臺灣博碩士論文加值系統

台灣位於歐亞板塊和菲律賓海板塊的交界處，平均每年有 3 萬多次地震發生。 因此，發展可靠的地震偵測方法是非常重要的。但是，傳統方法很容易發生誤報 或漏失偵測，必須定義一些其他的規則或是人工監督的方式來提高精準度。此外， 在吵雜環境中，傳統方法用於地震偵測的準確性也會降低。然而，地震的歷史資 料是非常豐富的，適合利用機器學習的演算法來偵測地震。為了解決這些傳統方 法有的缺點，本文提出兩種基於遞歸神經網絡模型用於地震偵測。神經網絡可以 從資料學習一些有用的規則而不是由人工定義來偵測地震，並能容忍於嘈雜的環 境。這些模型可以即時的偵測 P 波，S 波，和地震事件。有別於傳統神經網路用 於地震偵測方法，本文提出的模型不需要以一段波形為輸入，且可用於即時偵 測。在實驗中，不同架構的模型被比較，且結果顯示模型可於嘈雜的環境中偵測 地震。此外，模型比較於基於準則的地震偵測演算法且呈現較好的精準度。而且， 當地震波到達時，模型可以在平均 0.289 秒後偵測出地震，因此此模型非常適合 用於地震預警系統來發布地震警報。

Since Taiwan is located at the junction of the Eurasian Plate and the Philippine Sea Plate, on average more than 30 thousand earthquakes occur each year around Taiwan. Developing a reliable method for earthquake detection is very important. Nonetheless, false alerts and missing detection are prone to occur by traditional methods. Therefore, traditional methods need human supervision at all time or other criteria to verify the detection. The other shortcoming is that they will decrease the accuracy of earthquake detection in a noisy environment. Nevertheless, seismic data are plentiful. It is suitable to exploit machine learning techniques to detect earthquakes. To address these problems from traditional methods, this thesis proposed a recurrent neural network-based framework for earthquake detection. The neural network can learn some useful rules from data rather than manual design to detect earthquakes and has toleration in a noisy environment. The models are capable of detecting the P-wave, the S-wave, and the event in real-time. Different from the traditional neural network/machine learning-based methods for seismic detection, the models proposed in this thesis do not need to input a segment of waveforms and can be used for real-time detection. In experiments, different architectures are compared and the results show that these models are capable of detecting earthquake events in a noisy environment. In addition, the models are compared with the criterion-based algorithm and perform better accuracy of earthquake detection. Moreover, the models can detect earthquake events in an average of 0.289s after the P-wave arrival. It is suitable to apply for earthquake early warnings (EEW) system to issue warnings.

Abstract in Chinese .................................. iii
Abstract in English .................................. iv
Acknowledgements.................................. v
Contents........................................ vi
List of Figures..................................... ix
List of Tables ..................................... xi
1 Introduction.................................... 1
1.1 Background................................. 1
1.2 Motivation.................................. 2
1.3 Contributions ................................ 3
1.4 Organization ................................ 4
2 Related Work ................................... 5
2.1 Earthquake Detection............................ 5
2.2 Earthquake Early Warning......................... 6
2.3 Machine Learning Techniques for Earthquake Detection ......... 7
3 System Model................................... 10
3.1 The Architecture of Models......................... 10
3.1.1 Input Layer............................. 10
3.1.2 Hidden Layer............................ 12
3.1.3 Output Layer ............................ 13
3.2 Training and Optimization ......................... 14
4 Configuration ................................... 16
4.1 Dataset ................................... 16
4.2 Evaluation Metrics ............................. 20
4.3 Features Selection.............................. 22
4.4 Number of Neurons and Layers....................... 24
5 Experiments.................................... 26
5.1 Training Phase ............................... 26
5.2 Preliminary Results............................. 27
5.2.1 Single Model-p........................... 29
5.2.2 Single Model-s ........................... 32
5.2.3 Single Model-event......................... 35
5.2.4 Mixed Model............................ 37
5.3 Case Study ................................. 39
5.4 ROC Curve ................................. 41
5.5 Time Residuals ............................... 43
5.6 Comparison with the Criterion-based Algorithm . . . . . . . . . . . . . . 44
6 Conclusion..................................... 47
References....................................... 48
Letter of Authority .................................. 51
AppendixA...................................... 52
AppendixB...................................... 75

[1] R. V. Allen, “Automatic earthquake recognition and timing from single traces,” Bulletin of the Seis- mological Society of America, vol. 68, no. 5, p. 1521, 1978.
[2] R. Allen, “Automatic phase pickers: Their present use and future prospects,” Bulletin of the Seismo- logical Society of America, vol. 72, no. 6B, p. S225, 1982.
[3] B. K. Sharma, A. Kumar, and V. M. Murthy, “Evaluation of seismic events detection algorithms,” Journal of the Geological Society of India, vol. 75, pp. 533–538, Mar 2010.
[4] Z. E. Ross and Y. Ben-Zion, “Automatic picking of direct p, s seismic phases and fault zone head waves,” Geophysical Journal International, vol. 199, no. 1, pp. 368–381, 2014.
[5] A.Krizhevsky,I.Sutskever,andG.E.Hinton,“Imagenetclassificationwithdeepconvolutionalneural networks,” in Advances in neural information processing systems, pp. 1097–1105, 2012.
[6] Y.LeCun,L.Bottou,Y.Bengio,andP.Haffner,“Gradient-basedlearningappliedtodocumentrecog- nition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[7] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning representations by back-propagating errors,” nature, vol. 323, p. 533, 1986.
[8] F.U.Dowla,S.R.Taylor,andR.W.Anderson,“Seismicdiscriminationwithartificialneuralnetworks: Preliminary results with regional spectral data,” Bulletin of the Seismological Society of America, vol. 80, no. 5, p. 1346, 1990.
[9] T. Perol, M. Gharbi, and M. Denolle, “Convolutional neural network for earthquake detection and location,” Science Advances, vol. 4, no. 2, p. e1700578, 2018.
[10] A. Ruano, G. Madureira, O. Barros, H. Khosravani, M. Ruano, and P. Ferreira, “Seismic detection using support vector machines,” Neurocomputing, vol. 135, pp. 273 – 283, 2014.
[11] J. Doubravová, J. Wiszniowski, and J. Horálek, “Single layer recurrent neural network for detection of swarm-like earthquakes in w-bohemia/vogtland—the method,” Computers & Geosciences, vol. 93, pp. 138–149, 2016.
[12] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997.
[13] D.-Y.Chen,N.-C.Hsiao,andY.-M.Wu,“Theearthwormbasedearthquakealarmreportingsystemin taiwan,” Bulletin of the Seismological Society of America, vol. 105, no. 2A, pp. 568–579, 2015.
[14] R. Sleeman and T. van Eck, “Robust automatic p-phase picking: an on-line implementation in the analysis of broadband seismogram recordings,” Physics of the Earth and Planetary Interiors, vol. 113, no. 1, pp. 265 – 275, 1999.
[15] C.D.Saragiotis,L.J.Hadjileontiadis,andS.M.Panas,“Pai-s/k:Arobustautomaticseismicpphase arrival identification scheme,” IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 6, pp. 1395–1404, 2002.
[16] C.Baillard,W.C.Crawford,V.Ballu,C.Hibert,andA.Mangeney,“AnAutomaticKurtosis-BasedP- and S-Phase Picker Designed for Local Seismic Networks,” The Bulletin of the Seismological Society of America, vol. 104, pp. 394–409, Feb. 2014.
[17] Z.Ross,M.White,F.Vernon,andY.Ben-Zion,“Animprovedalgorithmforreal-times-wavepicking with application to the (augmented) anza network in southern california,” Bulletin of the Seismological Society of America, vol. 106, no. 5, pp. 2013–2022, 2016.
[18] R. M. Allen, P. Gasparini, O. Kamigaichi, and M. Böse, “The status of earthquake early warning around the world: An introductory overview,” Seismological Research Letters, vol. 80, no. 5, p. 682, 2009.
[19] O.Kamigaichi,M.Saito,K.Doi,T.Matsumori,S.Tsukada,K.Takeda,T.Shimoyama,K.Nakamura, M. Kiyomoto, and Y. Watanabe, “Earthquake early warning in japan: Warning the general public and future prospects,” Seismological Research Letters, vol. 80, no. 5, p. 717, 2009.
[20] Y.-M. Wu, D.-Y. Chen, T.-L. Lin, C.-Y. Hsieh, T.-L. Chin, W.-Y. Chang, W.-S. Li, and S.-H. Ker, “A high-density seismic network for earthquake early warning in taiwan based on low cost sensors,” Seismological Research Letters, vol. 84, no. 6, pp. 1048–1054, 2013.
[21] Y.-M.Wu,“Progressondevelopmentofanearthquakeearlywarningsystemusinglow-costsensors,” Pure and Applied Geophysics, vol. 172, pp. 2343–2351, Sep 2015.
[22] C.E.Johnson,A.Bittenbinder,B.Bogaert,L.Dietz,andW.Kohler,“Earthworm:Aflexibleapproach to seismic network processing,” IRIS Newsletter, vol. 14, no. 2, pp. 1–4, 1995.
[23] J.WangandT.-L.Teng,“Artificialneuralnetwork-basedseismicdetector,”BulletinoftheSeismolog- ical Society of America, vol. 85, no. 1, p. 308, 1995.
[24] S. Diersen, E.-J. Lee, D. Spears, P. Chen, and L. Wang, “Classification of seismic windows using artificial neural networks,” Procedia Computer Science, vol. 4, pp. 1572 – 1581, 2011.
[25] K.V.KislovandV.V.Gravirov,“Useofartificialneuralnetworksforclassificationofnoisyseismic signals,” Seismic Instruments, vol. 53, pp. 87–101, Jan 2017.
[26] W.ZhuandG.C.Beroza,“PhaseNet:ADeep-Neural-Network-BasedSeismicArrivalTimePicking Method,” ArXiv e-prints, Mar. 2018.
[27] Z. Li, M.-A. Meier, E. Hauksson, Z. Zhan, and J. Andrews, “Machine learning seismic wave dis- crimination: Application to earthquake early warning,” Geophysical Research Letters, vol. 45, no. 10, pp. 4773–4779, 2018.
[28] P. M. DeVries, F. Viégas, M. Wattenberg, and B. J. Meade, “Deep learning of aftershock patterns following large earthquakes,” Nature, vol. 560, no. 7720, p. 632, 2018.
[29] J.Wiszniowski,B.M.Plesiewicz,andJ.Trojanowski,“Applicationofrealtimerecurrentneuralnet- work for detection of small natural earthquakes in poland,” Acta Geophysica, vol. 62, pp. 469–485, Jun 2014.
[30] P.J.Werbos,“Backpropagationthroughtime:whatitdoesandhowtodoit,”ProceedingsoftheIEEE, vol. 78, pp. 1550–1560, Oct 1990.
[31] D.P.KingmaandJ.Ba,“Adam:Amethodforstochasticoptimization,”ArXive-prints,2014.
[32] T.LinandY.Wu,“Afastmagnitudeestimationforthe2011mw9.0greattohokuearthquake,”Seis-
mological Research Letters, vol. 83, no. 5, p. 831, 2012.
[33] D. M. Powers, “Evaluation: from precision, recall and f-measure to roc, informedness, markedness
and correlation,” Journal of Machine Learning Technologies, no. 1, pp. 37–63, 2011.
[34] M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Isard, Y. Jia, R. Jozefowicz, L. Kaiser, M. Kudlur, J. Levenberg, D. Mane, R. Monga, S. Moore, D. Murray, C. Olah, M. Schuster, J. Shlens, B. Steiner, I. Sutskever, K. Talwar, P. Tucker, V. Vanhoucke, V. Vasudevan, F. Viegas, O. Vinyals, P. Warden, M. Wattenberg, M. Wicke, Y. Yu, and X. Zheng, “TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems,” ArXiv e-prints, Mar. 2016.