| 研究課題/領域番号 |
18K04320
|
|---|
| 研究機関 | 横浜国立大学 |
|---|
研究代表者 |
シリンゴリンゴ ディオン 横浜国立大学, 先端科学高等研究院, 特任教員(准教授) (60649507)
|
|---|
| 研究期間 (年度) |
2018-04-01 – 2021-03-31
|
| キーワード | System Identification / Seismic Records / Sparse Representation / Wavelet Decomposition / Seismic Isolated Bridge |
|---|
| 研究実績の概要 |
Structural identification using sparse representation method with wavelet transform for time-varying identification was conducted. The method is based on continuous & discrete wavelet transforms(CWT-DWT) and sparse representation technique(SRT). Verification was conduced by finite element model with the time-varying condition of multi-span continuous seismically isolated bridge under earthquake is used as case study. The method is used to predict malfunction of seismic bearing using features developed based on CWT and DWT. The features can track changes in structural properties caused by bearing malfunction. Furthermore clustering technique of the selected features in SRT was developed to classify the clusters with bearing condition based on seismic response of pier and girder.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
The study has established relationship between structural condition with structural response by means of structural features extracted by CWT,DWT and SRT methods. It has been shown that the extracted features can be used as indicator structural condition assessment. The next step is develop classification strategy for these features by means of novel learning method such that identification of structural condition can be conducted automatically.
|
| 今後の研究の推進方策 |
In the third year, the research will focus on:
(1) Implementation of sparse-based structural identification method using FEM-generated data from a benchmark long-span bridge structure under various scenarios related to damage. More cases of Features Extraction and Unsupervised Learning Novelty Detection (ULND)will be tested. In this simulation, conditions of seismic response established in previous task and features associated with bridge condition. The parameters used in the Unsupervised Learning Method need
training to obtain optimum performance.
(2) Implementation of the method to real full scale seismic monitoring data.
(3) Dissemination of research results and paper writing.
|
| 次年度使用額が生じた理由 |
1. Software related expenses for conducting numerical simulations using extensive computer model of long-span bridge.
2. Attending conferences and workshop for dissemination of research
3. Site visit and gathering of data related to research.
|
