Classifiers based on machine learning algorithms trained through hybrid strategies have been proposed for structural health monitoring (SHM) of bridges. Hybrid strategies use numerical and monitoring data together to improve the learning process of the algorithms. The numerical models, such as finite-element (FE) models, are used for data augmentation based on the assumption of the existence of limited experimental data sets. However, a numerical model might fail in providing reliable data, as its parameters might not share the same underlying operating conditions observed in real situations. Meanwhile, the concept of transfer learning has evolved in SHM, in particular through domain adaptation techniques. The ability to adapt a classifier built on a well-known labeled data set to a new scenario with an unlabeled data set is an opportunity to transit bridge SHM from research to practice. Therefore, this paper proposes an unsupervised transfer learning approach for bridges with a domain adaptation technique, where classifiers are trained only with labeled data generated from FE models (source domain). Then, unlabeled monitoring data (target domain) are used to test the classification performance. As numerical and monitoring data are related to the same bridge, both domains are assumed to have similar statistical distributions, with slight differences caused by the uncertainties inherent to the FE models. The domain adaptation is performed using a transfer knowledge method called transfer component analysis, which transforms damage-sensitive features from the original space to a new one, called latent space, where the differences between feature distributions are reduced. This approach may increase the use of numerical modeling for long-term monitoring, as it overcomes some of the limitations imposed by the calibration process of FE models. The efficiency of this unsupervised approach is illustrated through the classification performance of classifiers built on source data with and without domain adaptation, and using the benchmark data sets from the Z-24 Bridge as the target data.
16 September 2023 in Mechanical Systems and Signal Processing
Bridges are built to last more than 100 years, spanning many human generations. Throughout their lifetime, their service requirements may change, or they age and often suffer a material degradation process that can lead to the need of retrofitting. In bridge engineering, retrofitting refers to the strengthening of existing structures to make them more resistant and to increase the lifespan of bridges. Retrofitting normally increases the stiffness of bridge components, which can cause significant changes in the global modal properties. In the context of structural health monitoring, a classifier trained with datasets before retrofitting will most likely output many outliers after retrofitting, based on the premise that the new observations do not share the same underlying distribution. Therefore, how can long-term monitoring data from one bridge (labeled source domain) be reused to create a classifier that generalizes to the same bridge after retrofitting (unlabeled target domain)? This paper presents a novel approach based on transfer learning in the context of domain adaptation on datasets from two real bridges subjected to retrofit and under-monitoring programs. Based on the assumption that both bridges are undamaged before retrofitting, the results show that transfer learning can support the long-term damage detection process based on a classification using an outlier detection strategy.
4 June 2025 in Mechanical Systems and Signal Processing
Machine learning methods used in Structural Health Monitoring applications still have generalization difficulties among structures, even when structures are nominally and topologically similar. The data sets present divergences between their probability distributions that do not allow the model’s generalization for damage detection. This issue is even more complex in situations where one wants to quantify damage levels through data sets collected from different structures. Transfer learning methods offer a solution to overcome those limitations, using relevant information from a labeled structure (source domain) to assist the analysis of another structure (target domain) under unknown conditions. Therefore, this paper proposes the use of transfer component analysis to mitigate divergences between the model/structure’s features, and the label consistency requirement is applied in combination with a Gaussian process regression model for damage quantification. The effectiveness of the estimated model improves when the labels consistency between domains is achieved, indicating the current damage level in the structure when the regression model achieves its best performance (lowest error). The proposed methodology is applied on the benchmark data of a three-story building structure from the Los Alamos National Laboratory using the knowledge from its numerical model under several conditions, where the complete information of its behavior is available. The results compare the analysis in the original space and after applying the proposed methodology, demonstrating an improvement of the performance in the damage detection and quantification steps.