Unsupervised learning-based framework for indirect structural health monitoring using adversarial autoencoder

The monitoring of a structural condition of steel bridges is an important issue. Good condition of infrastructure facilities ensures the safety and economic well-being of society. At the same time, due to the continuous development, rising wealth of the society and socio-economic integration of countries, the number of infrastructural objects is growing. Therefore, there is a need to introduce an easy-to-use and relatively low-cost method of bridge diagnostics. We can achieve these benefits by the use of Unmanned Aerial Vehicle-Based Remote Sensing and Digital Image Processing. In our study, we present a state-of-the-art framework for Structural Health Monitoring of steel bridges that involves literature review on steel bridges health monitoring, drone route planning, image acquisition, identification of visual markers that may indicate a poor condition of the structure and determining the scope of applicability. The presented framework of image processing procedure is suitable for diagnostics of steel truss riveted bridges. In our considerations, we used photographic documentation of the Fitzpatrick Bridge located in Tallassee, Alabama, USA.

Adversarial correlated autoencoder for unsupervised multi-view representation learning

An optimal sensor placement design framework for structural health monitoring using Bayes risk

Most damage-assessment strategies for dynamic systems only distinguish between undamaged and damaged conditions without recognizing the level or type of damage or considering unseen conditions. This paper proposes a novel framework for structural health monitoring (SHM) that combines supervised and unsupervised learning techniques to assess damage using a system’s structural response (e.g., the acceleration response of big infrastructures). The objective is to enhance the benefits of a supervised learning framework while addressing the challenges of working in an SHM context. The proposed framework uses a Linear Discriminant Analysis (LDA)/Probabilistic Linear Discriminant Analysis (PLDA) strategy that enables learning the distributions of known classes and the performance of probabilistic estimations on new incoming data. The methodology is developed and proposed in two versions. The first version is used in the context of controlled, conditioned monitoring or for post-damage assessment, while the second analyzes the single observational data. Both strategies are built in an automatic framework able to classify known conditions and recognize unseen damage classes, which are then used to update the classification algorithm. The proposed framework’s effectiveness is first tested considering the acceleration response of a numerically simulated 12-degree-of-freedom system. Then, the methodology’s practicality is validated further by adopting the experimental monitoring data of the benchmark study case of the Z24 bridge.

Continuous bridge monitoring generates long-duration, noisy, and multi-channel acceleration streams in which abnormal events are sparse yet safety-critical. This work targets reliable event classification under operational conditions while preserving the time-series nature of the data. A hybrid one-dimensional convolutional model with channel-wise recalibration and attention analyzes fixed-length windows from a cable-stayed bridge monitoring system, combined with peak picking, batching, and lightweight inference that suit day-scale records. Evaluation on a held-out test set shows balanced class performance: precision across classes ranges from 90.5% to 98.4%, and recall across classes ranges from 91.4% to 99.6%, yielding precision of 94.6% and recall of 94.9%. At the sensor level, classification accuracy ranges from 93.6% to 99.7%, indicating consistent behaviour across heterogeneous channels with different noise and placement conditions. The dataset is split into training, validation, and test subsets, and the test set remains untouched during model selection. The results suggest robust detection across classes and sensors, stable operation on continuous 24-h records, and practical readiness for field deployment in bridge monitoring.

In this paper, we present an Internet of Things framework for structural health monitoring. The proposed system detects the delimitation and debonding in composite concrete structures: more specifically, develop an IoT-Based Non-Destructive Test (NDT) to detect debonding between Carbon Fiber sheets and Concrete slabs. An Inter Digital Capacitance Sensor is designed and fabricated to function as the primary detection element. The sensor is embedded within an IoT node that manages the measurement process, captures the measurement location automatically, and performs essential data filtering operations with an attached graphical interface that allows basic control and early-access to the measured data. The IoT node integrates within the larger framework via wireless WiFi connection throughout which the data is transferred and control functions are administrated. The paper presents the system design details including mathematical modeling for the capacitance sensing element, finite element simulation results, and practical setup measured data.

This work proposes an optimal fiber optic sensor placement framework for structural health monitoring (SHM) applications. The framework is applied to an aircraft’s wing spar entirely made of composite materials. The damage of interest is debonding between laminates, which may cause local buckling that results in reduced structural load carrying capabilities. A high-fidelity finite element (FE) model is used as a synthetic data generator. The inputs to the model are loads and debonding damage parameters (size and location), and the outputs are uniaxial strain measurements and buckling eigenvalues. “Run time” surrogate models are created using different machine learning methods to overcome the high computational costs of each run of the physicsbased model. Then, Bayesian inference is used to estimate the damage parameters given strain measured at candidate sensor locations. These estimations are used to assess damage criticality, which is linked to buckling eigenvalues, and transformed into decisions. Bayesian optimization is used to select the candidates that minimize a utility function that considers the costs associated with making a certain decision plus the costs of acquiring and installing the SHM hardware (sensors, data acquisition system, etc.). The candidate with the lowest cost is selected. The resulting optimal sensor configuration is presented, consisting of the number of sensors to be deployed and their respective locations. The importance of defining an objective function that reflects the goal of the SHM system (e.g., maximizing the probability of detection, minimizing the probability of false alarms, or a balance of both) are also discussed.

Regular health monitoring of bridges is a vital process to prevent serious structural damage. Marker-based systems, which follow the trajectory of objects by placing a well-characterized pattern on their surface and identify them on photos or videos taken of these objects, have proven to be a cheap and flexible alternative for such tasks. In this work, we extend our previous laboratory implementation with a low-cost, fully automatic on-site installation at the bridge at Arosa Island, Galicia, Spain. Preliminary results presented in this paper show that our system is highly robust for the harsh climate of the installation site.

In recent years, the concept of bridge monitoring using indirect measurements from a passing vehicle has been rapidly developed. This concept is known as “drive-by bridge inspection”. Most of the methods proposed under this approach utilize the dynamic characteristics of the bridge as an indicator of damage, such as the natural frequency of the bridge. The natural frequency is often estimated using fast Fourier transform (FFT). However, FFT has a low frequency resolution at the condition of higher velocity of a passing vehicle; therefore, it is not appropriate to be used to monitor the frequency change caused by the degradation of the bridge structural integrity. This paper introduces a new frequency identification technique based on wavelet analysis. Wavelet transform is characterized by its high-frequency resolution and can, therefore, be used to visualize the bridge damage represented as changing the fundamental frequency of the bridge. The paper will implement this approach using an implicit Vehicle-Bridge Interaction (VBI) algorithm to simulate the passage of the inspection vehicle over the bridge. The acceleration signals are then processed using wavelet analysis to extract the bridge frequency. In addition, the study will investigate the use of a subtracted signal from two consecutive axles. The latter point has the advantage of substantially removing the effect of the road roughness from the recorded acceleration history.

Subspace (i.e. image, text or latent subspace) learning is one of the essential parts in cross-media retrieval. And most of the existing methods deal with mapping different modalities to the latent subspace pre-defined by category labels. However, the labels need a lot of manual annotation, and the label concerned subspace may not be exact enough to represent the semantic information. In this paper, we propose a novel unsupervised concept learning approach in text subspace for cross-media retrieval, which can map images and texts to a conceptual text subspace via the neural networks trained by self-learned concept labels, therefore the well-established text subspace is more reasonable and practicable than pre-defined latent subspace. Experiments demonstrate that our proposed method not only outperforms the state-of-the-art unsupervised methods but achieves better performance than several supervised methods on two benchmark datasets.

Deep learning-driven multi-level data fusion framework for predictive maintenance and structural health monitoring of concrete bridge decks

This study aims at developing a framework for assessing and tracking structural conditions by combining computer vision-based three-dimensional (3D) reconstruction and vibration measurement. Starting from a finite element (FE) model with unknown geometric and material properties, the framework seeks to obtain the accurate representation of the structure by applying a sequence of two types of operations: (1) image collection, 3D reconstruction, and the extraction of localized geometric properties, and (2) vibration measurement, operational modal analysis, and extraction of global stiffness properties. The first step, termed the computer vision step, is based on the alignment of the approximate initial finite element model to the 3D reconstruction, from which the area and second moment of area associated with each element of the model are retrieved by slicing the reconstructed mesh. The second step, termed the vibration measurement step, performs system identification using vibration measurement data, followed by the Bayesian model updating to calibrate the element stiffness. Every time one of those operations is performed, the new information is merged into all the information obtained previously. The framework is validated using a laboratory-scale bamboo cantilever beam. The experimental validation showed the potential of the proposed framework for effectively assessing and tracking localized structural conditions by heterogeneous sources of information.

To address the critical need for reliable safety assurance and efficient health assessment of in-service bridge structures, this study proposes a structural health evaluation method based on the Gaussian Mixture Model-Hidden Markov Model (GMM-HMM), aiming to tackle key challenges in automated monitoring and intelligent assessment of bridges. First, an automated monitoring system is established by deploying multi-type sensors (including displacement meters, strain gauges, inclinometers, and vibration sensors), enabling real-time data acquisition and forming a networked structural health monitoring (SHM) framework. Second, to overcome the issues of massive data volume and severe noise interference inherent in multi-sensor monitoring, a phased ‘anomaly detection-intelligent repair’ mechanism is developed for data preprocessing, ensuring the quality and reliability of input data. Subsequently, the proposed GMM-HMM is applied to practical bridge health assessment, using the Shangmadun Bridge in Jingmen (Hubei Province) as a case study. This application validates the method’s feasibility and effectiveness in quantifying structural health conditions, and a comparison with the GMM demonstrates the superiority of the GMM-HMM. Furthermore, to enhance model performance, the K-means algorithm is introduced to optimize the initial parameter selection of GMM-HMM. Compared with random initialization, this optimization reduces iterations by 48.4%, significantly mitigating the risk of local optima and improving convergence efficiency. The findings validate that the integrated approach provides a robust solution for automated, accurate, and efficient bridge health assessment, supporting data-driven decision-making in structural maintenance and safety management.

Deep learning and structural health monitoring: Temporal Fusion Transformers for anomaly detection in masonry towers

Zero-shot learning (ZSL) for visual recognition aims to accurately recognize the objects of unseen classes through mapping the visual feature to an embedding space spanned by class semantic information. However, the semantic gap across visual features and their underlying semantics is still a big obstacle in ZSL. Conventional ZSL methods construct that the mapping typically focus on the original visual features that are independent of the ZSL tasks, thus degrading the prediction performance. In this paper, we propose an effective method to uncover an appropriate latent representation of data for the purpose of zero-shot classification. Specifically, we formulate a novel framework to jointly learn the latent subspace and cross-modal embedding to link visual features with their semantic representations. The proposed framework combines feature learning and semantics prediction, such that the learned data representation is more discriminative to predict the semantic vectors, hence improving the overall classification performance. To learn a robust latent subspace, we explicitly avoid the information loss by ensuring the reconstruction ability of the obtained data representation. An efficient algorithm is designed to solve the proposed optimization problem. To fully exploit the intrinsic geometric structure of data, we develop a manifold regularization strategy to refine the learned semantic representations, leading to further improvements of the classification performance. To validate the effectiveness of the proposed approach, extensive experiments are conducted on three ZSL benchmarks and encouraging results are achieved compared with the state-of-the-art ZSL methods.