Journal of Civil Structural Health Monitoring最新文献

Limited research has been undertaken on the extant optimization models pertaining to the prediction of arch deformation in the course of erecting long-span concrete-filled steel tube (CFST) arch bridges. Moreover, CFST arch bridges stand as prototypical instances within the realm of bridge engineering’s segmental structures. This study focuses on the CFST arch bridge as a case study for deformation prediction. In pursuit of precise CFST arch bridge deformation prediction during construction, our investigation has formulated an arch-truss deformation prediction model and computational approach. Moreover, high-precision measuring robots (total stations) are usually utilized to monitor the deformation of arch bridges to obtain accurate deformation data for subsequent prediction studies. This model relies on the employment of the genetic algorithm (GA), sequential minimal optimization (SMO) algorithm, and an optimized support vector machine (SVM). Five input parameters, as utilized in the cable-stayed fastening-hanging cantilever assembly method, have been employed in constructing the model. The optimal parameter configuration for the SMO-SVM model was ascertained by employing an adaptive genetic algorithm to ensure an efficient and precise optimization process. Subsequently, the SMO-SVM model underwent training with the identified optimal parameter set. The prediction outcomes were subsequently verified through testing, facilitating the prediction of arch truss deformations during the construction of CFST arch bridges. To demonstrate the applicability of the proposed model, it was applied to the Pingnan Third Bridge, which is recognized as the world's longest-span CFST arch bridge (at the time of completion) and has a main span of 575 m. We also conducted a comparative study of the GA-SMO-SVM model in three distinct dimensions: various kernel functions, differing optimization algorithms, and alternative regression models. Our findings indicate that the GA-SMO-SVM model, which harnesses an adaptive genetic algorithm for efficient model optimization, achieves the most precise deformation predictions with a maximum absolute error of 8.86 mm, outperforming other models and achieving millimeter-level accuracy. Furthermore, the GA-SMO-SVM model operates with high efficiency, requiring approximately 1/64th of the time consumed by the SMO-SVM model optimized via a grid search (GS). This study validates the mechanism of the CFST arch bridge construction deformation prediction model founded on the GA-SMO-SVM algorithm through thorough model interpretation and analysis, presenting a pioneering approach to predicting arch deformations in the construction of extensive-span CFST arch bridges. Additionally, it offers a foundation for predicting deformations in other segmental structures.

This paper describes the experimental campaigns on the Tower of the Palazzo dei Vicari in Scarperia, a village in the Mugello area (Tuscany) exposed to high seismic hazards. The first campaign was carried out from December 2019 to January 2020, and the Tower underwent the so-called Mugello seismic sequence, which featured an M 4.5 earthquake. Other ambient vibration tests were repeated in June 2021 and September 2023 when another seismic sequence struck the area near Scarperia. These tests aimed to characterise the Tower’s dynamic behaviour under ambient and seismic excitations and check the response of the Tower over time. The experimental results were then used to calibrate a finite-element model of the Tower and estimate its seismic vulnerability. Several numerical simulations were conducted on the calibrated model using the NOSA-ITACA code for nonlinear structural analysis of masonry buildings. The dynamic behaviour of the Tower subjected to a seismic sequence recorded in 2023 by a seismic station at the base was investigated by comparing the velocities recorded along the Tower’s height with their numerical counterparts. Furthermore, several pushover analyses were conducted to investigate the collapse of the Tower as the load’s distribution and direction varied.

In recent years, urban road collapse incidents have been occurring with increasing frequency, particularly in populous cities. To mitigate road collapses, geophysical prospecting plays a crucial role in urban road inspections. Ground Penetrating Radar (GPR), a non-destructive technology, is extensively employed for detecting urban road damage, with manual interpretation of GPR images typically used to identify buried objects. Nonetheless, manual interpretation methods are not only inefficient but also subjective, as they rely on the interpreter's experience, thereby affecting the interpreting reliability. This study investigates the distribution and causes of road collapses and develops a deep learning-based intelligent recognition model using GPR detection images of urban roads in cities of the South China as original samples. The finding reveal that road collapses are concentrated in the months of July and August, mainly caused by pipe leakage and rainfall. Common anomalies in urban road GPR detection comprise seven types of target objects, including cavity, pipeline, etc., with standard GPR images acquired through outdoor field experiments. Utilizing GPR forward simulation and image augmentation methods to expand the sample size, as well as generating anchor box dimensions through clustering analysis, have all been proven to effectively improve the model's performance. The urban road GPR image intelligent recognition model, based on the YOLOv4 algorithm, achieves a detection accuracy of up to 85%, proving effective in GPR detection of urban roads in cities of North China. This research offers valuable insights for the future application of deep learning-based image recognition algorithms in urban road GPR detection.

Foundation scour is a critical phenomenon that may lead to the collapse of railway bridges. This issue is even more concerning in the current scenario where extreme weather events, such as floods, are becoming more severe and recurrent. Among different methodologies for assessing the structural integrity of railway bridges, vehicle-assisted monitoring has emerged as promising due to its low-cost and straightforward sensor installation compared to direct instrumentation of bridges. This paper provides a proof of concept of employing vehicle acceleration measurements from passing trains to detect the occurrence of bridge scour. To assess the effectiveness of accelerometer placement in data acquisition, vertical acceleration responses are collected from various positions throughout the vehicle and for different vehicles in the train, considering operational variabilities and measurement noise. A deep autoencoder model is used to process raw acceleration measurements collected during multiple train passages over a bridge affected by scour, where the scour damage is simulated as a local reduction in stiffness within a specific pier-foundation system. The difference between model-based and vehicle responses obtained from various observed events is the prediction error evaluated by the mean absolute error. The Kullback–Leibler divergence-based damage index is proposed to assess the number of vehicle-crossing events required to infer the damage. Finally, the approach’s accuracy is evaluated using Receiver Operating Characteristic curves. The results demonstrate that the applied methodology is highly effective in detecting both 5% and 10% levels of scour damage for sensors placed on the front and rear bogies of the first and last vehicles, without any prior data preprocessing.

In recent years, machine learning techniques have been available to predict and interpret the structural behavior of dams. Continuous monitoring of dam structure safety is vital in preventing possible damage. This study aims to predict the structural behavior by considering data collected for 13 years from instruments in the dam structure. Various machine learning methods are performed to account for the multi-non-linear relationships between dam displacement and the influential factors, thereby exploring the displacement laws of the dam. Three error metric indicators are employed for prediction, validation, and verification techniques to ensure the performance of models. Validation techniques include historical data validation, prediction validation, and the residual behavior over time. Predicting the structural behavior of the dam using the selected model requires data related to the input variables of the model. For this reason, the long short-term memory (LSTM) model, a robust algorithm for predicting time series variables, was used to predict the input variables. LSTM model provided acceptable predictions of changes in the input variables for these years. Additionally, the Boosted Regression Trees model, selected as the most accurate in the evaluation process, was employed to predict the structural behavior of the dam for periods not yet experienced by the dam, using these input variables. The predicted behavior of the dam demonstrated a strong ability to interpret the health of the dam structure and prevent possible damages. The effectiveness of the LSTM model was confirmed as a promising method in predicting time series input variables for ML models to predict dam displacements in the next years.

Protecting power infrastructure through visual surveillance can assure the safe operation of a power system, especially in unstructured environments where leaning utility poles are particularly inclined to cause large-area blackouts or even personal injury. Current methods place too much emphasis on detection and not enough on understanding leaning postures. However, due to the diversity and uncertainty of leaning utility poles, understanding them remains an urgent problem. Traditional posture estimation via three-dimensional (3D) point clouds is energy-intensive and costly, which limits its adoption in resource-constrained visual surveillance systems. In this study, we present a methodology to understand utility poles, and to estimate their leaning postures using a low-cost monocular camera. Edges and lines are extracted. Through their corresponding proximity and orientation, potential lines of utility poles are estimated. By analyzing relative geometric constraints between potential lines, utility poles are segmented and corresponding leaning angles are estimated, which is helpful to make risk-informed decisions to make leaning utility poles resilient. The approach requires neither prior training, nor calibration or adjustment of the camera’s internal parameters. It is robust against color and illumination associated with severe weather conditions. The percentage of correctly segmented pixels was compared to the ground truth, demonstrating that the method can successfully understand utility poles, meeting safety monitoring requirements for power infrastructure.

The prediction of suspender frequency and tension is difficult to solve due to the non-linear nature of suspender parameters. A method of predicting suspender frequency and tension using the generalized regression neural network (GRNN) model was proposed in this paper. It is necessary to select some suspender parameters as inputs into the model to solve the non-linear nature problem of the suspender parameters, such as length, mass unit per length, bending stiffness, fundamental frequency as well as tension, and to select the suspender frequency or tension as output. To consider the effect of different boundary constraints, analytical expressions of suspender parameters based on the singular perturbation method are derived and applied to train the models. Two different types of neural network models: back propagation neural network (BPNN) and radial basis function neural network (RBFNN), are also used to predict suspender frequency and tension to compare with the GRNN model. Datasets consist of measurements and literature samples are used to verify the models. Furthermore, R², MAE, and RMSE are used to compare the performance of the models. The results showed that the application of GRNN achieves higher accuracy in predicting suspender frequency and tension compared to BPNN and RBFNN.

Damage location vector (DLV) method is a model-based structural health monitoring approach that needs the frequency response–function response of the structure. A review of the literature indicates that although the DLV method accurately identifies the damage location in the single-damage structures, it does not work properly for the multi-damage. Accordingly, the aim of this research is to advance the DLV approach to increase its accuracy to detect the damage locations and severities in the multi-damage structures. In this regard, experimental and numerical studies were performed on the two-fixed ends steel beam having multiple damages with different intensities. During laboratory tests, the vibration response of steel beam specimens with multiple defects stimulated by hammer impact was measured. Different sensor locations were considered in the tests. A finite-element model of the steel beam was developed to calculate the dynamic response of undamaged beam under impact loading. Based on the fundamentals of hypothesis testing and data fusion, a threshold was derived to advance the DLV approach to detect the multiple damages. Moreover, the effect of sensor position on the performance of the DLV approach was investigated. The proposed method was also applied to a long-span box-shaped bridge to investigate its accuracy and efficiency for detecting damages in realistic complex structures. Moreover, the results obtained from the advanced DLV method were compared with other conventional methods, considering the effect of noise and different damage scenarios. The findings reveal that the advanced DLV approach proposed in this study accurately detects the defect locations and severities in the structures having multiple damages.

The present-day road traffic with the persistent change in the type and volume of vehicles needs to be specifically investigated for effective safety management of aging highway infrastructures. Actual traffic data can be implemented in refined procedures for stochastic simulation of road infrastructure performance, structural health monitoring (SHM), definition of weight limits on highways, and traffic-informed structural safety checks. While weigh-in-motion (WIM) systems had been widely used in many countries, their installation on Italian highways was mostly discussed and carried out only after the catastrophic collapse of the Polcevera bridge in 2018. This study presents a statistical data analysis, probabilistic models, and a simulation procedure for highway traffic, based on measurements of two WIM systems located along European route E45 close to Naples, Italy. Different limitations to maximum gross vehicle weight (GVW) were enforced at the locations of the two WIM systems, according to the Italian road code and the Italian guidelines for risk classification, safety assessment and monitoring of existing bridges, respectively. WIM data sets were filtered to exclude erroneous traffic data and vehicle classes defined according to the number of axles and axle distance were statistically characterised, allowing the derivation of probabilistic models for all traffic parameters of interest. A simulation methodology to generate random traffic load from the WIM data is also presented for its possible use in probabilistic performance assessment and traffic informed SHM of road infrastructures such as bridges.

Wire breakage in prestressed cylinder concrete pipes (PCCPs) due to various factors, such as corrosion, hydrogen embrittlement, material defects and overload, may lead to structural failure. Real-time detection of acoustic waves generated by wire breakage is possible using fiber optic sensors. Accurate determination of the first arrival time (FAT) of acoustic wave is vital for localizing wire breakages. A novel method based on the Bayesian optimal detector is proposed to automatically identify the FAT of near-wall acoustic wave. The FATs are subsequently fed into a localization model of wire breakage. The localization results are compared for the FAT of the proposed method and human subjective picking via model tests. The results show that compared with human subjective picking, the wire breakage localization of the proposed method can ensure the accuracy of the results. The maximum errors of the longitudinal and circumferential positions of the proposed method are 0.15 m and 0.02 m, respectively. The experimental results demonstrate that the FATs determined by the Bayesian optimal detector enable the accurate localization of wire breakage with noisy measurements. The proposed method overcomes the limitation of traditional picking methods in determining the FAT, which provides a promising tool for real-time monitoring of wire breakage in PCCPs.