Cross-fidelity nonlinear dynamic response predictions of steel frame buildings using CNN-LSTM deep learning models with transformer and attention mechanisms
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Seismic responses of building frames can be predicted using simplistic low fidelity (e.g., equivalent single-degree-of-freedom mass–spring–dashpot systems) or material mechanics-based high fidelity (e.g., fiber-section beam column or solid element finite element models) numerical models with a trade-off between prediction accuracy and computational efficiency. While low fidelity models have inherent limitations, their embedded computational efficiency and physics mechanism can be leveraged to couple with data-driven approaches to achieve high-fidelity seismic response predictions. This paper develops a novel cross-fidelity deep learning (DL) framework, which combines seismic ground motions (GM) and low fidelity structural responses as complementary inputs, to improve the accuracy and robustness in predicting high-fidelity nonlinear seismic responses of different steel frame buildings. The proposed models utilize hybrid architectures that integrate convolutional neural networks (CNN), long short-term memory (LSTM), transformer, and self-attention mechanisms to effectively capture time–frequency–magnitude dependencies inherent in seismic response data. Performance of these models is evaluated on three representative steel frame buildings in California and compared against six GM single-input DL models, as well as three dual-input models without having the CNN module. The proposed DL models with hybrid architectures and the cross-fidelity input mechanism consistently outperform other models, demonstrating significantly improved effectiveness in predicting the entire dynamic response history. Results indicate that integrating low-fidelity model responses as physics-guided inputs reduces prediction variance and enhances the reliability of time-series inference. This study highlights the potential of the proposed cross-fidelity DL approaches for improving seismic response predictions, which could be utilized to support downstream applications such as seismic risk assessment, rapid post-earthquake evaluation, and performance-based seismic design.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it