https://doi.org/10.12652/Ksce.2026.46.1.0001

조성국(Cho, Sung Gook);소기환(So, Gihwan);이정희(Lee Jeonghee);이홍표(Lee, Hongpyo)

Various types of equipment are installed in industrial facilities, including nuclear power plants (NPPs). In these facilities, concrete expansion anchors (CEAs) are commonly used to secure equipment to concrete floors. However, anchor bolts may loosen over time during plant operation due to equipment vibration and other factors, resulting in installation torque levels lower than those initially applied. Because the anchorage condition affects the boundary conditions that influence the dynamic characteristics of the equipment, the performance of CEAs used for securing important equipment should be evaluated. CEAs should be tightened to the installation torque recommended by the manufacturer, but in practice, they are often installed with torque values that differ from the standard. Therefore, it is necessary to evaluate the performance of CEAs subjected to over-torque or under-torque conditions compared with the manufacturer’s recommendations. In this study, tensile tests were conducted on M12 CEAs installed with various torque levels under the same conditions. The anchors were installed in uncracked, low-strength concrete blocks fabricated in accordance with ASTM standards and tested following the procedures specified in ACI 355.2. The failure modes of the CEAs were examined, and the concrete breakout strength was evaluated through tensile testing. Based on the experimental results, the effects of installation torque on the tensile performance of CEAs were analyzed and discussed, and the results were compared with the anchor strength equations provided in the current design standard.

https://doi.org/10.12652/Ksce.2026.46.1.0011

이유재(Lee, Yujae);김충길(Kim, Chunggil);이재훈(Lee, Jaehoon);방건혁(Bang, Geonhyeok);허광희(Heo, Gwanghee)

Bridges are critical components of road and railway transportation networks; however, they are continuously exposed to repetitive traffic loads, environmental effects, and abnormal actions such as overloading and vehicle impacts, which can accelerate structural deterioration and lead to serious safety risks. Conventional bridge assessment methods, including static load testing, dynamic characteristic analysis, and visual inspection, suffer from high cost, traffic interruption, indirect evaluation, and subjectivity, limiting their applicability to real-time load identification. This study proposes a dual-optimization-based digital twin framework to efficiently identify the magnitude and location of applied loads using measured displacement responses of bridges. The proposed approach first performs a dynamic optimization using measured dynamic response data to update model parameters, thereby reducing structural modeling uncertainties and improving physical consistency. Subsequently, a static optimization is conducted using measured displacement data to accurately estimate the load magnitude and its application location. Through this two-stage optimization process, the structural accuracy of the numerical model is enhanced and the discrepancy between measured and simulated responses is minimized, leading to improved reliability of load identification. To validate the proposed method, displacement measurement experiments were conducted on a scaled bridge model under various loading conditions. The experimental results demonstrate that the proposed digital twin effectively reproduces the measured structural responses and enables stable and accurate estimation of both load magnitude and location. In particular, the integration of dynamic and static measurement data through dual optimization significantly reduces estimation variability and provides consistent load identification performance under diverse measurement conditions.

https://doi.org/10.12652/Ksce.2026.46.1.0021

이재훈(Lee, Jae Hoon);허광희(Heo, Gwang Hee);전승곤(Jeon, Seung Gon);방건혁(Bang, Geonhyeok)

Under non-stationary dynamic excitations whose characteristics vary with time, such as seismic loads, the variability of structural responses increases significantly, which limits the ability to determine the presence or absence of damage using only raw measured signals. In this study, to address this limitation, an IMD-based damage indicator enhancement method grounded in multivariate statistical analysis is proposed, and the effect of changes in the covariance-matrix dimension on damage discrimination performance is experimentally investigated. To this end, a shake-table excitation test was conducted on a scaled cable-stayed bridge model using input motions simulating the El-centro earthquake record, and vertical acceleration responses were measured under an undamaged condition as well as under single- and combined-cable damage conditions. Based on the measured signals, IMD was computed by progressively expanding the covariance-matrix dimension, and the damage discrimination characteristics associated with the dimensional variation were comparatively analyzed. The results show that, as the covariance dimension increased, the increase in IMD values within the damaged intervals became more pronounced, and the separability between the undamaged and damaged states improved progressively. This indicates that expanding the scope of multivariate information enables more effective representation of response characteristics induced by damage. In addition, considering that IMD is a point-wise metric that may be sensitive to transient disturbances, the IMD time series was summarized using the RMS value; consequently, consistently larger RMS values were observed at sensors located near the damage. These findings confirm that the proposed indicator is applicable not only to damage detection but also as an auxiliary quantitative measure for damage localization.

https://doi.org/10.12652/Ksce.2026.46.1.0029

방건혁(Bang, GeonHyeok);이재훈(Lee, JaeHoon);전승곤(Jeon, SeungGon);허광희(Heo, GwangHee)

In this study, management information was extracted from overweight vehicles crossing an in-service bridge with the aim of establishing a database for long-term overweight vehicle monitoring. The target vehicles were limited to 3-axle and 4-axle freight trucks whose load distribution characteristics are similar to those of the standard design truck load (KL-510) specified in the Korean design code. The management information was defined as fundamental vehicle-level data that can be calculated and stored for overweight vehicle management, and includes the number of axles, traveling speed, individual axle loads, gross vehicle weight, and axle spacing. To ensure stable extraction of the management information, an overweight vehicle detection criterion based on axle-number-dependent load distribution characteristics was established. This criterion was employed as a primary filtering step to identify vehicles with a high likelihood of being subject to management among continuously passing traffic. Through this process, responses induced by passenger cars and light- to medium-duty vehicles were effectively excluded, enabling the systematic extraction of management information for overweight vehicles traveling in real time on the in-service bridge. The proposed management information extraction method was validated under a specific bridge type and strain measurement environment, and its capability to reliably reflect the operational characteristics of overweight vehicles was confirmed. However, rather than presenting a universally applicable detection criterion for all vehicle types and bridge conditions, the results of this study should be interpreted as a case study that experimentally demonstrates the feasibility of structuring overweight vehicle management information from a database construction perspective under similar conditions. The management information obtained in this study is expected to serve as fundamental data for future applications such as bridge fatigue damage assessment, overweight and overspeed enforcement, and the development of an overweight vehicle management database.

https://doi.org/10.12652/Ksce.2026.46.1.0039

최범균(Choi, Beom Gyun);·박종건(Park, Jong Gun);·윤창호(Yun, Chang-ho);·허광희(Heo, Gwang Hee)

In this study, the properties of three-phase composite electrically heated cement mortars incorporating multi-walled carbon nanotubes(MWCNT), micro steel fibers (MSF), and carbon fibers (CF) were systematically investigated. Electrically heated cement mortars have attracted increasing attention as an alternative technology for snow melting and ice prevention during winter conditions, and this study focused on the combined effects of conductive materials with different length scales. In particular, the synergistic effects resulting from the interactions among MWCNT, MSF, and CF were analyzed with respect to the microstructural characteristics, heating performance, and electrical behavior of the conductive mortars. The experimental results demonstrated that the three-phase composite system incorporating MWCNT, MSF, and CF exhibited a pronounced synergistic effect compared to single-phase or binary composite specimens. As the applied voltage and the dosage of MWCNT increased, the heating performance was significantly enhanced, while the electrical resistivity showed a marked decreasing trend. These results indicate that the formation of continuous conductive pathways plays a critical role in improving the self-heating performance of cement mortars. Scanning electron microscopy (SEM) observations revealed that MWCNTs were relatively uniformly dispersed within the cement matrix together with MSF and CF, forming conductive networks and acting as effective bridging elements between conductive fillers. However, localized agglomeration of MWCNTs was also observed, which resulted in non-uniform conductive pathways and consequently contributed to a reduction in heating performance. Overall, the findings of this study clarify the relationship between microstructure and heating behavior in three-phase conductive cement mortars and provide fundamental insights for the design of electrically heated pavement materials for snow-melting and deicing applications.

https://doi.org/10.12652/Ksce.2026.46.1.0045

서동주(Seo, Dong Ju);박종건(Park, Jong Gun);최범균(Choi, Beom Gyun);허광희(Heo, Gwang Hee)

To solve pedestrian safety problems caused by icing on sidewalks in winter, this study conducted an experimental study by applying a self-heating monitoring system that combines electrically conductive concrete sidewalk blocks with an electrical heating function and an Internet of Things (IoT) sensor. The developed electrically conductive concrete sidewalk blocks were reviewed through flexure strength and water absorption rate tests in accordance with KS F 4419 standards, and all of them were found to satisfy the criteria. In addition, the system measured the internal and surface temperatures of electrically conductive concrete sidewalk blocks by utilizing IoT sensors, and monitored the ambient temperature and humidity inside the freezer in real time. The test results confirmed that the IoT sensor-based monitoring system significantly increased the interior and surface temperatures of the concrete sidewalk block in a low-temperature environment (-20oC), and that self-heating was stably detected and transmitted. In particular, the ‘CNT0.6+CF1.0’ concrete sidewalk block exhibited the best self-heating performance, with internal and surface temperatures measured at 97.2oC and 65.6oC, respectively, when a voltage of 30 V was applied. These results show that the electrically conductive concrete sidewalk block and IoT-based monitoring system proposed in this study can be used as an effective self-heating and sensing technology to prevent freezing of sidewalks in winter.

https://doi.org/10.12652/Ksce.2026.46.1.0053

김훈관(Kim, Hun Kwan);박종건(Park, Jong Gun);송기창(Song, Ki Chang);허광희(Heo, Gwang Hee)

In this study, multi-walled carbon nanotubes (MWCNTs) were deposited onto polyethylene terephthalate (PET) fibers using a dip-coating process in order to impart electrical conductivity and Joule heating capability, with the ultimate goal of enhancing their applicability to winter road de-icing systems. By employing a simple and scalable coating technique, conductive networks of MWCNTs were successfully formed on the surface of PET fibers without altering the inherent flexibility and lightweight nature of the polymer substrate. The influence of varying MWCNT mixing ratios on surface morphology, electrical resistance, and heating performance was systematically investigated to establish structure?property relationships. As the MWCNT loading ratio increased, a more continuous and interconnected conductive network was observed on the fiber surface, leading to a pronounced reduction in electrical resistance. Correspondingly, under the application of direct current (DC) voltage, the coated fibers exhibited improved heat generation efficiency and higher steady-state temperatures. Scanning electron microscopy (SEM) analysis clearly confirmed the formation and uniformity of the MWCNT coating layers on the PET fibers. These results demonstrate that MWCNT-coated PET fibers possess stable electrical and thermal performance and highlight their potential as effective conductive and self-heating materials for infrastructure applications exposed to low-temperature environments, such as winter road surfaces and cold-region facilities.

https://doi.org/10.12652/Ksce.2026.46.1.0061

최준우(Choi, Junwoo);김종범(Kim, Jong-Beom)

The Korea Hydrographic and Oceanographic Agency (KHOA) operates an observation based real time rip current warning system to reduce rip current related safety incidents. This study statistically evaluated the predictive performance of rip current risk indices produced by the system for nine beaches using four years of data collected from 2021 to 2024. The existing risk index was first converted into a probabilistic index representing occurrence probability, and classification performance and forecast skill were comprehensively assessed using the Brier Skill Score (BSS), Accuracy, Probability of Detection (POD), False Alarm Ratio (FAR), False Positive Rate(FPR), the Youden index, the ROC curve, and the Area Under the ROC Curve (AUC). The results showed that AUC values ranged from 0.92 to 0.99 at most beaches, indicating excellent classification performance, while BSS values were generally greater than 0.36, demonstrating improved skill relative to a climatological reference. In particular, Naksan, Gyeongpo, and Mangsang Beaches exhibited high BSS and AUC values, confirming robust predictive performance. In addition, when the warning threshold probability(critical probability) was appropriately selected (typically in the range of 0.35 to 0.40), Accuracy remained at 84 to 95 percent while the balance between POD and FPR improved, suggesting enhanced operational efficiency. The data driven verification procedure proposed in this study, applied over multiple years, can contribute to establishing a methodology for evaluating rip current forecasts and to improving the reliability of rip current warning and forecasting systems.

https://doi.org/10.12652/Ksce.2026.46.1.0077

노정두(Noh, Jeongdu);오석진(Oh, Seokjin);서혁(Seo, Hyeok)

This study was conducted to identify the characteristics of traffic accidents involving Personal Mobility (PM) devices, which have recently emerged as a major mode of transportation in urban areas, and to determine the primary causes affecting accident severity. To this end, a total of 469 PM accident cases that occurred in Gwangju Metropolitan City from 2021 to 2023 were collected, and a multifaceted analysis was performed using Pearson correlation analysis, Logistic Regression, and the XGBoost algorithm. The analysis results showed that both the number of accidents and accident severity, calculated using the Equivalent Property Damage only method, exhibited an increasing trend each year. Statistically, injury severity was found to have the strongest correlation with road type. Furthermore, predictive models were constructed and compared by applying various ratios of training and test data; the results indicated that the XGBoost model demonstrated superior predictive performance over the Logistic Regression model in most evaluation metrics, including AUC, precision, and recall. In the analysis of variable importance, accident type and traffic law violations were identified as critical factors in the Logistic Regression model, whereas the driver’s age was found to be the most influential factor affecting accident severity in the XGBoost model. The findings of this study quantitatively present the rick factors of PM accidents and can serve as effective fundamental data for establishing customized traffic safety policies by local governments and for promoting a safety culture among users in the future.

https://doi.org/10.12652/Ksce.2026.46.1.0087

송민혁(Song, Minhyuk);조윤범(Jho, Youn-beom);김종우(Kim, Jongwoo)

Conventional pavement marking maintenance has relied on localized measurements of retroreflectivity and fragmented data management structures organized by individual devices or operational units. As a result, comprehensive analysis of long-term and segment-level measurement data has been limited, leading to persistent challenges in achieving efficient and consistent maintenance decision-making. To address these limitations, this study proposes a pavement marking maintenance framework that integrates a visualization platform and conversational analysis enabled by the Model Context Protocol (MCP), based on the analysis of mobile LiDAR-derived retroreflectivity data to support proactive and preventive maintenance decisions at the roadway segment level. First, domestic and international guidelines and institutional trends related to the observation and management of pavement marking retroreflectivity were reviewed to identify limitations in existing maintenance practices. Subsequently, a data-driven platform was designed to compensate for the blind spots of conventional maintenance systems by enabling the systematic storage, analysis, and visualization of pavement marking retroreflectivity data. The proposed visualization platform integrates data storage, analytical processing, and visualization functions within a unified environment. In addition, through MCP, Large Language Models (LLMs) are designed to utilize pavement marking retroreflectivity databases, maintenance manuals, and relevant regulations as contextual information, thereby providing maintenance-related insights through conversational analysis. The proposed approach is expected to substantially enhance the practical utilization of pavement marking retroreflectivity data and to improve the operational efficiency and proactive response capability of road maintenance by advancing automation and decision-support functions.

https://doi.org/10.12652/Ksce.2026.46.1.0095

박준영(Park, Jun Young);김종우(Kim, Jong Woo);조윤범(Jho, Youn Beom);정영우(Jung, Young Woo)

Accurate axle event detection in High-Speed Weigh-In-Motion (HS-WIM) systems is a fundamental requirement for the reliable estimation of individual axle loads and the accurate identification of overweight vehicles. However, conventional threshold-based axle event detection approaches are inherently sensitive to sensor offset drift, environmental noise, and limited responsiveness to low-weight axle loads, which frequently results in false and missed detections. To address these challenges, this study presents an automated axle load segment detection method in which time-series axle load signals acquired from HS-WIM systems are converted into two-dimensional graph representations and analyzed using a YOLOv8-based object detection framework. By learning the characteristic shapes of axle load segments directly from signal representations, the proposed method enables robust axle event detection without reliance on predefined threshold values or absolute signal magnitudes. The proposed model was trained and evaluated using real-world measurement data collected from a public road testbed under actual traffic conditions. The validation results demonstrated high detection performance, with Precision, Recall, and mAP@0.5 values of approximately 0.999, 1.000, and 0.995, respectively. Furthermore, an axle load segment detection accuracy of approximately 95 % was achieved on an independent test dataset. These results indicate that the proposed YOLO-based axle event detection method provides a reliable and practical solution for axle load measurement and overweight vehicle identification in HS-WIM systems.

https://doi.org/10.12652/Ksce.2026.46.1.0105

이태범(Lee, Tae-Beom);염상국(Yum, Sang-Guk);박민수(Park, Minsoo)

The frequency and severity of wildfires have been increasing due to extreme heat and drought conditions driven by climate change. Rapid and accurate assessment of burned areas immediately after wildfire events is essential for prioritizing restoration efforts and establishing effective disaster response strategies. This study proposes a post-processing framework that combines SegFormer with a superpixel-based non-iterative clustering technique for precise burned area estimation from high-resolution UAV imagery. SNIC-based superpixel aggregation was applied to the pixel-wise predictions from SegFormer to reduce salt-and-pepper noise and enhance spatial consistency along boundaries. Experimental results demonstrate that the proposed method improved precision from 84.22 % to 86.83 % and IoU from 88.17 % to 92.90 % compared to SegFormer alone, confirming overall improvements in the stability of damage masks and reliability of area estimation. Therefore, the proposed framework demonstrates the potential for rapid and reliable burned-area estimation in small- and medium-scale wildfire-affected areas and can serve as a practical analytical tool to support disaster response and recovery decision-making.

https://doi.org/10.12652/Ksce.2026.46.1.0111

성지연(Sung, Ji Yeon);나종호(Na, Jong Ho);공준호(Gong, Jun Ho);정유석(Jung, Yoo Seok);신휴성(Shin, Hyu Soung);오윤석(Oh, Yoon Seuk)

Construction sites in Korea continue to exhibit a high incidence rate of industrial accidents, while many accident countermeasure documents have been written in a largely formalistic manner, resulting in limited practical effectiveness. In this study, an automated system for generating recurrence prevention measures was proposed using an LLM-RAG framework based on accident reports from the Construction Safety and Health Information System (CSI) and the KOSHA construction safety guidelines. The proposed system was designed to promptly derive accident-type-specific recurrence prevention measures and provide concrete actions that can be applied to real construction sites. A Korean-specialized embedding model was employed to semantically vectorize both accident cases and guideline documents, and a retrievable knowledge base was constructed accordingly. Upon user query input, semantically similar accident cases and guideline documents were retrieved and provided to the large language model as reference materials, enabling evidence-grounded response generation. In the experimental evaluation, the suitability of the generated outputs was compared and reviewed across multiple accident-type queries, and the generated measures were assessed in terms of alignment with the recommendations specified in the guidelines. Based on an LLM-as-a-judge evaluation, an accuracy of approximately 93 % was achieved, suggesting that the generated recurrence prevention measures were appropriate for the given accident types and highly applicable in practice. In particular, the document-grounded outputs tended to demonstrate improved specificity and consistency compared to conventional countermeasure documents. The proposed approach is expected to contribute to the development of practical response strategies within increasingly digitalized construction safety management systems.