Machine Learning Based Real-Time Earthquake Signal Prediction
A Master of Science thesis in Mechatronics Engineering by Sara Tellab entitled, “Machine Learning Based Real-Time Earthquake Signal Prediction”, submitted in November 2020. Thesis advisor is Dr. Usman Tariq and thesis co-advisor is Dr. Mohammad AlHamaydeh. Soft copy is available (Thesis, Completion...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| التنسيق: | doctoralThesis |
| منشور في: |
2020
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | http://hdl.handle.net/11073/21367 |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| _version_ | 1864513440813416448 |
|---|---|
| author | Tellab, Sara |
| author_facet | Tellab, Sara |
| author_role | author |
| dc.contributor.none.fl_str_mv | Tariq, Usman AlHamaydeh, Mohammad |
| dc.creator.none.fl_str_mv | Tellab, Sara |
| dc.date.none.fl_str_mv | 2020-11 2021-03-16T07:16:45Z 2021-03-16T07:16:45Z |
| dc.format.none.fl_str_mv | application/pdf application/pdf application/pdf |
| dc.identifier.none.fl_str_mv | 35.232-2020.43 http://hdl.handle.net/11073/21367 |
| dc.language.none.fl_str_mv | en_US |
| dc.subject.none.fl_str_mv | Machine Learning Ground Motion Prediction NGAWest2 Database |
| dc.title.none.fl_str_mv | Machine Learning Based Real-Time Earthquake Signal Prediction |
| dc.type.none.fl_str_mv | info:eu-repo/semantics/publishedVersion info:eu-repo/semantics/doctoralThesis |
| description | A Master of Science thesis in Mechatronics Engineering by Sara Tellab entitled, “Machine Learning Based Real-Time Earthquake Signal Prediction”, submitted in November 2020. Thesis advisor is Dr. Usman Tariq and thesis co-advisor is Dr. Mohammad AlHamaydeh. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form). |
| format | doctoralThesis |
| id | aus_65e8c38372942a6c73cb4719a3fd7976 |
| identifier_str_mv | 35.232-2020.43 |
| language_invalid_str_mv | en_US |
| network_acronym_str | aus |
| network_name_str | aus |
| oai_identifier_str | oai:repository.aus.edu:11073/21367 |
| publishDate | 2020 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| spelling | Machine Learning Based Real-Time Earthquake Signal PredictionTellab, SaraMachine LearningGround Motion PredictionNGAWest2 DatabaseA Master of Science thesis in Mechatronics Engineering by Sara Tellab entitled, “Machine Learning Based Real-Time Earthquake Signal Prediction”, submitted in November 2020. Thesis advisor is Dr. Usman Tariq and thesis co-advisor is Dr. Mohammad AlHamaydeh. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).Processing the ground motion signal at an early stage is beneficial for issuing warnings, applying corrective measures and deploying first-responders teams, etc. As an earthquake starts, our proposed machine learning systems take in the first arriving points of a ground acceleration signal and predict the upcoming points in all three axes. The training, validation and testing data is acquired from the Pacific Earthquake Engineering Research Center (PEER) NGAWest2 database. It includes shallow crustal earthquakes with hypocenters less than 20 km deep. The research methodology applies different aspects of supervised and unsupervised learning. We analyze the metadata of previous earthquake records such as the magnitude, horizontal distance and peak ground acceleration (PGA). Moreover, we train various structures of artificial neural networks (ANNs) such as convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory (LSTM) networks and CNNLSTMs. The ANN model serves as a baseline for performance evaluation of the other models. We rely on the sliding window approach to split the acceleration signal. It was found that the best model for short term prediction was the LSTM model for a prediction horizon of ten timesteps. It yielded a root mean squared error (RMSE) of 8.43e6 which is a 95.2% improvement in performance compared to the baseline that yielded an RMSE of 1.74e4 . In addition, the prediction time for the CNN model is 0.49 , which makes it the fastest model. Moreover, the CNN, ANN and CNNLSTM models experimented with in this work, yielded real-time performance. The other models can also produce faster predictions using more GPUs or a supercomputer.College of EngineeringMultidisciplinary ProgramsMaster of Science in Mechatronics Engineering (MSMTR)Tariq, UsmanAlHamaydeh, Mohammad2021-03-16T07:16:45Z2021-03-16T07:16:45Z2020-11info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisapplication/pdfapplication/pdfapplication/pdf35.232-2020.43http://hdl.handle.net/11073/21367en_USoai:repository.aus.edu:11073/213672025-06-26T12:22:57Z |
| spellingShingle | Machine Learning Based Real-Time Earthquake Signal Prediction Tellab, Sara Machine Learning Ground Motion Prediction NGAWest2 Database |
| status_str | publishedVersion |
| title | Machine Learning Based Real-Time Earthquake Signal Prediction |
| title_full | Machine Learning Based Real-Time Earthquake Signal Prediction |
| title_fullStr | Machine Learning Based Real-Time Earthquake Signal Prediction |
| title_full_unstemmed | Machine Learning Based Real-Time Earthquake Signal Prediction |
| title_short | Machine Learning Based Real-Time Earthquake Signal Prediction |
| title_sort | Machine Learning Based Real-Time Earthquake Signal Prediction |
| topic | Machine Learning Ground Motion Prediction NGAWest2 Database |
| url | http://hdl.handle.net/11073/21367 |