https://doi.org/10.12652/Ksce.2026.46.4.0313

이정우(Lee, Jeongwoo) ; 정일문(Chung, Il-Moon) ; 홍성훈(Hong, Sung Hun)

Analytical solutions for a line-width stream in an unbounded aquifer have been widely used to estimate streamflow depletion induced by groundwater pumping. However, these solutions neglect stream width by treating the stream as a line, limiting their applicability to wide streams. To address this limitation, this study develops an analytical model for estimating streamflow depletion under finite-width stream conditions. A two-dimensional unsteady groundwater flow equation was formulated by incorporating a source term representing stream leakage and a sink term representing groundwater pumping. An analytical solution for groundwater drawdown was derived using the Fourier-Laplace transform. Based on this, streamflow depletion was formulated in the Laplace domain as an integral over stream width, and the corresponding time-domain solution was obtained via numerical inverse Laplace transform. The developed analytical solution was validated by comparison with numerical results obtained using MODFLOW. The effect of stream width was then evaluated under various hydrogeological conditions, including transmissivity, storage coefficient, streambed hydraulic conductivity, and the distance from the stream boundary to the pumping well. The results showed that the difference between the two analytical solutions decreased as the pumping duration increased. Overall, however, the line-with solution tended to overestimate streamflow depletion compared with the finite-width solution. Furthermore, it is found to be appropriate when the distance between the stream boundary and the pumping well exceeds approximately twice the stream width. The proposed analytical model is applicable to streams of any width, making it a more versatile and widely applicable analytical tool.

https://doi.org/10.12652/Ksce.2026.46.4.0321

김숙희(K im, Suk Hee) ; 정가형(Jeong, Ga Hyung) ; 김성준(Kim, Sung Jun)

Urbanization and increasing automobile use have led to a continuous rise in parking demand in metropolitan areas; however, conventional parking policies, which rely on fixed supply, have limitations in responding effectively to spatiotemporal variations in demand. This study redefines Flexible Parking as a parking management system designed to allow parking supply and demand to be mutually adjusted in response to temporal and spatial changes, and proposes a data-driven procedure for selecting pilot sites for flexible parking, using Suwon City as a case study. The site selection procedure consists of six sequential filtering stages: road geometric structure, safety hazard areas, illegal parking hotspots, block-level parking supply-demand ratio, land use-based parking demand patterns, and hourly traffic volume. A total of 84 initial candidate segments were examined through this stepwise analysis, resulting in 11 segments identified as feasible pilot sites. The final feasibility was determined by combining quantitative priority ratings with field survey evaluations. In result, The total installable capacity was estimated at approximately 340 parking spaces across the 11 segments. This study contributes to the literature by conceptually distinguishing flexible parking from shared parking and reservation-based parking as a broader operational framework, and by presenting an ex-ante, data-driven site selection methodology that can serve as a practical decision-support tool for future evidence-based urban parking policy.

https://doi.org/10.12652/Ksce.2026.46.4.0335

김다인(Kim, Dain) ; 정수민(Jung, Sumin) ; 박민건(Park, Mingeon) ; 허태민(Heo, Taemin)

With the accelerated aging of bridge infrastructure in South Korea, there is an increasing need for a system to rationally prioritize inspection, maintenance, and reinforcement under limited personnel and budgets. However, deterioration prediction models in existing Bridge Management Systems (BMS) have primarily been utilized as independent analysis modules. This reveals a limitation in fully integrating the “evidence provision” and “context-based decision support” required in the actual decision-making stages. To address this, this study presents the applicability of a two-stage integrated decision support framework. This framework combines a machine learning classification model to screen high-risk bridges by predicting safety grades, and a Retrieval-Augmented Generation(RAG)-based question-answering system to provide contextual evidence by retrieving and summarizing past inspection records. A comparative analysis of algorithms (Decision Tree, Random Forest, and XGBoost) using data from approximately 8,890 national highway bridges demonstrated that while XGBoost showed higher overall accuracy, Random Forest was selected as the final prediction model for its superior performance in minimizing the omission of C-grade bridges requiring urgent inspection. Furthermore, the developed RAG system assists administrators' judgment by summarizing past damage recurrences and recommended actions, alongside the current and future (2036) prediction results from the Random Forest model. Across the five query categories Evaluation results demonstrated high faithfulness (93.4 %). By seamlessly connecting prediction, evidence provision, and decision support into a single continuous workflow rather than mere deterioration prediction, this study is expected to contribute to the advancement and practical applicability of future BMS.

https://doi.org/10.12652/Ksce.2026.46.4.351

이강혁(Lee, Kanghyeok) ; 조재홍(Cho, Jaehong) ; 김영환(Kim, Yeonghwan) ; 강상혁(Kang, Sanghyeok)

The construction industry records the highest proportion of occupational fatalities among all industries, and risk assessment has been made legally mandatory to strengthen on-site safety management and prevent industrial accidents. However, risk assessments are often conducted in a perfunctory manner due to limitations in manpower and time, and evaluation consistency may be compromised by reliance on the subjective judgment of individual safety managers. To improve the effectiveness of risk assessment and enhance workers' risk awareness, this study constructed a vector database from accident case and legal data using the Text-Embedding-3-Large model. Relevant cases were retrieved through retrieval-augmented generation (RAG)-based similarity search, and a multimodal generative AI-based support system was developed to automatically generate risk assessment sheets after verifying the relevance of accident cases using GPT-4o. The system also integrates legal information search, additional accident case search, and accident case-based visual material generation using the gpt-image-1 model. A survey conducted with nine safety managers with risk assessment experience yielded scores of 4.56 for educational effectiveness, 4.33 for efficiency, 4.11 for satisfaction, 3.89 for reliability, and 3.78 for effectiveness, confirming its potential for improvement over conventional methods. Respondents also noted that the system reduces repetitive documentation burden by enabling rapid access to accident cases and legal information through simple inputs, and that accident case-based visual materials were effective in improving understanding of hazard factors and supporting worker safety education. The proposed system can serve as a tool to enhance the practical effectiveness of risk assessment and support worker safety education at construction sites, and future research should expand its application to a wider range of trades and construction environments.

https://doi.org/10.12652/Ksce.2026.46.4.0363

박영훈(Park, Younghoon)

In this study, a Three-Multihead semantic segmentation framework was proposed for the simultaneous segmentation of crack, rebar exposure, and spalling in reinforced concrete structures. The proposed model incorporates Knowledge Distillation (KD), Pairwise Prior, and Extra Loss to improve composite-damage segmentation performance by transferring knowledge from single-damage teacher models, reflecting spatial interactions among damage classes, and suppressing overlapping activations. Experimental results showed that the proposed model achieved a 13.5 % improvement in mean Dice compared with the baseline model and a 20.0 % improvement compared with the Multi-Damage Representation model. The ablation study further demonstrated that Pairwise Prior and Extra Loss effectively reduced inter-class interference and overlapping predictions. To evaluate performance under real composite-damage conditions, a dataset consisting of 44 composite-damage images was constructed. Without additional training, the proposed model achieved a mean Dice score of 0.35, which increased to 0.57 after fine-tuning using the composite-damage dataset. Class-specific threshold optimization provided a modest calibration effect, resulting in a final mean Dice score of 0.58 while improving the prediction balance across damage classes. These results indicate that composite-damage data-based learning is effective for improving model adaptation to real composite damage environments and suggest that the proposed Three-Multihead framework is a promising approach for enhancing composite-damage segmentation performance.

https://doi.org/10.12652/Ksce.2026.46.4.0377

방은석(Bang, Eun Seok) ; 정상원(Jeong, Sang Won) ; 임신혁(Yim, Sin Hyuk)

This study presents a preliminary evaluation of a ground-proximate magnetic survey system mounted on a quadruped robot (Unitree Go2 EDU Plus), equipped with a fluxgate magnetometer (Sensys MagDrone R3) and a multi-frequency RTK-GNSS, for landmine detection at border-area construction sites. Conventional manual detection poses significant safety risks to operators, while UAV-based magnetic surveys face limitations in securing sufficient magnetic anomaly signals from landmines when low-altitude flight is restricted by vegetation or terrain relief. By maintaining a constant sensor height of 0.5 m above the ground surface, the proposed system is expected to offer a relatively favorable signal-to-noise ratio (SNR) compared to UAVs under conditions where low-altitude flight is difficult due to vegetation. A data acquisition and processing workflow was established, comprising grid-based traverse surveys, zero-phase low-pass filtering (fc = 2 Hz) for gait-induced noise suppression, and target localization using the total horizontal derivative(THD). Field verification was conducted in both a controlled environment and a riverside floodplain using M15 anti-tank and M16 anti-personnel practice mines. Under the tested conditions, magnetic anomaly signals were identified for all targets. Landmine types were distinguished using peak-to-peak amplitude and the spatial extent of magnetic anomaly zones, with THD based position estimation errors within 0.37 m. These results provide an initial confirmation that the quadruped robot-based magnetic survey system has potential applicability for landmine detection at border-area construction sites under the tested conditions. Further validation under actual burial conditions, diverse terrain environments, and various noise sources is required before practical deployment.