Purpose This study aims to develop a system reliability-based topology optimization framework for structural layout design problems involving multiple, dependent failure modes. The goal is to enhance structural safety while efficiently addressing the influence of failure dependence. Design/methodology/approach A D-vine copula model is employed to characterize complex statistical dependencies among limit state functions. To overcome computational challenges arising from evolving topologies and nonlinear performance functions, a quantile-based decoupling strategy is proposed, along with an adaptive reliability allocation method. Findings The proposed methodology is validated through two numerical examples: a planar bridge and a simplified turbine disk. Comparative studies are conducted among deterministic, component-level, system-level and independent reliability-based topology optimization formulations. Results demonstrate that accounting for failure dependence leads to more rational and safer designs. Neglecting such dependence can result in overly conservative or unsafe structural layouts. Originality/value This work provides a novel system reliability-based design framework that integrates failure dependence modeling using D-vine copula model into topology optimization. Although limited to two representative case studies, the findings offer meaningful insights into the trade-offs among different RBTO strategies and emphasize the importance of capturing failure dependence in system-level reliability design.

Purpose Fatigue mechanisms and subsequent crack growth represent the predominant causes of fracture failures in gas turbines. Investigating fatigue crack growth and the fracture behavior of critical gas turbine components is essential for predicting and extending their service life. The paper aims to discuss this issue. Design/methodology/approach Fatigue crack growth tests were conducted on gas turbine combustor liner materials between 500 °C and 860 °C to determine high-temperature crack growth rates. Testing employed a sinusoidal waveform at 10 Hz with a stress ratio of R = 0.1 and a maximum load of 8000 N. Three-dimensional crack growth was simulated using the Paris law with a linear elastic constitutive model, while fracture behavior was evaluated via the J-integral under an elastoplastic framework, considering different start-up conditions (hot, warm and cold starts). Findings The results indicate that, under the given conditions, the combustor liner exhibits the shortest service life during cold starts, only 24.45% of warm start values and a mere 1.14% of hot start conditions, primarily due to the maximum stress intensity factor range. The crack growth rate is relatively slow near the inner surface but accelerates significantly near the outer surface. This phenomenon occurs because the initial crack at the dilution hole edge experiences compressive stress on the inner surface and tensile stress on the outer surface. Originality/value Based on these findings, this study recommends structural reinforcement of the combustor liner's outer surface to improve durability and operational safety under various start-up conditions. Highlights

Purpose With the rapid advancement of computer vision and deep learning, crack detection has transitioned from manual inspection to automated approaches. However, challenges such as varying illumination and environmental noise continue to hinder detection accuracy. This study aims to enhance crack segmentation performance and robustness under complex imaging conditions through noise-augmented training and rigorous model comparison. Design/methodology/approach Noise-augmented versions of public benchmark datasets were employed to train selected segmentation models, thereby enhancing their robustness to illumination variations and noise interference. To evaluate model generalization, a challenging dataset containing 434 images featuring diverse infrastructure types and camera angles was constructed. Two deep learning frameworks, DeepLabv3+ and SegNet, were implemented with various pre-trained backbones, resulting in seven distinct architectures, such as DeepLabv3+Inception-ResNet-v2, for comparative performance analysis. Findings Models trained on noise-augmented datasets exhibited notable improvements in Mean Intersection over Union (MIoU) and F1-score compared with their non-augmented counterparts. Specifically, the DeepLabv3+Inception-ResNet-v2 model achieved the most significant progress and the best overall performance, demonstrating respective increases of 0.7% in Accuracy, 3.2% in Recall, 15.3% in Precision, 4.4% in F1-score and 5.0% in MIoU on the test set. Furthermore, evaluation on the 434-image dataset confirmed the model's high robustness. Originality/value These findings indicate that the proposed network, DeepLabv3+Inception-ResNet-v2, has strong potential for crack segmentation tasks in basic infrastructure, suggesting its applicability in real-world engineering scenarios.

Purpose In this study, a new method for identifying damage in steel beams is presented. Ease of use, high accuracy, calculation volume reduction, output-only and reduction of health monitoring costs have been the main criteria in presenting the new methodology. Design/methodology/approach A new methodology for identifying damages in steel beams is presented based on the use of a Cone-Shape kernel distribution and machine learning. This research helps to improve the accuracy and ability to detect damage in steel beams. Findings To evaluate and ensure the performance of the proposed method, the results of k-fold as well as the results of the percentages of damage obtained from the test scenario were used. The results showed that all introduced machine learning algorithms have significant accuracy in identifying damage. By comparing the results obtained from all machine algorithms, it was found that the MLP Neural Network algorithm has a higher detection accuracy than other algorithms in identifying the intensity and location of damages. The very high capability of the presented methodology in damage detection is such that it detects the presence of several damages at the same time with different severities with very high accuracy. Originality/value The use of bilinear time–frequency distribution combined with machine learning algorithms is a completely new methodology for damage detection in steel beams. The capability of the presented methodology in damage detection is very remarkable. In such a way that the methodology can detect the presence of several damages at the same time with different severities with very high accuracy.

Purpose This study aims to investigate how high-temperature oxidation affects the microstructure and failure mechanisms of SiCf/SiC composites. Understanding the oxidation influences on the mechanical properties and failure modes is crucial for improving the performance and lifespan of SiCf/SiC composites in extreme environments. Design/methodology/approach In situ testing techniques were employed based on synchrotron radiation X-ray computed tomography. Combined with deep learning and the digital volume correlation method, the microstructural changes and crack propagation behavior of SiCf/SiC composites were characterized under different oxidation states. Additionally, scanning electron microscopy was used to analyze the interfacial evolution process, which was further correlated with the observed failure modes. The study focuses specifically on the transition from fiber pull-out damage to brittle fracture behavior as a result of oxidation. Findings The results demonstrate that oxidation significantly alters the microstructure and failure behavior of the SiCf/SiC composites. The formation of annular pores at the interface and the subsequent generation of SiO2 on the fiber and matrix surface degrade the material's mechanical properties. The study reveals that short-term oxidation promotes fiber pull-out damage, while prolonged oxidation induces the brittle fracture mode due to the formation of SiO2, which enhances the bond between the fiber and matrix but limits crack deflection. Originality/value This study provides novel insights into the relationship between oxidation-induced microstructural changes and the mechanical performance of SiCf/SiC composites. The findings contribute to a deeper understanding of the material's behavior under high-temperature oxidative environments and provide guidance knowledge for the design, optimization and application of SiCf/SiC composites in aerospace and other high-performance applications.

Purpose Complex structures are subjected to multiple sources of uncertainty during service life, and their intrinsic failure mechanisms are complicated. It is important to achieve accurate response prediction and reliability analysis as a prerequisite for ensuring structural safety. Design/methodology/approach An improved metaheuristic algorithm enhanced adaptive neuro-fuzzy inference system (ANFIS) (short for IMA-EANFIS) is developed based on ANFIS, C-means clustering and improved snow geese algorithm (ISGA). In this method, the ANFIS is used to describe the black-box relationship between input variables and output responses; the C-means clustering partitions the input space to control the number of fuzzy rules and avoid the “rule explosion”; the ISGA is employed to replace the gradient descent method to achieve global optimization of the ANFIS premise parameters and reduce the risk of trapping in local optima. Findings The results demonstrate that the IMA-EANFIS approach exhibits excellent modeling and simulation performance. Originality/value The proposed IMA-EANFIS method can provide instructive valuable insights for the reliability design and operational support of engineering structures.

Purpose The study aims to propose a new multiaxial fatigue damage parameter. Design/methodology/approach Using the proposed stress analysis method proposed in part 1 and results of the design experimental program, the proposed formula is introduced. Findings The performance of the proposed formula shows encouraging accuracy. This performance is evaluated via the comparison of the proposed formula with other well-known ones. Research limitations/implications The proposed formula requires fatigue input data for non-proportional loadings in the low-cycle regime to predict the fatigue life in that regime. While accessing the available data in high-cycle fatigue is sufficient for life prediction of the non-proportional loadings in the high-cycle regime. Practical implications The paper introduces the concept of the normal-shear stresses interaction in a multiaxial fatigue field. The importance of this phenomenon is evaluated via several available data. Originality/value This paper fulfils an identified need to predict the fatigue life under the proportional and non-proportional multiaxial loadings.

Purpose Steel box girders, a fundamental component in modern bridge construction, experience complex thermal behavior under solar radiation that significantly impacts their structural performance. Design/methodology/approach This study examines temperature distribution patterns in scaled steel box girder models using controlled radiation angles to simulate varying solar exposure conditions. Through experimental testing of three scaled models with different geometric parameters, temperature fields were measured using thermistors and thermal imaging techniques. Findings The results reveal that beam width predominantly affects vertical temperature gradients, with wider sections experiencing temperature differentials up to 8.2 °C. Beam height primarily influences lateral temperature gradients, particularly at higher radiation angles, generating temperature differences up to 6.8 °C. Originality/value Computational validation using COMSOL finite element software enabled the development of a predictive model for vertical temperature field distribution. These findings advance the understanding of thermal behavior in steel box girders and provide a quantitative basis for evaluating temperature-induced stresses in bridge design under various environmental conditions.

Purpose This research explores the potential of employing DIC method to assess the load-deflection behavior of lightweight reinforced concrete (LWRC) beams enhanced with near-surface-mounted (NSM) glass fiber-reinforced polymer (GFRP) bars. Design/methodology/approach This research explores the potential of employing DIC method to assess the load-deflection behavior of LWRC beams enhanced with NSM GFRP bars. An experimental study was presented to investigate the behavior of the strengthened LWRC beams. The lightweight expanded clay aggregate (LECA) was utilized to substitute the natural coarse aggregate (NCA) in concrete mixture to produce LWRC beams. The experimental program comprised 14 LWRC beams subjected to four-point loading until a failure. The beams are classified into two groups: the first group comprised of seven LWRC beams reinforced to fail by flexural mode, while the second group included seven LWRC beams reinforced to fail by shear mode. Within each group, one beam served as the control, lacking any reinforcement, while the remaining samples were strengthened using NSM GFRP bars arranged and configurated in various patterns. In addition, this study uses the DIC method to capture the strain profile and load-displacement behavior of the NSM strengthened beam specimens with deferent locations, numbers, diameters, bond lengths, spacings, angle of inclinations and materials of the NSM GFRP rods. The experimental results showed improvements in the load-carrying capacity of the LWRC beams when NSM GFRP bars are used for flexural and shear strengthening, compared to that of control specimens. Further, the DIC results were compared with those obtained experimentally, revealing that the DIC method offers higher accuracy compared to visual inspection, particularly in the analysis of cracking loads. This DIC capability enables early detection of potential issues before cracks happen in the specimen. Findings The experimental results showed improvements in the load-carrying capacity of the LWRC beams when NSM GFRP bars are used for flexural and shear strengthening, compared to that of control specimens. Further, the DIC results were compared with those obtained experimentally, revealing that the DIC method offers higher accuracy compared to visual inspection, particularly in the analysis of cracking loads. Research limitations/implications The study is applied on simply reinforced concrete beams under two-points loads. Practical implications This DIC capability enables early detection of potential issues before cracks happen in the specimen. Social implications The study helps in sustainable using of reinforced concrete building. Originality/value The study employed the DIC method to assess the load-deflection behavior of LWRC beams enhanced with NSM GFRP bars.

Purpose Rapid and accurate prediction of crack growth paths in materials using machine learning can significantly reduce the resource consumption of traditional phase-field simulations and offer a novel solution for risk assessment of engineering structures, with high potential for industrial application. Design/methodology/approach Two hybrid crack growth path prediction models are proposed, integrating convolutional neural networks (CNN) and recurrent neural networks (RNN). The first is the Densely Connected Convolutional Networks and Gated Recurrent Neural Network combination (DG), while the second is a mixed 3D and 2D convolution model (3-2Dmix). Considering the spatiotemporal continuity of crack growth, CNN and RNN, with their respective spatial and temporal dimensions advantages, are selected for modeling. Findings Experimental results demonstrate that the DG combination performs better in crack growth prediction, characterized by fast convergence and high accuracy. In contrast, the 3-2Dmix combination struggles with temporal sequence processing, leading to suboptimal prediction accuracy. Originality/value The proposed model, combining CNN and RNN, introduces a novel technical framework for crack growth prediction and contributes to the risk assessment of engineering structures.