Abstract
Proper placement of load-bearing elements plays a critical role in earthquake-resistant building design. Design mistakes and deficiencies in the arrangement of columns, shear walls, and beams may reduce seismic performance, increase torsional effects, cause frame discontinuities, and lead to inefficient or uneconomical structural solutions. Therefore, early identification of such deficiencies directly from structural floor plans can provide important support during both preliminary design and rapid assessment stages. This study presents an automated methodology for detecting, classifying, and evaluating load-bearing elements in RC building floor plans by integrating image processing and deep learning techniques. The YOLO object detection algorithm was employed to identify structural components within floor plan images. A dataset comprising 500 RC building floor plans was developed, and structural elements were manually annotated and labeled for model training and validation. Unlike previous approaches that mainly focus on limited structural components, the proposed model detects columns, shear walls, and four different beam types according to their support conditions. In addition, detected bounding box coordinates are converted into real-world dimensions using grid distances obtained from drawings, allowing the cross-sectional dimensions of each element to be determined. The model achieved high detection performance, particularly for columns, shear walls, and main beams, with accuracy values exceeding 93%. The results show that the proposed approach can accurately obtain information from structural floor plans and provide a promising decision-support system.
Address
(1) Ozan İnce, Erkin Eren, Burak Çakil, Muhammed Atar:
Department of Civil Engineering, Fırat University, Elazig, Türkiye;
(2) Perihan Karaköse:
Department of Electronic and Automation, Bartin University, Bartin, Türkiye;
(3) Elif Gökçe İnce:
3Department of Civil Engineering, Erzincan Binali Yıldırım University, Erzincan, Türkiye.
Abstract
Stiffened panels are widely used in engineering due to their high strength-to-weight ratio. However, parametric correlation analysis and optimization of such panels typically require extensive datasets, and even finite element analysis (FEA) suffers from high computational costs and time-consuming processes. To address these challenges, this study proposes a method based on the Laminate Smeared Stiffener Method (LSSM), which simplifies data acquisition and structural optimization. By equivalently representing multiple stiffened panels of varying dimensions as a single laminated panel, the LSSM eliminates the need for repetitive geometric modeling, meshing, and stiffness matrix generation in FEA. Parametric FEA was employed to extract the maximum deformation and buckling critical load for each dimensional configuration, and a corresponding dataset was constructed. Four regression algorithms were trained on this dataset, with Ridge regression demonstrating the highest prediction accuracy. This algorithm was selected to establish a surrogate model for stiffened panel FEA. Subsequently, the optimization design of stiffened panels was carried out using a genetic algorithm with the surrogate model as the fitness function. The results demonstrate that the LSSM efficiently generates the large datasets required for optimization. In single-objective optimization, while maintaining the original design volume, the optimized panels exhibited a 47.97% improvement in anti-deformation capability and an 87.33% increase in buckling load capacity. For multi-objective optimization, the volume was reduced by 27.60% without compromising performance. This method significantly reduces computational costs and optimization times compared to traditional FEA-genetic algorithm approaches, offering an efficient and accurate solution for stiffened panel optimization.
Key Words
genetic algorithm; laminate smeared stiffener method; ridge regression; stiffened panels; surrogate model
Address
(1) Chen Guo, Zheng Yang, Yanchao Yue:
School of Human Settlements and Civil Engineering, Xi'an Jiaotong University, Xi'an 710000, China;
(2) Chen Guo:
Earthquake Research Institute, The University of Tokyo, Tokyo 113-0032, Japan.
Abstract
This article proposes a method for detecting surface cracks in building concrete based on improved YOLOv8. By introducing deformable attention mechanism (DAttention) in the backbone network, the crack growth trend can be dynamically focused; Enhance multi-scale feature expression capability in neck design (Cross scale Feature Fusion Module, CCFM); Embedding Efficient Channel Attention (ECA) in the head to enhance the weight of key features; And replace the Complete Intersection over Union (CIoU) loss function with the Scale invariant Intersection over Union (SIoU) loss function to optimize the bounding box regression process. The experimental results show that our method achieved a detection accuracy of 88.4%, a recall rate of 95.2%, and an average precision mean (mAP) of 96.4% on a self built dataset, which is significantly improved compared to the benchmark YOLOv8 model. This method is capable of extracting crack features in a complete and continuous manner, effectively identifying subtle cracks and suppressing background interference, providing reliable technical support for the health assessment and safety risk prevention of building structures.
Key Words
attention mechanism; building concrete; crack detection; feature fusion; improved YOLOv8; loss function
Address
School of Architectural Engineering, Wuhan City Polytechnic Wuhan 430068, China.
Abstract
Vibration-based Structural Health Monitoring (SHM) systems are susceptible to false alarms, particularly due to environmental influences. This study proposes that nonlinear static pushover Finite Elements (FE) analysis outputs can be used as complementary data to existing methods in Smart Building (SB) decision support systems. The proposed hybrid approach has the potential to reduce false alarm rates compared to decision-making based solely on modal parameter changes, statistical methods, and machine learning. To this end, in addition to these approaches, a secure proposal based on traditional procedures for integrating nonlinear structural analysis results into intelligent SHM systems is presented. A global performance evaluation of a core system was conducted to obtain local damage distribution and degrees at the element level. In the SHM system, the focus was on using element damage level thresholds obtained from system analysis as threshold values, rather than general limits in standards or solely the results of the structural system's specific global performance analysis. Today, thanks to advancements in hardware and software technologies, performing and disseminating these analyses is much easier than in the past. In the near future, the information obtained from these analyses will inevitably be further utilized in SHM systems. Although the study did not include experimental validation and model updating, it was shown that using the proposed physical thresholds in conjunction with existing modal-based SHM methods has the potential to reduce false alarm rates and provides a theoretical basis for rapid decision-making mechanisms after earthquakes.
Key Words
earthquake engineering; finite element modeling; hazard monitoring; nonlinear structural behavior; seismic protection; Structural Health Monitoring (SHM); system identification
Address
(1) Varol Koç:
Ondokuz Mayis Univercity, Faculty of Engineering, Department of Civil Engineering, Samsun, Türkey;
(2) Hasan Yılmaz:
Ministry of National Defense, Ankara, Turkey.
open access
