Paper Link: https://ieeexplore.ieee.org/document/9328806

DOI: 10.1109/TSMC.2020.3048696

Being able to forecast structural response (displacement) for a short period of time in the future is very crucial for safety of civil infrasturctures under extreme condidtions, like earthquakes. We developed an RNN encoder-decoder structure with temporal attention for sequence forecasting.

Architecture of the proposed deep learning network.

An example of forecasted results for a time window.