|
| | Abstract | | We applied a convolutional neural network for detection of first arrival time of seismic wave to natural earthquake data observed in Hachijojima. The network has been developed using the Deep Learning by CALTEC, (Ross et al., 2018). Hachijojima data was obtained by our observation system that our university has installed for volcanic disaster prevention. This network has been made up with some convolutional layers and fully connected layers and predicts respective first arrivals of P-wave and S-wave independently from seismic records. Applied results to Hachijojima data show a good match with published from JMA and NIED in the case of relatively big earthquake. However, in the other applied results, unnatural predictions were often found from the viewpoint of pairing of P-wave and S-wave. Our presentation provides some detailed explanation about this network and also provides future development for exact detection of first arrival of seismic wave. |
|
|
|
| |
|