2019 Fiscal Year Final Research Report

Applicability of machine learning to tsunami source estimation for exploring devastated area

Research Project

PDF

Project/Area Number	17H03316
Research Category	Grant-in-Aid for Scientific Research (B)
Allocation Type	Single-year Grants
Section	一般
Research Field	Hydraulic engineering
Research Institution	Kansai University
Principal Investigator	Takahashi Tomoyuki 関西大学, 社会安全学部, 教授 (40261599)
Co-Investigator(Kenkyū-buntansha)	奥村与志弘関西大学, 社会安全学部, 准教授 (80514124) 河野和宏関西大学, 社会安全学部, 准教授 (60581238)
Project Period (FY)	2017-04-01 – 2020-03-31
Keywords	機械学習 / ニューラルネットワーク / 津波波源断層モデル / 津波シミュレーション / 津波観測 / 津波被害
Outline of Final Research Achievements	For effective disaster management, prompt and accurate tsunami source estimation is important not only for updating tsunami warnings but also for determining emergency responses. Although tsunami waveform inversion is a powerful tool for estimating tsunami sources, expert analysis of a large amount of data by trial-and-error methods is needed. Machine learning is an effective means of extracting features from complex big data and determining underlying laws related to different phenomena. In the study, a model to estimate tsunami source from observed data by using machine learning was developed. Using the deep learning algorithm, a neural network model was constructed. The training data were based on a dataset generated using the fault models and the virtually observed water level changes. The trained model was validated by examining whether it could determine the tsunami source parameters for the different scenarios from the virtually observed data.
Free Research Field	水災害
Academic Significance and Societal Importance of the Research Achievements	巨大津波災害においては、様々な物理的・人的要因が複雑に影響しあって被害を拡大させる。よって、大量の複雑な情報を分析する必要があり、人工知能の活用が期待される分野であるが、現状ではほとんど導入されていない。本研究は津波研究の中心となる津波波源に対して機械学習の適用を試みたものであり、その学術的意義は高いと考えられる。また、津波波源は津波警報やハザードマップなどの津波防災技術の基礎になるものである。よって、津波波源を機械学習により迅速に調べることができるようになれば、津波防災実務への貢献も期待される。