P-WAVE DETECTION USING DEEP LEARNING IN TIME AND FREQUENCY DOMAIN FOR IMBALANCED DATASET

• Main Author: Aditya Sugondo, Rhesa
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55253
• Institution: Institut Teknologi Bandung

Early Tsunami and earthquake warning system needs a good Automatic First Arrival Picking(AFAP) Sub-System to determine the earthquake arrival time. This sub-system has a time-domain earthquake signal as the input and the arrival time of the earthquake as the output. There are several methods of AFAP that are used widely nowadays, one of them is Short-Term Average/Long-Term Average (STA/LTA) fused with the Auto-Regressive Coefficient (AR-AIC) method. Even though this method is real-time, its performance still relatively low. With similar characteristics between the seismic signals with image data, utilizing Deep Learning on AFAP can further increase its performance. The seismogram channels can be seen as the image height and the signal at a certain window can be seen as the image width. Unfortunately, these image data will be considered as an imbalanced dataset because the amount of P-wave data are less than the non Pwave data. Proposed in this research, a Deep Learning with Time domain and Frequency domain as inputs with SMOTE oversampling method. Deep Learning is used because of its ability to generalize well on a huge dataset while SMOTE is used to overcome the imbalanced dataset problem. With SMOTE, the amount of Pwave data will increase without using duplication. With this proposed system, the accuracy is 99.3%, the Root Mean Square Error is 0.202 seconds, and the maximum execution time is 0.17 seconds with the periodic time of 0.4 seconds. With those results, the AFAP system has good results for estimating the first arrival earthquake time.