4th International Conference on Structural Health Monitoring and Engineering Structures (SHM&ES 2025)

Table of Contents

1. Structural Health Monitoring

   1. Frontmatter

   2. Enhancing Vibration-Based Failure Identification in Beam Structures Using Statistical Features and Machine Learning

      Long Viet Ho, Ba Ho-Xuan, Toan Vu-Van

      Abstract

      Early diagnosis of structural damage, particularly in identifying its location, is essential for timely repair and maintenance. A vibration-based approach is effective, as damage alters a structure’s dynamic properties. Among these, mode shape-based methods offer faster, simpler localization than frequency-based ones. This study proposes a statistically based approach to enhance damage localization by applying a threshold to suppress false peaks in undamaged areas. Numerical studies on two beam-like structures confirm its superior accuracy compared to the modal curvature and mode shape curvature square methods. The method's robustness is validated under varying conditions, such as different mode numbers, sensor sparsity, and damage levels. To quantify damage extent, an artificial neural network (ANN) model optimized using a stochastic algorithm is employed. The optimized ANN achieves less than 2% error, even with added white Gaussian noise. The findings confirm the efficiency and reliability of the proposed approach in both localizing and quantifying structural damage.

   3. Forecasting the Ultimate Load Capacity of Flat Slabs with Artificial Neural Networks

      Hieu-Phuong Vu, Tien-Thuy Nguyen, Hoang-An Le

      Abstract

      Flat slabs are increasingly popular in modern construction due to their beamless design and ability to optimize space. They help reduce story height and maximize usable floor area. However, accurately predicting the ultimate load capacity of flat slabs is still challenging, influenced by factors such as geometry, materials, and load conditions. This study applies an Artificial Neural Network (ANN) model to predict the ultimate punching shear load of fiber-reinforced concrete slabs, based on 232 experimental data samples. The model consists of four hidden layers and is trained using advanced techniques to enhance generalization capability. Results show that the model achieves high prediction accuracy, with a coefficient of determination R² = 0.936 and a mean absolute percentage error (MAPE) of 11.88%. These findings demonstrate that ANN is an effective tool for predicting the punching shear capacity of flat slabs.

   4. Modal Strain Energy and Convolutional Neural Network-Based Damage Identification in Plate-Like Structures

      Ngoc-Tuan-Hung Bui, Thanh-Cao Le, Van-Sy Bach, Tran-Huu-Tin Luu, Manh-Hung Tran, Chi-Khai Nguyen, Duc-Duy Ho

      Abstract

      The modal strain energy (MSE)-based technique is a highly effective approach for damage identification. In this study, it is chosen among vibration-based techniques to presented a method for identifying damage in plate-like structures via a convolutional neural network. The finite element method (FEM) is utilized to analyze the free vibration of the plate to obtain the natural frequencies and mode shapes of six initial bending modes. These data serve as the primary input for the presented method. To validate the feasibility of the presented method, this study investigates a simply supported aluminum plate. The results reveal that the presented method successfully identifies the damage in the plate by utilizing the appropriate modal strain energy data and establishing a damage threshold.

   5. Detecting Multiple Damages in I-Section Steel Beams Using an Improved Mode Shape Curvature Change-Based Method

      Khanh-Hoang Vu, Duc-Duy Ho, Manh-Tung Dinh, Quang-Thien Ho, Trieu-Vy Nguyen, Hong-Huan Chiem, Van-Sy Bach, Manh-Hung Tran

      Abstract

      In this study, an improved mode shape curvature change-based technique is introduced for detecting the appearance and location of damage in I-section steel beams. The vibration of a steel beam is numerical analyzed in both undamaged and damaged cases using the finite element (FE) method, where damage is introduced by reducing the flexural rigidity of the corresponding beam elements. Next, the damage location is determined based on a defined damage threshold. This research is conducted in two steps of structural health monitoring (SHM): (i) evaluating the occurrence of damage using a vibration characteristics-based method; and (ii) proposing an appropriate damage threshold for detection and assessing its effectiveness through a set of damage detection indices. The findings indicate that the mode shape curvature change-based method has high precision in the damage detection in both the damage’s appearance and location.

   6. Factors Affecting the Structural Health of French Colonial Architecture in Vietnam

      Le Duy Thanh

      Abstract

      The French colonial period in Vietnam left behind a significant architectural heritage, particularly in major cities such as Hanoi and Ho Chi Minh City. These structures are not only aesthetically valuable but also hold deep historical significance, contributing to the formation of the current urban identity. However, the number of French colonial architecture (FCA) has been declining significantly. One of the main reasons for this deterioration is the lack of structural health monitoring methods in management, leading to degradation, damage, and even safety risks. This study focuses on French colonial architecture in Hanoi, analyzing factors affecting structural health and proposing appropriate monitoring methods to preserve and maintain the value of this architectural heritage.

   7. Compressed Sparse Regression for Anchored Design of Experiments and Sensor Placement in Structure Health Monitoring

      Yunpeng Zhu, Lianyuan Cheng, Liangliang Cheng

      Abstract

      This study investigates sensor placement for condition monitoring in complex systems, focusing on capturing dominant dynamic responses that indicate abnormal conditions. Traditional sensor placement methods often rely on costly distributed sensors and heuristic strategies, which are not efficient in capturing the most informative response characteristics. To address these challenges, a data-driven Design of Experiment (DoE) approach is proposed, leveraging system science principles to optimize sensor allocation systematically. The implementation of this framework is formulated as a sparse regression problem, enabling an efficient selection of sensor locations that maximize information gain while minimizing redundancy. To solve this problem, a newly developed Compressed Orthogonalized Least Squares (Comp-OLS) algorithm is introduced. In order to validate the proposed approach, a case study on the DoE of a Duffing system is conducted. Compared with the commonly used Pivoting QR Factorization (PQRF) method, the results demonstrate that the Comp-OLS-based framework significantly enhances sensor placement efficiency, ensuring comprehensive coverage of system dynamics while anchoring the locations of required sensors. This study demonstrates the potential of data-driven DoE for improving condition monitoring in various engineering applications, offering a scalable and effective solution for sensor placement challenges.

   8. Predicting Building Energy Consumption Considering Climate Change Using 6D BIM and Machine Learning

      Tran-Hieu Nguyen, Do Thi Mai Dung

      Abstract

      This paper proposes a framework that integrates 6D Building Information Modeling (6D BIM) with machine learning techniques to predict building energy consumption under different climate scenarios. Firstly, a machine learning model is trained using the dataset generated from the 6D BIM-based parametric study. Next, a regression model is constructed based on the simulation results with the weather data according to climate change scenarios RCP 2.6, RCP 4.5 and RCP 8.5. Combining two models allows for the forecasting of future building energy demand up to the year 2100. A case study of 2-storey private house in Hanoi is presented to illustrate the proposed framework. This research underscores the potential of advanced digital tools and data-driven methods to support building design and operation in an era of environmental uncertainty.

   9. Optimization of Sensor Locations for Homogeneous Beams in Structural Health Monitoring Using Isogeometric Analysis and Differential Evolution

      Quan M. Lieu, Khanh D. Dang, Tam T. N. Do, Tan T. Nguyen, Anh H. Nguyen, Van Hai Luong, Qui X. Lieu

      Abstract

      This work introduces a numerical approach to the optimization of sensor placement for homogeneous beams in Structural Health Monitoring (SHM). In which, the displacements through the beam height are represented by a generalized shear deformation theory (GSDT) based on the third-order polynomial function. Meanwhile, the displacements through the beam length are approximated by B-spline functions within the Isogeometric analysis. Accordingly, the position of measurement sensors concerning degrees of freedom (DOFs) defined at control points is determined by maximizing the sum of the terms in the Modal Assurance Criterion (MAC) which is built by eigenvectors obtained by the full model and a model order reduction (MOR) utilizing the second-order Neumann series expansion (SNSE). Differential Evolution (DE) is utilized as an optimizer. A simply supported beam is investigated to illustrate the current methodology’s reliability. Obtained results have indicated that the current paradigm can be utilized for the sensor location optimization of other structures with potential applications to the SHM.

   10. An Adaptive DNN-Assisted Metamodel for Damage Detection of Steel Frames Based on Incomplete Frequencies and Mode Shapes with Limited Training Datasets

       Vin Nguyen Thai, Du Dinh Cong, Duy Khuong Ly, Thao Nguyen Trang, Trung Nguyen-Thoi

       Abstract

       This study presents an adaptive metamodel approach, assisted by a Deep Neural Network (DNN), for damage detection in steel frames based on incomplete frequency and mode shape data with limited training datasets. The proposed method integrates model order reduction (MOR) and a multi-stage process to enhance efficiency and accuracy. Initially, the Modal Strain Energy Change Ratio (MSECR), calculated from incomplete modal data, is employed to eliminate low-risk damage candidates by leveraging a second-order Neumann series expansion-based MOR (NSEMR-II) technique. This significantly reduces the neural network architecture of the DNN model used in subsequent stages. The DNN is trained on frequencies and mode shapes simulated using the Finite Element Method (FEM), corresponding to measured degrees of freedom (DOFs). Iteratively refining damage candidates through a damage threshold, the method improves diagnostic accuracy while maintaining low computational demands and requiring only moderately sized datasets. The simplified DNN models effectively identify both the location and severity of damage using data from limited sensors, even under high noise conditions. Numerical examples on steel frame structures validate the approach’s efficiency and practicality for structural health monitoring applications.

   11. TPE-Optimized Neural Network Framework for Predicting Settlement of Nodular Pile Foundations

       Hung La, Tan Nguyen, Khiem Quang Tran

       Abstract

       This study presents a data-driven approach to predicting settlement of statically loaded pile foundations. We employ an Artificial Neural Network (ANN) model whose hyperparameters are fine-tuned using the Tree-Structured Parzen Estimation (TPE) method. The training procedure uses data obtained from physical tests, incorporating key factors such as nodular pile size, vertical loading conditions, and soil resistance characteristics. The results demonstrate that the optimized ANN model provides robust predictive performance. This approach offers valuable potential to improve the reliability of geotechnical design practices.

   12. Damage Detection of Trusses Utilizing Free Vibration Signals and Convolutional Neural Network Relied on Model Order Reduction

       Tan T. Nguyen, Quan M. Lieu, Trong V. Trinh, Tam T. N. Do, Qui X. Lieu, Khanh D. Dang

       Abstract

       This study presents an approach for damage quantification of trusses utilizing free vibration signals and Convolutional neural network (CNN) relied on model order reduction (MOR). The input data consists of the values of eigenvectors extracted from several important degrees of freedom (DOFs) instead of all ones, collected from numerical simulations under various random damage scenarios. The output is the truss members’ randomly assumed damage ratios. The Modal strain energy-relied index (MSEI) is applied to eliminate members with a low probability of damage, aiming to reduce the data dimension for CNN. Thereby, its accuracy of predicting the damage detection is improved with the capability of automatically extracting features from CNN, this method significantly reduces the computational cost in training and testing compared to traditional methods. The methodology is validated on a 2D truss model under two damage scenarios programmed in Python. The results are promising for providing the method's potential applications to structural health monitoring (SHM).

   13. Monitoring Column and Shear Wall Shortening in High-Rise Buildings

       Khiem Van Giang, Hien Manh Nghiem

       Abstract

       This study examines the vertical shortening behavior of reinforced concrete columns and shear walls in a 55-story high-rise building. Field measurements were carried out during construction using embedded sensors to monitor time-dependent deformations caused by creep, shrinkage, and elastic shortening. Shortening data were collected at multiple levels, specifically the 16th, 39th, and 49th floors, using embedded sensors installed in both columns and shear walls. Results show that vertical shortening is more pronounced at lower levels due to accumulated loads and sustained deformation over time. Columns exhibited slightly greater shortening than shear walls at corresponding locations, highlighting the influence of axial flexibility differences between structural elements. Symmetrical sensor pairs demonstrate consistent behavior, validating the structural design's uniformity. One notable exception was an abnormally large shortening at a shear wall location, suggesting localized effects that warrant further investigation. The findings emphasize the importance of differential shortening assessment to minimize long-term deformation-related issues, such as slab distortion and joint misalignment, in tall building construction.

   14. Suggesting a Procedure of Technical Diagnosing a Thin-Walled Horizontally Curved Steel Bridge

       Tham Hong Duong

       Abstract

       This article suggests a procedure for a technical diagnosis of a horizontally curved bridge (HCB). The structure is so significant in torsional warping and flexural actions that its response to the so-called ‘unusual component of internal forces’ bimoment, torsional warping, and flexural torsional effects is complicated to master. To understand the responses that need to be measured, large-span structures having different configurations of the cross-section, static, or dynamic loadings are intentionally combined. Firstly, a straight structure at the same span length and cross-section is studied as a comparative model; secondly, three finite element models of HCB with multi-cell and thin-walled cross-sections subjected to torsional and flexural are developed to get data of responses. By conducting a modal analysis and a simulated test with rather strong excitation at midspan, results indicate that it is relevant to install accelerometers at somewhere midspan and velocity transducers at supports; besides, flexural and torsional vibrations dominate simultaneously in the higher modes of curved structures. Some other findings are that the sharper the curvature is, the more notable the torsional vibration that appears in the first mode. Finally, the dynamic analysis for exploring the responses of the structure in the hope of finding a base for technical diagnosis of the structure using accelerometers, and strain gauges together with other remote instruments for displacement measuring tests, and non-destructive tests for the connection and status of assembly components. From these abovementioned results, some relevant locations, instrumentation design, and main technical specifications are suggested.

   15. Damage Classification of Steel Frames Using Long Short-Term Memory and Fully Convolutional Network Models

       Truong Thanh Chung, Tran Tien Son, Le Van Vu, Luong Nguyen-Duc, Tran Ngoc Hoa

       Abstract

       In the field of Structural Health Monitoring (SHM), the application of deep learning models for analyzing time-series data has garnered significant attention. One-dimensional convolutional neural networks (1DCNN) are commonly used but face limitations in effectively handling long datasets. Therefore, this study proposes a novel approach by combining 1DCNN with the Squeeze-and-Excitation (SE) mechanism (SE-1DCNN) and Long Short-Term Memory (LSTM) networks to accurately classify structural damage. This combination leverages the spatial feature extraction and attention mechanism of SE-1DCNN alongside LSTM’s capability to process long-term time-series data. The model is trained and evaluated using an experimental dataset collected from a steel frame structure instrumented with multiple accelerometers under various damage scenarios. The proposed SE-1DCNN-LSTM model achieves an accuracy of 96.7% on the training set and 95.3% on the test set, outperforming the traditional 1DCNN-LSTM model. These results confirm that integrating SE-1DCNN and LSTM enhances damage classification accuracy and demonstrates strong potential for real-world SHM applications.

   16. Reliable and Interpretable AI for CFST Column Safety Assessment

       Tran-Trung Nguyen, Thanh Cuong-Le

       Abstract

       This work suggests a hybrid framework to predict the dependability of concrete-filled steel tube (CFST) columns under axial stress by combining Monte Carlo Simulation (MCS), the CatBoost gradient boosting technique, and SHAP explainability. The model was trained using a dataset of 663 experimental CFST samples; the regression target was the computed failure probability P_f using Monte Carlo Simulation (MCS). Based on the dependability metric β, the CatBoost model effectively classified all samples into safety categories and achieved high predicted accuracy. According to SHAP analysis, geometric parameters—especially outer diameter, wall thickness, and column length—had the highest impact on expected failure probability. The dataset often revealed that a significant fraction fell below the accepted safety threshold \(\beta = 3.0\), which emphasizes the importance of design review in many respects. Moreover, a decision tree classifier was constructed to extract rule-based safety reasoning, providing a precise tool for informed engineering decisions. The proposed framework offers an accurate, interpretable, and computationally efficient alternative to conventional dependability evaluation techniques, leveraging transfer learning and semi-empirical modeling. It lays a strong basis for future applications to eccentric loading situations.

   17. Detectability Analysis of Structural Defects Using Lamb Waves: A Frequency-Based Approach for Structural Health Monitoring

       Juan Brazalez, Airton Nabarrete

       Abstract

       Structural Health Monitoring (SHM) plays a crucial role in ensuring the integrity and longevity of aerospace structures. This study investigates the detectability of structural defects using Lamb waves, focusing on the impact of defect size on wave propagation characteristics. Numerical simulations were conducted on an aluminum plate embedded with a sensor-actuator network, evaluating the interaction of fundamental Lamb wave modes (A0 and S0) with defects of 2, 4, and 8 mm in diameter. Frequency spectrum analysis revealed that larger defects lead to significant energy attenuation, spectral shifts, and mode conversion, particularly influencing the dispersive nature of the A0 mode. Detectability maps derived from FFT energy loss highlight the sensitivity of different sensor locations to damage, demonstrating that defect size and wave scattering influence signal degradation. The findings confirm that Lamb wave-based SHM effectively enables early defect detection and damage quantification. The results support the optimization of sensor placement and excitation frequency selection to enhance defect characterization.

   18. RTK and PPK Method in Automatic Bridge Monitoring

       Viet Ha Nguyen, Ngoc Quang Vu

       Abstract

       This paper investigates and evaluates the effectiveness of an automatic monitoring solution based on the GNSS-RTK method, in comparison with GNSS-PPK, for structural monitoring applications. The study employs two Comnav GNSS-N3 receivers, multi-frequency, multi-channel devices, operating autonomously on the professional CDMonitor platform. Results show that, at short distances, the accuracy of the RTK solution is comparable to that of the PPK approach. These findings serve as a foundation for selecting a suitable monitoring method for structures where the distance between the base station and the antenna mounted on the structure is a critical factor.

   19. Multi-damage Identification in Three-Dimensional Frame Structures via a Combined MSE-Based Method and PSO Algorithm

       Van-Sy Bach, Duc-Duy Ho, Thanh-Cao Le, Khanh-Hoang Vu, Manh-Tung Dinh, Manh-Hung Tran, Tran-Huu-Tin Luu

       Abstract

       This paper presents a multi-damage identification approach for three-dimensional frame structures that combines the modal strain energy (MSE)-based method with the particle swarm optimization (PSO) algorithm. Firstly, the potential damage locations are identified by the modal strain energy-based index (MSEBI). This index is calculated from the difference in the MSE values of the elements corresponding to the two states prior to and after the damage’s occurrence. In order to improve accuracy and reduce the limitations of the noise elements in determining damage location, the MSEBI index is determined from the first six vibration mode shapes. Secondly, the PSO with a function of objective variables based on the vibration modal strain energy (MSE) values determines the damage level of the elements identified in the first step. The accuracy and reliability of the proposed method are evaluated by analyzing a four-story three-dimensional frame structure with 63 elements, considering three different damage scenarios. The obtained results confirm that the proposed method accurately identifies the location and severity of multi-damage in the three-dimensional frame structures.

   20. Quantitative Assessment of Damage in Cementitious Beams via Acoustic Emission Technique (AET)

       Tam Nguyen-Tat

       Abstract

       The purpose of this study is to contribute to a deeper understanding of degradation mechanisms in concrete, mortar, and cement-paste beams subjected to mechanical loading, through the application of the Acoustic Emission Technique (AET). To this end, displacement-controlled three-point bending tests were conducted on three notched beams of identical shape and dimensions. The objective was to establish correlations between damage modes observed during each loading cycle and the corresponding AE activity. For data analysis, a background noise filtering technique was first applied to the raw AE signals collected during testing. The filtered signals were then subjected to a clustering process using the k-means algorithm to categorize them into distinct groups based on their characteristics. Following this, damage classification was performed on the filtered data using the RA method, providing insight into the nature and evolution of damage within the beams.