Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Spatiotemporal Deep Learning for Bridge Response Forecasting
Accurate prediction/forecasting of the future response of civil infrastructure plays an essential role in health monitoring and safety assessment. However, the complex latent dynamics within the field sensing measurements makes the forecasting task challenging. To this end, this paper leverages the recent advances in deep learning and proposes a spatiotemporal learning framework to forecast structural responses with strong temporal dependencies and spatial correlations. The key concept is to establish a convolutional long-short term memory (ConvLSTM) network to learn spatiotemporal latent features from data and thus establish a surrogate model for structural response forecasting. The proposed approach is applied to predict the strain response for a concrete bridge with over three-year measurements available. A comparative study is also conducted against a traditional temporal-only network to highlight the forecasting performance of the proposed approach. Convincing results demonstrate that the ConvLSTM approach is a promising, reliable, and computationally efficient approach that is capable of accurately forecasting the dynamical response of civil infrastructure in a data-driven manner.
Spatiotemporal Deep Learning for Bridge Response Forecasting
Accurate prediction/forecasting of the future response of civil infrastructure plays an essential role in health monitoring and safety assessment. However, the complex latent dynamics within the field sensing measurements makes the forecasting task challenging. To this end, this paper leverages the recent advances in deep learning and proposes a spatiotemporal learning framework to forecast structural responses with strong temporal dependencies and spatial correlations. The key concept is to establish a convolutional long-short term memory (ConvLSTM) network to learn spatiotemporal latent features from data and thus establish a surrogate model for structural response forecasting. The proposed approach is applied to predict the strain response for a concrete bridge with over three-year measurements available. A comparative study is also conducted against a traditional temporal-only network to highlight the forecasting performance of the proposed approach. Convincing results demonstrate that the ConvLSTM approach is a promising, reliable, and computationally efficient approach that is capable of accurately forecasting the dynamical response of civil infrastructure in a data-driven manner.
Spatiotemporal Deep Learning for Bridge Response Forecasting
Zhang, Ruiyang (Autor:in) / Meng, Libo (Autor:in) / Mao, Zhu (Autor:in) / Sun, Hao (Autor:in)
31.03.2021
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
A Spatiotemporal Deep Learning Approach for Urban Pluvial Flood Forecasting with Multi-Source Data
| DOAJ | 2023