2023 Fiscal Year Final Research Report
Fundamental study on structural damage detection applying machine learning methods to sensor data
| Project/Area Number |
19K04583
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 22020:Structure engineering and earthquake engineering-related
|
|---|
| Research Institution | Tokai University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2019-04-01 – 2024-03-31
|
| Keywords | 損傷検知 / 機械学習 / センサーデータ / オートエンコーダ / 振動実験 |
|---|
| Outline of Final Research Achievements |
Assuming that structural response data (vibration data) is obtained by instrumented sensors, it is possible to automatically and immediately detect damage to structures by applying autoencoders as a machine learning method.
Specifically, damage to the structure was simulated by reducing stiffness, and an attempt was made to detect responses with damage from learning responses without damage. As a further application, it was shown that it is possible to obtain criteria for determining whether a structure should be immediately taken out of service by having a structure equipped with a strong motion seismometer learn the response to small and medium earthquake motions that occur sometimes, and then applying an autoencoder to the nonlinear response of the structure when a large earthquake occurs.
|
| Free Research Field |
地震工学・構造工学
|
| Academic Significance and Societal Importance of the Research Achievements |
構造物の損傷検知の必要性は，インフラメンテナンスのように，損傷が徐々に進行する場合と，例えば，地震などの災害時のように急激に進行する場合が考えられるが，いずれの場合においても対応する技術者が不足する今日の状況の中，本研究成果を適用することで，自動，かつ，即時に損傷検知が可能となる．本研究は構造物の損傷を検知し，次の詳細検査段階へと進めるための１次スクリーニング手法として有益であると考えられ，これにより，技術者不足の問題の解消につながることが期待される．
|
