In an era of massive construction, damaged and aging infrastructure are becoming more common. Defects, such as cracking, spalling, etc., are main types of structural damage that widely occur. Hence, ensuring the safe operation of existing infrastructure through health monitoring has emerged as an important challenge facing engineers. In recent years, intelligent approaches, such as data-driven machines and deep learning crack detection have gradually dominated over traditional methods. Among them, the semantic segmentation using deep learning models is a process of the characterization of accurate locations and portraits of cracks using pixel-level classification. Most available studies rely on single-model knowledge to perform this task. However, it is well-known that the single model might suffer from low variance and low ability to generalize in case of data alteration. By leveraging the ensemble deep learning philosophy, a novel collaborative semantic segmentation of concrete cracks method called Co-CrackSegment is proposed. Firstly, five models, namely the U-net, SegNet, DeepCrack19, DeepLabV3-ResNet50, and DeepLabV3-ResNet101 are trained to serve as core models for the ensemble model Co-CrackSegment. To build the ensemble model Co-CrackSegment, a new iterative approach based on the best evaluation metrics, namely the Dice score, IoU, pixel accuracy, precision, and recall metrics is developed. Results show that the Co-CrackSegment exhibits a prominent performance compared with core models and weighted average ensemble by means of the considered best statistical metrics.

Structural damage inspection is a key structural engineering technique that strives for ensuring structural safety. In this regard, one of the major intelligent approaches is the inverse analysis of structural damage using evolutionary computation. By considering the recent advances in this field, an efficient hybrid objective function that combines the global modal kinetic and modal strain energies is introduced. The newly developed objective function aims to extract maximum dynamic information from the structure and overcome noisy conditions. Moreover, the original methods are usually vulnerable to the associated high multimodality and uncertainty of the inverse problem. Therefore, the oppositional learning (OL) for population initialization and convergence acceleration is first adopted. Thereafter, the unified particle swarm algorithm (UPSO) mechanism is combined with another newly developed algorithm, the gradient-based optimizer (GBO). The new algorithm, called the oppositional unified particle swarm gradient-based optimizer (OL-UPSGBO), with the convergence acceleration feature of (OL), enhances balanced exploration-exploitation of UPSO, and the local escaping operator of GBO is designed to specifically deal with the complex inverse analysis of structural damage problems. To authenticate the performance of the OL-UPSGBO, the complex benchmark set of CEC 2017 is adopted to compare the OL-UPSGBO with several original metaheuristics. Furthermore, the developed approach for structural damage identification is tested using several damage scenarios in a multi-story frame structure. Results show that the developed approach shows superior performance and robust behavior when tackling the inverse analysis of structural damage.

In the era of massive construction, damaged and aging infrastructure are becoming more common. Defects, such as cracking, spalling, etc., are main types of structural damage that widely occur. Hence, ensuring the safe operation existing infrastructure through health monitoring has emerged as an important challenge facing structural engineers. In recent years, intelligent approaches, such as data driven machine and deep learning crack detection, gradually dominate over traditional methods. Among them, the semantic segmentation using deep learning models is a process of characterization of accurate location and portrait of cracks using pixel level classification. Most available studies rely on single model knowledge to perform this task. However, it is well-known that the single model might suffer from low variance and low ability to generalize in case of data alteration. By leveraging the ensemble deep learning philosophy, a novel corporative semantic segmentation of concrete cracks method called Co-CrackSegment is proposed. Firstly, five models, namely the U-net, SegNet, DeepCrack19, DeepLabV3-ResNet50, and DeepLabV3-ResNet101 are trained to serve as core models for the ensemble model Co-CrackSegment. To build the ensemble model Co-CrackSegment, a new iterative approach based on the best evaluation metrics, namely the dice score, IoU, pixel accuracy, precision, and recall metrics is developed. Results show that the Co-CrackSegment exhibits a prominent performance compared to core models and weighted average ensemble by means of the considered best statistical metrics.

The 300 m-level high arch dams built in China have different degrees of valley-narrowing deformation in the initial impoundment. This unexpected phenomenon may increase the long-term safety risk of high arch dams. There are several explanations for the valley-narrowing deformation phenomenon, and one of the explanations—that it is caused by a drop in bedrock temperature—is still under debate. In this paper, coupled thermo-hydromechanical (THM) simulations of the temporal and spatial evolutions of the bedrock temperature and seepage fields of the Xiluodu arch dam site were performed. The results showed that the mechanism of the bedrock temperature drop after impoundment can be divided into two types: short-term type and long-term. For the short-term type, the maximum temperature drop of 9 °C is caused by the heat exchange between the bedrock surface layer and the reservoir water during the initial impoundment. For the long-term type, the maximum temperature drop of 15 °C in front of the grouted curtain is caused by the long-term infiltration of cold water from upstream to downstream bypassing the grout curtain. The monitoring data showed that the valley-narrowing deformation mainly occurred in the initial impoundment, which corresponds to the short-term type. Notably, the valley-narrowing deformation may be overestimated if the long-term type of bedrock temperature drop is considered. Taken together, these findings demonstrate that the bedrock temperature change due to impoundment is dominated by the short-term type and can be applied to study the effect of temperature on valley-narrowing deformation after impounding.

The concrete panel of earth-rock dams in cold regions tends to crack due to the combination effect of non-uniform foundation settlement, ice expansion loads, and freeze-thaw damage. In this work, simulations are designed to investigate the effects of freeze-thaw damage degrees on the fracture behavior caused by the partial detachment and ice expansion loads on concrete panels. Results show that the range of detached panels and freeze-thaw damage degree are the dominant factors that affect the overall load-bearing capacity of the panel and the failure cracking modes, whereas the panel slope is a secondary factor. The failure cracking modes are found to be a combination of type II fracture modes, including compression-shear and tension-shear cracking. Besides, with the increase of detachment area and the degree of freezing-thawing damage, the concrete panels are prone to appear bending tension cracks at the bottom and a more serious smeared crack layer at the top.

As a discontinuous deformation method, the block element method (BEM) characterizes a material’s elastoplastic behavior through the constitutive relation of thin-layer elements between adjacent blocks. To realistically simulate rock damage paths, this work improves the traditional BEM by using random Voronoi polygonal grids for discrete modeling. This approach mitigates the distortion of damage paths caused by regular grids through the randomness of the Voronoi grids. As the innovation of this work, the iterative algorithm is combined with polygonal geometric features so that the area–perimeter fractal dimension can be introduced to optimize random Voronoi grids. The iterative control index can effectively improve the geometric characteristics of the grid while maintaining the necessary randomness. On this basis, a constitutive relation model that considers both normal and tangential damage is proposed. The entire process from damage initiation to macroscopic fracture failure in rocks is described using two independent damage surfaces and a damage relationship based on geometric mapping relationships. The analysis results are in good agreement with existing experimental data. Furthermore, the sensitivity method is used to analyze the influence of key mechanical parameters in the constitutive model.