An Information‐Density‐Aware Strategy Search Framework for Dam Displacement Reconstruction: Dynamically Synergizing HST‐Inspired Sparse Parametric Modeling and Data‐Driven Nonlinear Mapping
Structural Control and Health Monitoring · 2026-01 · 2026 (1)
DOI: 10.1155/stc/3048344 ↗Kurz gefasst
Fachartikel in der Zeitschrift Structural Control and Health Monitoring, online erschienen am 19.11.2026. Verfasst von Minshui Huang und 4 weiteren Autor:innen. Die Arbeit nennt 61 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Guangxi University.
Abstract (Originalsprache)
Dam displacement monitoring data have varying sampling intervals and complex missing patterns, which are difficult to accommodate simultaneously using existing methods. An information‐density‐aware strategy search framework (ADAF‐C) is proposed. HST‐inspired environmental features are used as inputs to parallel polynomial LASSO and CatBoost branches. Sparse parametric and nonlinear nonparametric representations are established, respectively, through pairwise interaction expansion with L1 regularization and symmetric decision trees with gradient boosting. The branch predictions are adaptively fused and further corrected through multiview systematic correction. A search over the numerical parameters is performed using CMA‐ES, while fusion modes and applicable correction strategies are evaluated using the mean RMSE from expanding‐window temporal validation. Monitoring data from two in‐service concrete dams are used for validation. The original 1‐day series of Dataset 2 is uniformly downsampled to construct 3‐day, 5‐day, and 7‐day series for same‐dam sampling‐interval analysis, while Dataset 1 with a 15‐day sampling interval is treated as an independent engineering case. SVR, RF, LightGBM, and LSTM are tuned using training‐only chronological validation. The lowest mean RMSE and highest mean R 2 are reported for ADAF‐C on all three test sets of both datasets. Relative to the baseline with the lowest mean RMSE, reductions of 20.47%–28.85% are obtained for Dataset 1 and 30.65% and 11.47% for Dataset 2 under continuous and random missing, respectively. Under periodic missing in Dataset 2, a mean RMSE of 0.0865 mm is obtained with ADAF‐C, close to 0.0867 mm with linear interpolation. Fusion modes are selected according to sampling interval and missing pattern. At the 5‐day interval, the best transferred RMSE obtained without rerunning CMA‐ES is 29.21%–165.63% above the CMA‐ES reference, indicating limited transferability. In the training‐input noise experiment, the mean RMSE of ADAF‐C is lower than that of LightGBM at SNR ≥ 10 dB, while performance decreases at 5 dB.
Förderung
- National Natural Science Foundation of China
- Guangxi University
Zitieren
APA
Huang, M., Xu, X., Wan, N., Deng, Z., & Gu, J. (2026). An Information‐Density‐Aware Strategy Search Framework for Dam Displacement Reconstruction: Dynamically Synergizing HST‐Inspired Sparse Parametric Modeling and Data‐Driven Nonlinear Mapping. Structural Control and Health Monitoring, 2026(1). https://doi.org/10.1155/stc/3048344
Vancouver
Huang M, Xu X, Wan N, Deng Z, Gu J. An Information‐Density‐Aware Strategy Search Framework for Dam Displacement Reconstruction: Dynamically Synergizing HST‐Inspired Sparse Parametric Modeling and Data‐Driven Nonlinear Mapping. Structural Control and Health Monitoring. 2026;2026(1). doi:10.1155/stc/3048344
BibTeX
@article{Huang2026,
author = {Huang, Minshui and Xu, Xiaoyu and Wan, Neng and Deng, Zhihang and Gu, Jianfeng},
title = {An Information‐Density‐Aware Strategy Search Framework for Dam Displacement Reconstruction: Dynamically Synergizing HST‐Inspired Sparse Parametric Modeling and Data‐Driven Nonlinear Mapping},
journal = {Structural Control and Health Monitoring},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/stc/3048344}
}
EUVIMED