https://doi.org/10.11112/jksmi.2025.29.6.1

하수경(Soo-Kyoung Ha) ; 최기선(Ki-Sun Choi)

This study conducted a pull-off test to evaluate the bond performance of a quasi non-combustible external thermal insulation composite system (ETICS) cast onto concrete substrates under repeated freeze?thaw conditions. The specimens were manufactured as sandwich panels composed of quasi non-combustible EPS cores with calcium carbonate-based RCB exterior finishing materials on both sides and were exposed to up to 500 freeze? thaw cycles. The results showed that the average bond strength under all cycle conditions satisfied the ETAG 004 minimum requirement of 0.08 MPa, maintaining approximately 22.5% surplus strength even after 500 cycles. Cohesive failure within the insulation core was observed in most specimens, with adhesive failure between dissimilar materials appearing in only one case. These findings suggest that the cast-in-place method ensures reliable bonding at the insulation-to-concrete interface. However, potential quality variations within the composite panel highlight the need for precision in the manufacturing process. Overall, the cast-in-place attachment method is considered an effective approach for securing structural durability and reliability of ETICS affixed to external walls under prolonged freeze?thaw exposure.

https://doi.org/10.11112/jksmi.2025.29.6.11

한국인(Guk-In Han) ; 김동환(Dong-Hwan Kim) ; 조민수(Min-Su Jo) ; 임수아(Su-A Lim) ; 김길희(Kil-Hee Kim)

Brittle punching-shear failure remains the governing limit state in flat-plate slab?column connections. This experimental study investigates the effect of shear-reinforcement configuration on punching resistance, ductility, and failure behavior. Four half-scale specimens were tested under concentric column loading: one unreinforced reference slab and three specimens reinforced with shear bars arranged in orthogonal, radial, and diagonal-hoop configurations (reinforcement ratio ρsr ? 0.5?1.0%). Structural responses such as load?displacement behavior, reinforcement strain, and crack patterns were evaluated. While all configurations produced only a modest increase in initial punching strength (3?8% above the reference), they significantly enhanced post-peak ductility, with ultimate rotation capacities increasing by 2.1?2.4 times. The failure mode transitioned from brittle cone breakout to a ductile crushing governed by tensile-membrane action. Comparisons with predictions from KDS 14-2025, ACI 318-19, and Eurocode 2 revealed that existing design formulas overestimated strength by 15?28%, primarily due to overreliance on flexural reinforcement ratio and neglect of configuration-induced membrane effects. The findings highlight the need for refined design provisions that explicitly incorporate the influence of reinforcement configuration on punching performance.

https://doi.org/10.11112/jksmi.2025.29.6.21

박지원(Jiwon Park) ; 장민우(Minwoo Chang)

This study applies finite element (FE) model updating to estimate the material properties of aging bridges and evaluates the performance of optimization algorithms. The Deokpyeong 1 Bridge, for which initial design documents are unavailable, was analyzed using vertical acceleration data to identify its dynamic parameters and update the FE model. The acceleration data measured from 15 sensor locations are divided into every ten minutes, which results in 1152 datasets. The modal parameters were identified for each dataset. Two optimization methods-Genetic Algorithm (GA) and Particle Swarm Optimization (PSO)-were implemented to minimize the error between the modal parameters identified from the measured dataset and those obtained from the FE model. The GA demonstrated approximately 7 seconds faster computation time, whereas the PSO achieved lower frequency errors and MAC values exceeding 0.8, indicating superior estimation accuracy overall. Furthermore, the effect of temperature variation on bridge material properties was investigated. Results showed that both the elastic modulus of the main structural components and the spring stiffness of bridge bearings fluctuated with temperature variations. These findings underscore the importance of FE model updating for structural health monitoring and maintenance of aging infrastructure.

https://doi.org/10.11112/jksmi.2025.29.6.30

김일순(Il Sun Kim) ; 최소영(So Yeong Choi) ; 양은익(Eun Ik Yang)

This study examined the feasibility of deep learning?based concrete damage classification under small-scale data conditions. Four representative CNN models?GoogLeNet, ResNet-50, EfficientNet-B0, and MobileNetV2?were employed using a dataset of 3,000 images representing three types of damage: crack, efflorescence, and rebar exposure. The number of training images was varied at 100, 250, 500, and 1,000, and performance was evaluated in terms of accuracy, F1-score, training time, and t-SNE visualization. The experimental results showed that all models exhibited a performance saturation point around 500 images. ResNet-50 and EfficientNet-B0 achieved high accuracy (around 92%) and distinct cluster separability, while MobileNetV2 demonstrated real-time applicability owing to its lightweight structure and fast computation. Among the damage types, rebar exposure achieved the highest classification accuracy, whereas efflorescence showed relatively lower accuracy with greater variability. Overall, this study confirms that reliable classification performance can be achieved with more than 500 images, and provides practical criteria for selecting CNN models for field applications under limited data conditions. Future research should focus on expanding data diversity, validating with real-world images, and applying advanced preprocessing and augmentation techniques.

https://doi.org/10.11112/jksmi.2025.29.6.39

이성진(Sung-Jin Lee) ; 이창길(Changgil Lee) ; 김현기(Hyun-Ki Kim) ; 유용래(Yong-Rae Yu) ; 안호준(Hojune Ann)

When accidents occur in large railway infrastructures such as retaining walls, it is crucial to promptly record and digitally archive initial site conditions. However, in emergencies, deploying high-cost equipment like drones or LiDAR is difficult, and sites are often cleared quickly, leading to loss of key evidence. This study explores a rapid digital archiving method using a general-purpose smartphone. Video captured with a smartphone was processed into 3D point clouds via photogrammetry and used not only for site documentation but also for qualitative deformation identification and quantitative dimension estimation through scale correction. Experiments with tilt and bulging specimens showed that even with sparse data and without absolute scale, deformation occurrence and patterns were visually confirmed, and dimensional changes were measurable after scale correction. Error analysis indicated that tilt specimens had the largest deviation in height (up to 7.4 cm; MAE 5.1 cm length, 5.8 cm height, 1.0 cm depth, angle:0.75°), while bulging specimens maintained 2?3 cm errors in length and depth with minor variation in height (MAE 3.5 cm). These findings demonstrate that despite limitations, smartphone-based rapid imaging can detect deformation and capture approximate dimensions, offering a practical complementary tool for digital archiving and prompt accident site analysis. Future research should focus on improving data quality through dedicated smartphone applications and conducting comparative validation against high-precision measurement equipment.

https://doi.org/10.11112/jksmi.2025.29.6.49

김승우(Seungwoo Kim) ; 최웅규(Woonggyu Choi) ; 나상일(Sangil Na) ; 박승희(Seunghee Park)

The acceleration of climate change has increased the frequency of localized heavy rainfall and torrential rain, raising the risk of flooding in urban areas. Existing flood monitoring methods based on manual surveys and sensors have limitations such as a restricted application range, high operation, maintenance, and management costs, and a lack of real-time capability. This study introduces a computer vision?based method capable of detecting urban flooded areas in real time and quantitatively estimating flooded areas. A YOLOv8-based deep learning model is used to simultaneously detect flooded areas, traffic signs, traffic lights, and the actual size of the reference objects combined with pixel information in the image is utilized to calculate the flooded area in square meters. The training dataset was constructed by combining data collected from AI-Hub, and the final model achieved a mAP of 0.548 and an f1-score of 0.59. In addition, a field experiment conducted in areas frequently affected by flood showed an error rate of approximately 5.63% compared to the area estimation based on satellite imagery, demonstrating high reliability.

https://doi.org/10.11112/jksmi.2025.29.6.58

강진석(Jin Seok Kang) ; 김재현(Jae Hyun Kim) ; 정호성(Hoseong Jeong) ; 양은빈(Eun Bin Yang) ; 김강수(Kang Su Kim)

Reinforced concrete (RC) beam-column joints play a critical role in ensuring the structural safety and performance of buildings. However, failures in these regions often occur in a brittle manner, potentially leading to sudden and catastrophic collapse. Existing design code equations for joint shear strength are typically simple and conservative, but they often diverge from experimental results due to their limited consideration of influential parameters such as transverse reinforcement and longitudinal beam reinforcement. To address this gap, recent studies have increasingly applied machine learning techniques to model the complex interactions among these variables. In this study, a machine learning-based approach was adopted to develop an estimation equation for the shear strength of RC beam-column joints. Preliminary variables were selected based on structural design codes and prior research, and their relative importance was assessed using the random forest method. Key variables identified through this process were then used in genetic programming to derive the final estimation equation. The performance of the proposed equation was evaluated by comparing its accuracy with that of existing design code equations and previously developed models. The results demonstrated that the proposed shear strength equation for RC beam-column joints showed superior accuracy, indicating its reliability and potential for practical application in structural design.

https://doi.org/10.11112/jksmi.2025.29.6.70

심승보(Seungbo Shim)

Cracks in aging concrete structures pose significant threats to structural stability and public safety. However, it is difficult to secure sufficient crack image datasets that reflect diverse deterioration conditions in real-world scenarios, which leads to performance degradation of crack detection neural networks when applied on-site. To address this data scarcity problem, this study employs a generative AI approach based on Stable Diffusion. Specifically, an inpainting fine-tuning method using DreamBooth was applied to synthesize realistic cracks within masked regions, and additional training datasets were generated through combinations of different label images. For evaluation, 4,867 concrete crack images were divided into training, validation, and test sets with ratios of 15%, 15%, and 70%, respectively, and four segmentation models?DDRNet, RegSeg, SwinFormer(T), and PoolFormer(S24)?were assessed. The quality of the generated images was evaluated using the Frechet Inception Distance metric, confirming visually consistent and high-quality crack synthesis. In terms of detection performance, incorporating synthetic data improved the average F1-score by 5.73% compared with the baseline, with PoolFormer(S24) achieving the highest improvement of 11.96%. In conclusion, the proposed synthetic data-based augmentation method effectively enhances crack detection accuracy across diverse network architectures and environments, providing a practical and generalizable alternative for real-world applications.

https://doi.org/10.11112/jksmi.2025.29.6.80

노병철(Byeong-Cheol Lho) ; 박종현(Jong-Hyeon Park)

This study investigates the strength development mechanism of Portland Limestone Cement (PLC) as a sustainable alternative to Ordinary Portland Cement (OPC). PLC replaces a portion of the clinker with finely ground limestone, reducing CO₂ emissions while maintaining comparable mechanical performance. Experimental and analytical findings show that the compressive strength of PLC is governed by the combined effects of physical and chemical mechanisms, including the filler effect, nucleation effect, and carboaluminate formation. The optimal limestone replacement ratio of 10? 15% and increased fineness effectively offset the dilution of clinker, while the incorporation of supplementary cementitious materials (SCMs) such as fly ash and slag further improves long-term strength and durability through synergistic reactions. Performance-based mix optimization considering limestone fineness, clinker composition, water-to-binder ratio, and curing conditions is essential to achieve equivalent or superior performance to OPC. The results provide a scientific framework for designing low-carbon, high-performance concrete and support the implementation of Korea’s 2050 Carbon Neutrality Roadmap by providing a scientific framework for mix design and standardization in the construction materials sector.

https://doi.org/10.11112/jksmi.2025.29.6.92

문병갑(Byung-Kab Moon) ; 김경주(Kyoung-Ju Kim) ; 조두용(Doo-Yong Cho)

Truck Mounted Attenuators (TMAs) are employed on highways during maintenance or emergency repair operations, as well as for the inspection and management of bridge structures, in order to simultaneously protect road workers and vehicle occupants in the event of a collision. In the United States and Europe, the performance of TMAs is evaluated through full-scale crash tests; however, domestic standards remain limited compared to international criteria, particularly in terms of the weight ranges of support truck and test vehicles. This study analyzed occupant safety indices?namely the Theoretical Head Impact Velocity (THIV), Post-Impact Head Deceleration (PHD), and Acceleration Severity Index (ASI)?along with roll-ahead distance, based on crash tests conducted with combinations of test vehicle weights (1.3 ton and 2.0 ton) and support truck weights (5 ton and 10 ton). The results showed that an increase in test vehicle weight led to a decrease in THIV, while PHD exceeded the threshold criteria. Furthermore, a reduction in support truck weight caused the roll-ahead distance to increase by up to eleven times, thereby heightening the risk of secondary accidents for workers positioned ahead of the support truck. These findings are expected to serve as fundamental data for the improvement of future TMA performance evaluation standards.

https://doi.org/10.11112/jksmi.2025.29.6.102

강성원(Seong-Won Kang) ; 김동환(Dong-Hwan Kim) ; 조민수(Min-Su Jo) ; 김형국(Hyeong-Gook Kim) ; 김길희(Kil-Hee Kim)

Bond failure of longitudinal beam reinforcement passing through interior beam-column joints in reinforced concrete (RC) structures critically affects structural safety under cyclic loading. This study investigates the effect of concrete compressive strength on bond strength and evaluated the bond behavior of such reinforcement through experiments on eight RC beam?column joint specimens. To accurately simulate the stress state within the interior joint, tensile force was applied to the beam reinforcement at one end, while compressive force was applied to both the reinforcement and the beam cross-section at the opposite end. The experimental program examined the bond stress?slip relationship and the strain distribution along the beam reinforcement. All specimens were detailed to satisfy the column width requirements specified in KDS 14 20 80; however, bond failure was observed in several specimens, indicating that the current code provisions may not be sufficient to prevent bond deterioration under cyclic loading. The bond failure initiated on the tension side and gradually progressed toward the compression side as the load increased. The use of high-strength concrete enhanced resistance to ring tension, delayed the deterioration of bond stress on the compression side, and consequently improved the overall bond performance of the joint.

https://doi.org/10.11112/jksmi.2025.29.6.110

김일순(Il Sun Kim) ; 최소영(So Yeong Choi) ; 양은익(Eun Ik Yang)

This study quantitatively evaluated the interpretability of deep learning?based concrete damage classification models using Grad-CAM and compared the results with performance metrics to establish fundamental criteria for practical applications. Three representative CNN models? GoogLeNet, ResNet-50, and EfficientNet-B0?were tested with varying dataset sizes (750, 1500, 3000 images) and Grad-CAM threshold values (0.3, 0.5, 0.7). Model performance was assessed using accuracy and F1-score, while interpretability was evaluated with the Grad-CAM?based Damage Ratio. The experimental results showed that both performance and interpretability improved as the dataset size increased; however, a trade-off between the two metrics was observed. EfficientNet-B0 achieved the highest accuracy, whereas GoogLeNet produced wider activation regions with a higher Damage Ratio. In addition, threshold 0.5 yielded the most balanced results in terms of interpretability and noise suppression. In conclusion, this study highlights the importance of balancing performance and interpretability in deep learning?based structural damage diagnosis and proposes baseline criteria for model and threshold selection. Future research should focus on enhancing interpretability by incorporating diverse damage types and real-world structural data.

https://doi.org/10.11112/jksmi.2025.29.6.119

김봉균(Bong-Kyun Kim) ; 이병재(Byung-Jae Lee)

In order to enable broader and more direct applications of calcium silicate based cement (CSC), this study investigated the production of mortars based on rapid-hardening cement binders incorporating CSC. The effects of mixture proportion and incorporation of anhydrite gypsum on the fluidity, setting characteristics, and compressive strength of the mortars were evaluated. The results indicate that the fluidity and strength development characteristics of the rapid-hardening binder can be controlled by adjusting both the mixture ratio and the amount of anhydrite gypsum added. Furthermore, the fluidity of cement mortars containing CSC remained generally consistent regardless of the CSC replacement ratio, but the compressive strength at 4 hours and at 7 days decreased as the CSC content increased. Nevertheless, even with up to 40% CSC incorporation, the compressive strength satisfied the requirements for concrete repair materials, suggesting potential for practical use in repair applications.

https://doi.org/10.11112/jksmi.2025.29.6.127

이선호(Sunho Lee) ; 김상우(Sang-Woo Kim) ; 이정윤(Jung-Yoon Lee)

This paper presents an experimental evaluation of the effect of hanger reinforcement on reinforced concrete deep beams subjected to indirect loading from a transverse beam. Experiments were conducted on a directly loaded specimen and two indirectly loaded specimens. One of the indirectly loaded specimens was reinforced with hangers to investigate the shear strength contribution and the change in load transfer path. The shear span-to-overall depth ratio of the main beam was consistently 1.5 for all specimens. Experimental results showed that the indirectly loaded specimen exhibited a shear strength up to 0.28 times that of the directly loaded one. However, the indirectly loaded specimen with hanger reinforcement showed a strength more than double that of the unreinforced specimen, confirming a recovery in shear performance. Additionally, the analysis of crack patterns and the concrete strut revealed that the hanger reinforcement successfully altered the load transfer path. Despite the strain in the hanger reinforcement approaching yield, its actual contribution to shear strength was limited to approximately 59% of the calculated value, indicating a failure to fully restore the shear strength to the level of the directly loaded specimen. This suggests that while hanger reinforcement is effective in enhancing the shear strength of deep beams under indirect loading, there are limitations to its performance.

https://doi.org/10.11112/jksmi.2025.29.6.134

박예진(Yaejin Park) ; 임형진(Hyung Jin Lim)

This study presents the development of a non-destructive Glass Fiber Reinforced Polymer (GFRP) rebar identification technique based on Ground Penetrating Radar (GPR) B-Scan images (radargrams). Conventional steel rebars are vulnerable to corrosion and require costly repairs and maintenance, while GFRP rebars are gaining interest due to their lightweight, non-corrosive, and high tensile strength characteristics. GPR is a non-destructive inspection technique for concrete structures using electromagnetic waves. The existence, location, and diameter of steel rebars can be identified using the electromagnetic waves reflected from them, owing to significant dielectric constant differences with concrete. However, for GFRP rebars, it becomes challenging because the dielectric constant difference between GFRP rebars and concrete is much less than that of steel rebars. Thus, in this study, the existence and location of GFRP rebars are estimated. First, electromagnetic waves reflected from GFRP rebars were enhanced by averaging multiple spatially obtained GPR radargram images to detect GFRP rebars. Then, the location of GFRP rebars was estimated using conventional hyperbola fitting of the reflected electromagnetic waves. The performance of the developed technique was experimentally evaluated using concrete specimens with actual GFPR rebars. The results indicate that GFRP rebars are more clearly identified with this technique, enabling approximate depth estimation.

https://doi.org/10.11112/jksmi.2025.29.6.142

김민수(Minsu Kim) ; 강성근(Seong Keun Kang) ; 황순규(Soonkyu Hwang)

This introduction proposes a pattern recognition-based point cloud data processing technique for the automated detection and visualization of bolt loosening in steel structures. The proposed technique detects loosening by quantifying bolt height based on distance information acquired from a depth vision camera. The automated detection of bolt joint loosening proceeds by first extracting the bolt joint region from point cloud data based on depth data. Subsequently, to ensure data quality, a planarization process is performed using Principal Component Analysis (PCA). Finally, to facilitate the recognition of the bolt region, a coordinate transformation is applied relative to the steel plate, thereby aligning the data. To detect bolt loosening, we quantified bolt height using a k-d tree-based radius search and the DBSCAN algorithm, and subsequently visualized the loosening. The proposed algorithm minimizes noise and distortion caused by on-site environmental variables, thereby enabling reliable inspection results even in outdoor field environments. Implemented using a portable inspection hardware system, the proposed method was experimentally validated on bolted joints of steel structures, demonstrating its ability to quantitatively detect and visualize bolt loosening with a a Root Mean Square (RMS) error within 1 mm. Furthermore, lab-scale experiments were conducted to evaluate the accuracy of bolt loosening detection across a range of separation distances (100 mm to 500 mm) and measurement angles (30° to 90°), thereby establishing the reliability of the technique for practical field applications.

https://doi.org/10.11112/jksmi.2025.29.6.153

최병일(Byoung-Il Choi) ; 이상영(Sang-Young Lee) ; 송건호(Keon-Ho Song) ; 심윤태(Yun-Tae Sim) ; 박세준(Se-Jun Park)

In Korea, ground subsidence accidents are occurring in the upper layers during tunnel construction. Such as the Bujeon-Masan double-track railway and the Shinansan Line in the metropolitan area. For various reasons, The initial collapse occurred inside the tunnel due to the internal lining being unable to withstand the load, and ground subsidence occurred as a result of the failure to support the upper stratum. As a result of domestic and foreign tunnel accidents analysis, it was found that most accidents occur during tunnel construction and the main reason is the load acting on the back of the tunnel lining. In general, when designing for urban tunnel construction, the design should be carried out a properly considering the load on the tunnel, and construction will be necessary considering high-quality materials and construction procedures. In this paper, risk assessment was conducted according to the cavity size at the back of the segment lining for TBM tunnel which is mainly applied when constructing tunnels under the metropolitan area passing of the city.

https://doi.org/10.11112/jksmi.2025.29.6.164

강준구(Jungu Kang) ; 김동준(Dongjun Kim) ; 최하진(Hajin Choi) ; 이철우(Cheolwoo Lee) ; 김홍섭(Hongseop Kim) ; 송호민(Homin Song)

This study developed a method to effectively detect and automatically classify major defects that may occur behind the containment liner plate (CLP) of nuclear power plant containments. For this purpose, the impact-response technique was applied by striking the specimen surface and measuring the vibration response using an accelerometer. Analysis of the acquired signals revealed that defect characteristics were not clearly distinguished in the frequency domain, whereas the time-domain signals clearly differentiated not only between sound and defective regions but also among defect types. Accordingly, the time-domain signals were employed as input data for a one-dimensional convolutional neural network (1-D CNN) model, which classified CLP backside defects including voids, kissing bonds, and wall-thinning. The proposed model achieved an accuracy of approximately 98% on the validation dataset and 100% on the test dataset. These results suggest that the proposed approach can complement existing nondestructive evaluation techniques and is expected to contribute to the structural integrity assessment and safety assurance of nuclear power plant containments.

https://doi.org/10.11112/jksmi.2025.29.6.172

현창진(Chang-Jin Hyun) ; 김효정(Hyo-Jung Kim) ; 이방연(Bang-Yeon Lee) ; 김윤용(Yun-Yong Kim)

In this study, mixtures for 3D printing of alkali-activated slag (AAS)-based strain-hardening cementitious composites (SHCC) were developed. The rheological properties, buildability, and post-curing mechanical performance of each mixture were systematically evaluated according to the type and dosage of alkali activators and admixtures. Working time was secured by analyzing the change in rheology based on the activator type, and extrudability and shape retention performance were optimized by controlling the content of the viscosity-modifying agent (VMA) and ethylene vinyl chloride (EVCL). The experimental results showed that the mixture using the Type-B activator (Type-B13%) secured sufficient working time over 150 minutes, demonstrating high applicability for the 3D printing process. It was confirmed that the most stable 8-layer buildability was achieved when 8 wt% of EVCL was added. SHCC specimens extracted from blocks printed with the optimized mixture exhibited distinct strain-hardening behavior and multiple micro-cracks in uniaxial tension tests, recording a high average tensile strain capacity of 4.9%.

https://doi.org/10.11112/jksmi.2025.29.6.180

양우정(Woo-Jung Yang) ; 권태윤(Tae-Yun Kwon) ; 박준석(Jun-Seok Park) ; 김민철(Min-Cheol Kim) ; 안진희(Jin-Hee Ahn)

This study analyzed the effect of changes in member and inspection item weighting factors on the overall safety rating within the bridge condition evaluation system. According to investigation of domestic bridge accidents, 47.83% of the incidents were caused by substructure failures. The current evaluation system assigns relatively low weighting factors to substructures, which may not adequately reflect their structural importance. To analyze the characteristics of the existing safety rating system, various scenarios involving adjustments to the weighting factors of substructures and other components were compared. According to the analysis, increasing the weighting factors assigned to the substructure and durability item resulted in a relative decrease in the weighting factor ratios of other members, whereas decreasing the we.ighting factors assigned to miscellaneous components led to an increase in the weighting factor ratios of other members. And, change in the weighting factor ratios of other members was relatively small compared to change in the adjusted weighting factors. Furthermore, changes in the weighting factor ratios by member and inspection item affected the overall bridge safety grade such that the grade changed depending on whether the relative weighting factor ratio changes exceeded the range that determine the safety grade. When the weighting factor ratio changes was relatively small, the determination range were not altered, and consequently, the overall safety grade remained unchanged. This indicates that the weighting factor of a specific member or inspection item directly influences the overall evaluation result; therefore, it is necessary to determine appropriate weighting ratios by quantitatively assessing the relative influence of each member or component on the overall safety grade. In addition, for determining the bridge safety grade, it may be appropriate to reflect structural characteristics by separating the superstructure and substructure for independent evaluations, or by evaluating them separately and then determining the overall bridge grade.

https://doi.org/10.11112/jksmi.2025.29.6.189

유용래(Yong-Rae Yu) ; 이창길(Changgil Lee) ; 김현기(Hyun-Ki Kim) ; 안호준(Hojune Ann)

Railway cut slopes are critical yet traditionally inspected by labor?intensive visual surveys and text?based reports, which limit objectivity, efficiency, and long?term data use. This study develops a digital?twin?based smart maintenance platform that creates and manages three?dimensional models of railway slopes. Drone and mobile images are processed with SfM/MVS photogrammetry and deep?learning?based region?of?interest extraction to build lightweight, high?density 3D point clouds. Within the platform, inspectors can measure distance, slope, and deformation, detect geometric changes through cloud?to?cloud comparison, and automatically generate inspection reports consistent with domestic railway and slope safety regulations. All images, models, and reports are stored and managed in a cloud environment to support project?level history tracking. To validate the platform’s performance, field experiments were conducted at the Osong Railway Test Line. Results showed that 3D?based length measurements differed from design or survey values by about 0.35% (track gauge error 0.005?0.012 m; stepped retaining?wall height error 0.15?0.17 m), and slope angle errors were within 1°. These results confirm the technical feasibility and practical potential of the proposed platform for standardized, data?driven management of railway slopes.

https://doi.org/10.11112/jksmi.2025.29.6.199

신현규(Hyunkyu Shin)

This study compares and analyzes the performance of deep learning-based image processing models for automatic crack detection on building exteriors. To overcome the limitations of traditional visual inspections, which are time-consuming, labor-intensive, and prone to subjectivity, various CNN- and Transformer-based object detection models were evaluated in terms of detection accuracy, inference speed, and model size. The study particularly focuses on the practicality of lightweight models for real-world applications, experimenting with YOLOv11, Faster R-CNN, EfficientDet, and RT-DETR models. Using a public crack dataset, the results show that the YOLOv11 series achieves a high level of accuracy while maintaining fast inference speeds, making it suitable for real-time field inspections. Additionally, a novel Trade-off Score (TO-score) metric was proposed to quantitatively balance accuracy and speed for practical model selection. This research provides valuable insights for optimizing crack detection models in environments with limited computational resources.

https://doi.org/10.11112/jksmi.2025.29.6.207

정재용(Jae-Yong Jeong) ; 이진희(Jin-Hui Lee) ; 윤현도(Hyun-Do Yun) ; 박완신(Wan-Shin Park) ; 최원창(Won-Chang Choi)

This study aimed to address the carbon emissions issue in the cement industry by developing biochar based artificial lightweight aggregates (BALA) and evaluating their applicability as structural construction materials through mortar incorporation. While biochar possesses high carbon sequestration potential, its direct application in concrete poses physical limitations. To overcome these challenges, a coating technique combining bentonite and cement was employed in the fabrication of BALA. Experimental results demonstrated that BALA exhibited physical properties such as particle size distribution, bulk density, and water absorption that were comparable to or superior to those of commercial lightweight aggregates (LWA). In particular, the BT10M specimen achieved a compressive strength of 36.8 MPa, an elastic modulus of 16.5 GPa, and an ultimate compressive strain of 0.0039 mm/mm, indicating excellent mechanical performance and ductility. Microstructural analysis further revealed that BALA with 10 to 20% bentonite content formed dense and uniform coating layers, contributing to enhanced strength and durability. The findings confirm that biochar based aggregates can be effectively applied as structural lightweight aggregates while simultaneously serving as a viable carbon sequestration strategy in the construction industry.

https://doi.org/10.11112/jksmi.2025.29.6.217

이지구(Ji-Gu Lee) ; 김기영(Ki-Young Kim) ; 이태형(Tae-Hyung Lee) ; 이종훈(Jong-Hun Lee) ; 손훈(Hoon Sohn)

This study proposes a displacement estimation method for high-speed railway bridges by integrating measurements from a collocated accelerometer and a frequency-modulated continuous-wave (FMCW) millimeter-wave radar. Conventional accelerometer-based double integration is highly susceptible to low-frequency drift, whereas radar measurements often suffer from noise, line-of-sight limitations, and phase wrapping when the structural displacement exceeds half the radar wavelength. To address these issues, the proposed system automates radar target selection and determines an optimal conversion factor by comparing low-pass-filtered radar displacements with high-pass-filtered accelerometer-based displacements. An accelerometer-aided phase-unwrapping algorithm is then employed to reconstruct continuous phase histories, enabling accurate displacement estimation even under large-amplitude vibrations. Finally, a finite impulse response (FIR) filter is used to fuse low-frequency radar displacement with high-frequency accelerometer displacement, producing a refined displacement signal with reduced noise and drift. Field tests conducted on an operational high-speed railway bridge in South Korea demonstrate strong agreement with laser Doppler vibrometer measurements, with a maximum root mean square error below 0.1 mm. These results confirm the accuracy and robustness of the proposed sensing system and highlight its suitability for real-time structural monitoring under dynamic train loads.

https://doi.org/10.11112/jksmi.2025.29.6.227

변준영(Junyoung Byun) ; 유정빈(Jungbhin You) ; 홍성남(Sungnam Hong) ; 박선규(Sun-Kyu Park) ; 이창준(Changjun Lee) ; 박승희(Seunghee Park)

In this study, anchor bolt and U-jacket anchorage methods were implemented to prevent debonding between the TRM reinfrocement and the concrete substrate, and the flexural performance of TRM-strengthened beams was compared according to the anchorage method. Experimental results confirmed that TRM strengthening is most effective near the yielding stage, and thus the yielding stage was defined as the strengthening limit state by synthesizing the results of this study with previous research findings. The flexural performances depending on anchorage configuration were evaluated at this strengthening limit state. No debonding was observed in beams strengthened with either anchor bolts or U-jackets. For beams without preloading, TRM-strengthened beam anchored the anchor-bolt exhibited approximately 11% higher flexural capacity than the U-jacket. When preloading was considered, the performance difference between the two anchorage methods was reduced. However, TRM-strengthened beams anchored the anchor bolt still demonstrated a 6% higher flexural capacity. Additionally, the incorporation of anchorage effectively prevented the reduction in strengthening efficiency caused by preloading. It was also verified that TRM strengthening provides effective flexural enhancement when applied to beams with damage levels below 80% of the yield load.

https://doi.org/10.11112/jksmi.2025.29.6.236

이의성(Eue-Sung Lee) ; 서지석(Ji-Seok Seo) ; 김윤용(Yun-Yong Kim)

This study evaluates the carbon reduction potential of high-performance clinker (HPC) and high-performance clinker cement (HPCC) by partially replacing limestone in the clinker raw mix with CaO-rich KR slag, a by-product of the steelmaking process. KR slag is used as a non-carbonate raw material, and clinker-specific CO2 emission factors were calculated using a Tier 3 methodology for various KR slag replacement levels and raw mix compositions, which were then compared with those of ordinary Portland cement (OPC) clinker. The results showed that when approximately 4.32% KR slag was incorporated into the raw mix to produce HPC, the process-related CO2 emissions of this clinker were about 98.5% of those of OPC clinker, indicating that partial substitution of clinker raw materials with KR slag can reduce process CO2 emissions. Using this clinker, HPCC was produced by adding 10% supplementary cementitious materials (5% ground granulated blast-furnace slag and 5% limestone powder), and the same replacement level was applied to OPC for comparison. Durability tests confirmed that the carbonation and freeze?thaw resistance of HPCC were comparable to those of OPC, demonstrating that HPCC with a high level of supplementary materials can achieve performance equivalent to OPC while having potential significance as a carbon-reduction technology in the cement industry.

https://doi.org/10.11112/jksmi.2025.29.6.246

에르덴바타르 아미나(Amina Erdenebaatar) ; 다르한바트 할리오나(Khaliunaa Darkhanbat) ; 허인욱(Inwook Heo) ; 최승호(Seung-Ho Choi) ; 김강수(Kang Su Kim)

This study conducted fire simulations for typical office buildings by considering heat release rate, compartment floor area, and fuel type as primary variables, and established a database of ceiling-level time-temperature histories. Based on the constructed dataset, an Artificial Neural Network (ANN) model and a regression-based simplified time-temperature prediction equation were developed. To examine the applicability of the proposed models, an additional fire simulation was performed for a new office building, and the predicted results were compared with the simulation outputs. The mean absolute percentage error (MAPE) of both models was found to be below approximately 23%, indicating a reasonable level of predictive performance. The developed models are expected to serve as practical tools for fire resistance evaluation and as foundational data for performance-based fire design in office buildings.