Utilizing a Machine Learning-Based Surrogate Model for Structural Reliability Analysis
Résumé
Structural Reliability Analysis (SRA) is a crucial area of research in civil and mechanical engineering. One of the primary challenges of this field is accurately quantifying the uncertainty of complex physicalsystems. However, traditional methods can be computationally expensive as they rely on information about the Limit State Function (LSF). To address this issue, this study uses a machine learning-based surrogate model for reliability analysis. The basic idea is to develop a model equivalent to the performance function. This research adopts a Machine Learning-based surrogate model to assess the propagation of structural uncertainties with Tuned Mass Dampers (TMD). Various machine learning models, including Random Forest, Gradient Boosting, Xtrem Gradient Boosting, and Neural Network, are used for prediction, treating the task as a regression problem. The study explores different structures with TMD, considering the effects of longitudinal vibration and transverse shear in a beam like structure. The researchers conducted dynamic response simulations under seismic excitation using a beam-like model. Analytical and numerical examples are presented to evaluate the benchmarking performance of different machine-learning methods. The simulation results confirm the efficiency of the proposed surrogate mechanics.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |