Damage Detection for a Cantilevered Steel I-Beam through Deep-Learning Methods: LSTM, Multivariate Time-Series Transformer, and LSTM-Based Autoencoder
Bibliographic record
Abstract
The structural integrity of steel trusses and I-beams is of vital importance for preventing the potential collapse of steel bridges when subjected to extraordinary forces. Thus, identifying damage to I-beams, which cannot be noticed in typical inspections, based on their measured response, would enable early damage detection, and would trigger the necessary mitigation measures to restore the structural integrity of the bridge. This investigation built a vast database of structurally damaged cantilever I-beams, in which openings of various degrees and locations were placed along the beams to emulate reductions in stiffness. Both damaged and undamaged I-beams were modeled using Abaqus software, facilitated by Python scripting. Three deep-learning algorithms were trained, validated and tested with the healthy and damaged I-beam cases: long short-term memory (LSTM), a LSTM-based autoencoder, and multivariate time-series transformers (MTTs), for which the input data consisted of acceleration responses recorded at specific points on the top flange of both undamaged and damaged I-beams subjected to harmonic dynamic loads. To enhance adaptation for field monitoring data, random normal noise was introduced into the acceleration responses before running the machine learning (ML) damage identification algorithms. The three algorithms demonstrated exceptional ability to accurately distinguish between the damaged and the undamaged I-beams. Furthermore, the location of the damage on the beam was identified by the LSTM and MTT algorithms, which had the best accuracy for damage localization. Finally, a comparative analysis of the three algorithms was conducted to clarify the optimal quantity of data points required to attain reliable results.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".