| Title |
Deflection Prediction of Girdersusing Bi-LSTM Network Based on
Strain History Data |
| Authors |
박상원(Park, Sangwon) ; 장민우(Chang, Minwoo) ; 야즈단파나 오미드(Yazdanpanah, Omid) |
| DOI |
https://doi.org/10.12652/Ksce.2024.44.6.0743 |
| Keywords |
양뱡향 LSTM; 매개변수; 처짐; 변형률; 하중 재하시험 Bidirectional LSTM; Paramater; Vertical deflection; Strain sensor; Steady load test |
| Abstract |
Precise diagnosis of structural condition is essential indicator for the sustained operation of infrastructure. The load tests are widely used to assess the serviceability and safety condition of bridges that are continually exposed to various forms of repeated loading. Recently, advancement in smart diagnostic technologies have introduced methods for assessing structural responses using image data. However, when utilizing image data, challenges such as distortions from lighting and angles, as well as difficulties in collecting information from blind spots, can hinder the comprehensive understanding of structural behavior. To address these limitations, this study presents an artificial intelligence (AI) based method for estimating vertical deflections using strain data measured during static tests on girders. The strain data were obtained from load tests conducted on 11 girders at the Hybrid Structural Testing Center of Myongji University. The deep learning model was developed using a bidirectional Long Short-Term Memory (LSTM) network, with multiple input parameters including girder type, static load, and strain data. The performance of AI model was further enhanced through hyperparameter optimization, and its accuracy was verified through consistency evaluation. Additionally, the effectiveness of the proposed method was evaluated by analyzing prediction accuracy and computational time in relation to the quantity of training data. |