmund and Rolf Häfner. I would like to thank the CTBTO for financing my research position at the institute for the development of the OSI SAMS software. This allowed me to have a very meaningful program- ming project and I could apply outcomes of the thesis there. The CTBTO also gave me the Young Scientist award and the possibility to attend various very interesting UN training exer- 14 cises, especially the IFE . Peter Labak together with Gregor Malich of the CTBTO were always good partners who trusted me with the development of the software. The thesis could not have been possible without the people deploying and maintaining the seismic stations and creating the ground truth data: 1 . Basel dataset: Martin Häge and Marco Walter deployed the stations. Martin Häge and Patrick Blascheck did the manual analysis with an extensive ground truth of mul- tiple thousand events by Patrick Blascheck. The external data with an additional ground truth was recorded and made available by the SED. 5 2 . DGMKdataset: UlrichSchwadererwiththehelpofPatrick Blascheck, GregorMokelke, MarcSchmid, AlexanderKich- erer, Tom Rohloff and Juan-Carlos Santoyo deployed the initial DGMK network. Later maintenance and exten- sions were done by Zaneta Gurbisz and Gregor Mokelke supported mainly by Marc Schmid, Alexander Kicherer, Marco Walter and me. Uwe Niethammer generated the syntheticseismogramsforthisdataset. Thefundingforthe 761 project came from DGMK, Project : ExxonMobil Pro- duction Deutschland GmbH, DEA Deutsche Erdoel AG, GDF SUEZ E&P Deutschland GmbH, Wintershall Holding GmbH. External data was provided by BGR and WEG. 3 14 . IFE dataset: The CTBTO organized the training exercise and made the data available. Network deployment, main- tenance and data screening was done by the OSI SAMS team. Special thanks goes to the team members Nicolai Gestermann and Martin Häge. 4 . PISCO dataset: The GFZ made the data available publicly 2010 with an accompanying ground truth (GEOFON, ). Günter Asch provided the dataset initially together with additional information. 5 . The data of the nuclear test of North Korea is available from IRIS. Although I re-implemented most of the algorithms from scratch during this thesis, I would like to thank all the soft- ware developers whose open-source libraries I used. I used Python to create most of the plots with mainly the libraries NumPy, SciPy, matplotlib and seaborn. I used ObsPy (Megies 2011 et al., ) for MiniSEED reading, acquisition of external data via SeedLink and FDSN webservices and to test some of my trig- ger implementations. I used the Generic mapping tools (GMT, 1998 Wessel & Smith ) to create the PISCO map. Scikit-learn 2011 (Pedregosa et al., ) was used for the Random Forest and 2015 Naive Bayes classifiers. Keras (Chollet, ) was used for the conventional, recurrent and convolutional neural networks. The 6 PCA plots were created with the JFreeChart Java plotting library. The thesis was written in LATEX and the PGF and TikZ packages were used to combine Figures and add explanations. Last but certainly not least I would like to thank my wife Alice for being such a wonderful partner during all these years. 7 CONTENTS 1 introduction 31 2 challenges of automatic processing 37 3 outline of test datasets 41 31 42 . Basel . . . . . . . . . . . . . . . . . . . . . . . . . . 32 43 . DGMK . . . . . . . . . . . . . . . . . . . . . . . . . 33 14 45 . IFE . . . . . . . . . . . . . . . . . . . . . . . . . . 34 48 . PISCO . . . . . . . . . . . . . . . . . . . . . . . . . . 4 algorithmic building blocks 51 41 52 . Single stations . . . . . . . . . . . . . . . . . . . . . 411 . . Single station event detection and phase 52 picking . . . . . . . . . . . . . . . . . . . . . 412 . . Single station event classification by pat- 66 tern recognition . . . . . . . . . . . . . . . . 42 89 . Coincidence processing . . . . . . . . . . . . . . . . 43 91 . Coherence analysis . . . . . . . . . . . . . . . . . . 431 92 . . Beamforming . . . . . . . . . . . . . . . . . 432 105 . . Combined array spectrograms . . . . . . . 433 . . Comparison of array algorithms at one 105 mini-array . . . . . . . . . . . . . . . . . . . 5 clustering by unsupervised pattern recogni - tion 111 51 114 . Feature extraction . . . . . . . . . . . . . . . . . . . 511 115 . . Interpretation . . . . . . . . . . . . . . . . . 52 117 . Application of clustering algorithms . . . . . . . . 53 118 . PCA . . . . . . . . . . . . . . . . . . . . . . . . . . . 531 118 . . Cluster analysis with PCA . . . . . . . . . . 532 119 . . Feature reduction . . . . . . . . . . . . . . . 533 125 . . Visualization by back-transformation . . . 534 125 . . Variations along principal components . . 54 127 . SOM . . . . . . . . . . . . . . . . . . . . . . . . . . . 541 133 . . Visualization of SOM nodes . . . . . . . . . 542 133 . . Event locations of SOM nodes . . . . . . . 9 543 135 . . Amplitude invariance . . . . . . . . . . . . 544 136 . . Supervised classification with SOM . . . . 6 waveform coherence for seismic event loca - - tion 145 61 146 . Conventional processing . . . . . . . . . . . . . . . 62 148 . Source scanning . . . . . . . . . . . . . . . . . . . . 621 153 . . Characteristic function (CF) . . . . . . . . . 622 157 . . Weighting with Fisher ratio . . . . . . . . . 623 158 . . Normalization of characteristic functions . 624 . . Finalstackingoftraveltimecorrectedchar- 161 acteristic functions . . . . . . . . . . . . . . 63 164 . Application . . . . . . . . . . . . . . . . . . . . . . . 631 164 . . Basel Deep Heat Mining project monitoring 632 166 . . Gas field monitoring in northern Germany 633 . . CTBTO on-site inspection Integrated Field 2014 172 Exercise . . . . . . . . . . . . . . . . . 7 conclusions and outlook 181 71 185 . Outlook . . . . . . . . . . . . . . . . . . . . . . . . . 72 188 . Algorithm implementations . . . . . . . . . . . . . 189 References 207 Appendices a combined real time detector 209 - b noise removal by simplified source scanning211 c visual seismic event screening by super sono - - grams 215 c1 215 . Abstract . . . . . . . . . . . . . . . . . . . . . . . . . c2 216 . Introduction . . . . . . . . . . . . . . . . . . . . . . c21 217 . . Nanoseismic Monitoring . . . . . . . . . . . c22 219 . . Heumoes slope example dataset . . . . . . c3 219 . Super-sonogram event screening and classification c31 221 . . Sonogram calculation steps . . . . . . . . . c32 223 . . Super-sonograms . . . . . . . . . . . . . . . c33 225 . . Signal classification . . . . . . . . . . . . . . c4 227 . Software . . . . . . . . . . . . . . . . . . . . . . . . c5 232 . Conclusions and discussions . . . . . . . . . . . . c6 233 . Acknowledgements . . . . . . . . . . . . . . . . . . 10

Description:
NumPy, SciPy, matplotlib and seaborn. I used ObsPy ( The thesis was written in LATEX and the PGF and TikZ packages References. 189 .. Page 18 . eignis registrieren und/oder das Signal-Rausch-Verhältnis sehr .. 0. -10. -20. -30. Elevation [km]. DGMK stations. Manual locations. 50m. 50m.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.