No abstract

No abstract

No abstract

The structural health monitoring relies on the continuous observation of a dynamic system over time to identify its actual condition, detect abnormal behaviors, and predict future states. The regular changes in environmental factors have been reported as one of the main challenges for the application of structural health monitoring systems. These influences in the structural responses are in general nonlinear, affecting the damage-sensitive features in the most varied forms. The usual process to remove these normal changes is referred to as data normalization. In that regard, principal component analysis is probably the most studied algorithm in structural health monitoring, having numerous versions to learn strong nonlinear normal changes. However, in most cases, not all variability is properly accounted for via the existing nonlinear principal component analysis approaches, resulting in poor damage detection and quantification performances. In this article, a new paradigm based on deep principal component analysis, rooted in the deep learning field, is presented to overcome these limitations. This approach extracts the most salient underlying feature distributions by stacking multiple feedforward neural networks trained to learn an identity mapping of the input variables, where the network inputs are reproduced into the outputs. Similar to the traditional nonlinear principal component analysis–based approach, our approach identifies a nonlinear output-only model of an undamaged structure by comprising modal features into an internal bottleneck layer, which implicitly represents the independent environmental factors. The proposed technique is validated through the application on a progressively damaged prestressed concrete bridge and a three-span suspension bridge. The experimental results demonstrate that capturing the most slight nonlinear variations in the data can lead to improved data normalization and, consequently, better damage detection and quantification performances.

During the service life of bridges, the bridge management systems (BMSs) seek to handle all performed assessment activities by controlling regular inspections, evaluations, and maintenance of these structures. However, the BMSs still rely heavily on qualitative and visual bridge inspections, which compromise the structural evaluation and, consequently, the maintenance decisions as well as the avoidance of bridge collapses. The structural health monitoring appears as a natural field to aid the bridge management, providing more reliable and quantitative information. Herein, the machine learning algorithms have been used to unveil structural anomalies from monitoring data. In particular, the Gaussian mixture models (GMMs), supported by the expectation-maximization (EM) on the parameter estimation, have been proposed to model the main clusters that correspond to the normal and stable state conditions of a bridge, even when it is affected by unknown sources of operational and environmental variations. Unfortunately, the performance of the EM algorithm is strongly dependent on the choice of the initial parameters. This paper proposes a hybrid approach based on a standard genetic algorithm (GA) to improve the stability of the EM algorithm on the searching of the optimal number of clusters and their parameters, strengthening the damage classification performance. The superiority of the GA-EM-GMM approach, over the classic EM-GMM one, is tested on a damage detection strategy implemented through the Mahalanobis-squared distance, which permits one to track the outlier formation in relation to the chosen main group of states, using real-world data sets from the Z-24 Bridge, in Switzerland.

No abstract

During the service life of engineering structures, structural management systems attempt to manage all the information derived from regular inspections, evaluations and maintenance activities. However, the structural management systems still rely deeply on qualitative and visual inspections, which may impact the structural evaluation and, consequently, the maintenance decisions as well as the avoidance of collapses. Meanwhile, structural health monitoring arises as an effective discipline to aid the structural management, providing more reliable and quantitative information; herein, the machine learning algorithms have been implemented to expose structural anomalies from monitoring data. In particular, the Gaussian mixture models, supported by the expectation-maximization (EM) algorithm for parameter estimation, have been proposed to model the main clusters that correspond to the normal and stable state conditions of a structure when influenced by several sources of operational and environmental variations. Unfortunately, the optimal parameters determined by the EM algorithm are heavily dependent on the choice of the initial parameters. Therefore, this paper proposes a memetic algorithm based on particle swarm optimization (PSO) to improve the stability and reliability of the EM algorithm, a global EM (GEM-PSO), in searching for the optimal number of components (or data clusters) and their parameters, which enhances the damage classification performance. The superiority of the GEM-PSO approach over the state-of-the-art ones is attested on damage detection strategies implemented through the Mahalanobis and Euclidean distances, which permit one to track the outlier formation in relation to the main clusters, using real-world data sets from the Z-24 Bridge (Switzerland) and Tamar Bridge (United Kingdom).

Summary In most real‐world monitoring scenarios, the lack of measurements from damaged conditions requires the application of unsupervised approaches, mainly the ones based on modal features estimated from raw vibration data through traditional system identification methods. Although numerous successful applications using modal parameters have been reported, they have demonstrated to be insufficient to estimate a robust set of damage‐sensitive features. Inspired by the idea of compressed sensing and deep learning, an intelligent two‐level feature extraction approach using stacked autoencoders over pre‐processed vibration data is proposed. This procedure can improve the performance of traditional damage detection classifiers by compressing modal parameters into a smaller set of highly informative features when considering information entropy metrics. The proposed technique demonstrates significant improvement in the performance of damage detection and classification approaches when evaluated on the well‐known monitoring data sets from the Z‐24 Bridge, where several damage scenarios were carried out under rigorous operational and environmental effects.