ABSTRACT To accurately predict the tensile fatigue life of reinforced thermoplastic pipes (RTPs) and clarify the mechanism of fatigue damage evolution, this research proposes a multiscale analytical model based on continuum damage mechanics (CDM). Built upon the three‐dimensional constitutive relations of RTPs and the load–deformation compatibility conditions, the model solves the stress–strain responses of individual plies at the macroscopic level and achieves cross‐scale stress decomposition from plies to fiber/matrix phases via a bridging model. Within a cycle increment framework, the CDM evolution equations are used to iteratively update stresses and damage variables, enabling prediction of residual stiffness evolution and fatigue life. Damage parameters are calibrated using composite fatigue experiments, and the predicted fatigue life and key responses are validated against experiments and finite element simulations. The results demonstrate that the proposed model can predict the tensile fatigue life of RTPs under different load ratios with good accuracy and further reveal stress redistribution and damage evolution laws across different scales, providing an efficient multiscale analysis framework for interpreting RTP fatigue mechanisms and conducting structural assessment.

Asphalt pavement distress detection plays a pivotal role in highway maintenance, providing an essential basis for optimizing maintenance strategies and allocating funding. Consequently, quick detection and efficient identification of distress are crucial for enhancing the quality of highway maintenance. This study aims to acquire high-precision distress data using 3D laser point cloud technology, identify distress types via the YOLO algorithm, and extract geometric features such as length and angle. Specifically, a recognition method based on 3D laser point cloud images is proposed, where point cloud data are converted into planar images for processing. Experimental results indicate that the laser point cloud detection achieves millimeter-level precision, the distress recall rate exceeds 85%, and the identification precision reaches 79.5%, demonstrating satisfactory detection accuracy and efficiency.

Detecting and measuring cracks are crucial for ensuring the safety of civil infrastructures. Traditional fully supervised methods require an amount of high-precision labeled data, making them time-consuming to deploy. In this paper, a semi-supervised learning network model designed for crack segmentation is proposed to address this issue, which incorporates a mutual consistency constraint and a boundary loss function. The mutual consistency constraint enables the model to utilize information from unlabeled data, thereby improving its performance and efficacy. Meanwhile, the boundary loss function enhances the model's ability to predict images when the background pixels outnumber the crack pixels. To comprehensively evaluate the model performance, a new dataset featuring various environmental interferences is constructed. The model optimal hyperparameters and architecture are determined through the experiments and an ablation study. To highlight the advantages of the proposed network model, its prediction results are compared with other segmentation models. It is demonstrated that the proposed model delivers high-precision and robust results while requiring less labeled data.

To ensure the accurate applications of the cohesive zone models (CZMs), it is crucial to calibrate the interface parameters. In this study, considering both the experimental and numerical data including the load-displacement curves and crack length of interface, the identification of interface parameters is conducted through an inverse identification framework based on the multi-island genetic algorithm (MIGA). Due to the fact that the debonding in experiments may occur at random locations and expand in various directions, a nonlinear multi-objective function taking the load-displacement curve and crack length as input data are presented to formulate the identification problem. A numerical model that reproduces the four-point bending test is established, and the interface elements are implemented based on the strength-based cohesive zone model (ST-CZM). The interface parameters of the FRP-concrete in four-point bending model are identified using the inverse identification framework. The failure modes observed in the experiments are accurately reproduced by the numerical model with the identified parameters, thereby verifying the accuracy and effectiveness of the proposed method.

In this paper, a Lennard-Jones potential based three-dimensional cohesive zone model (LJCZM3D) is proposed, and its correctness is verified through various loading combinations of patch tests. Based on MD, the phase-field model, and LJCZM3D, a three-dimensional sequential multiscale damage modeling approach for graphene reinforced epoxy nanocomposites is established. The MD simulations are employed to calibrate the parameters of LJCZM3D. Additionally, a 3D graphene-reinforced epoxy RVE model is developed using a complex Python program. The process involves randomly placing graphene in 3D space, identifying and extracting graphene elements, inserting zero-thickness cohesive elements on both sides of the graphene, and applying 3D periodic boundary conditions. Thereafter, the phase field damage model and LJCZM3D are used to simulate the damage of epoxy resin and the mechanical behavior of the graphene-epoxy interface, respectively. Finally, the effects of graphene volume fraction, aspect ratio, orientation, and surface buckling on the mechanical performance of the 3D RVE model are analyzed through a detailed parameter study.Highlights A new 3D Lennard-Jones potential-based cohesive zone model is proposed. A 3D sequential multiscale damage modeling approach is established. Extensive MD simulations are employed to obtain the key parameters of the model. The model is implemented and validated through the development of the UMAT subroutine. The influences of several graphene morphological factors are investigated.

ABSTRACT Fatigue damage in asphalt pavements is a critical issue affecting the durability and safety of road infrastructure. Traditional fatigue testing methods, such as the indirect tensile fatigue test, fail to replicate the alternating tension-compression stress fields experienced in real-world conditions, leading to inaccuracies in fatigue life predictions. This study investigates the bidirectional splitting fatigue test as an alternative method to better simulate the stress state of asphalt pavements. Using cylindrical AC-13 asphalt specimens subjected to varying stress ratios (0.3, 0.4, 0.5, and 0.6) at 20°C and 10 Hz frequency, the research evaluates vertical displacement trends and fatigue life. Results reveal that bidirectional splitting induces a more realistic stress response, with reduced permanent deformation and slower fatigue progression compared to indirect tensile testing. At lower stress ratios, bidirectional splitting enhances material durability by leveraging compressive stresses for crack healing, whereas higher stress ratios lead to shear failures. These findings underscore the bidirectional splitting test’s potential to improve fatigue performance assessment, paving the way for more resilient asphalt mixtures. The practical implications of these findings lie in their potential application to real-world pavement design and maintenance. Future research should explore its applicability to various asphalt types and real-world loading conditions.

A new mixed-mode cohesive zone model based on Lennard-Jones potential (LJCZM) is proposed to simulate the interface failure between graphene and epoxy matrix. The values of model parameters are obtained from a large number of molecular dynamics simulations, and a UMAT subroutine is programmed and validated to introduce this model into the ABAQUS platform. This process spans from the nanoscale to the microscale, which provides a new routine for the multiscale damage modeling of the graphene reinforced epoxy nanocomposite at microscale. In addition, the continuous damage phase-field model is used to simulate the matrix damage, and the values of model parameters are determined from the molecular dynamic simulations of the bulk epoxy at nanoscale. At last, the effects of parameters such as volume fraction, aspect ratio, orientation, and curvature of graphene nanoplatelets are investigated. The results indicate that the nanocomposite reinforced with high content and large aspect ratio graphene nanoplatelets presents the lower ultimate stress and fracture strain. In addition, the orientation and waviness of the graphene also significantly affect the mechanical properties of the nanocomposites. The nanocomposite reinforced with graphene platelets with greater waviness has higher stiffness and strength but lower toughness. The rationality and effectiveness of the model are verified through comparison with other existing results.

Coal gangue (CG) is an aluminosilicate solid waste of low reactivity that limits its utilization in cement. The effects of 30 wt.% replacement of slag with thermal-activated coal gangue (CG) on the properties of supersulfated cement (SSC) were studied and compared with CG-blended (30 wt.%) Portland cement. It was found that thermal activation of the CG increases the pozzolanic reactivity due to the amorphization of the kaolinite, reducing the binding energy of the Si-O tetrahedra/Al-O octahedra that leading to a higher dissolution rate of the aluminate and silicate phases. The pozzolanic reactivity of CG plays a different role in blended cement and SSC: a higher reactive CG contributes to the greater formation of the C-S-H gel and compressive strength development in blended cement. But the high and low reactive CG achieve the compressive strength of 35.10 MPa and 32.20 MPa at 28 days, which are both comparable to the plain SSC of 36.33 MPa. The findings of this work imply the importance of filling the understanding the role of chemical compositions/reactivity of coal gangue for the preparation of low-carbon binders.

In this study, the compatibility of phase change material(PCM) in asphalt was investigated using several methods. Initially, solubility parameters, radial distribution functions, and relative concentrations of PCM modified asphalt were examined through molecular simulations to explore its compatibility. Subsequently, the phase structures of different PCM modified asphalt were analyzed using fluorescence microscopy experiments on both upper and lower samples. Finally, the compatibility differences among PCM modified asphalt samples with varying naphthenic oil contents were evaluated through softening point and Brookfield viscosity tests. The results indicate that higher naphthenic oil content in encapsulated PCM leads to closer solubility parameter alignment with the base asphalt. At a naphthenic oil content of 50% in encapsulated PCM, significant interactions occur between PCM and asphalt components. The concentration distribution curve of PCM molecules in the asphalt approaches 1.0, indicating excellent compatibility between encapsulated PCM and asphalt. Overall, the results from fluorescence microscopy and conventional tests are consistent with simulation outcomes. The only discrepancy observed is that conventional experimental results for the PCM-60% sample slightly outperform those for the PCM-50% sample.

Cellulose nanocrystals (CNCs) were obtained from the extraction and bleaching of jute cellulose as the enhancer, 2-hydroxypropyl-β-cyclodextrin (HP-β-CD) as the carrier, the flavonoids-anthocyanidins and cinnamaldehyde as the bioactive agent, and finally a novel kind of polylactic acid (PLA)-based composite membrane was derived by electrostatic spun method. With the increasing concentration, HP-β-CDs cooperated with CNCs to regulate or control the release rate of bioactive compounds, which had a synergistic effect on the performance of the PLA matrix. The mechanical strength of PLA-3.2 composite with tannic acid (TA) surface cross-linking was 29.6 % higher than neat PLA, and could also continuously protect cells from oxidative stress and free radicals. In addition, excellent cell biocompatibility was found, and attributed to the interaction between bioactive compounds and cell membrane. In addition, we also found two excellent properties from our experimental results: obvious intelligent color reaction and good antibacterial ability. Finally, PLA-3.2 composites could be degraded by soil and are conducive to plant root growth. Hence, this work could solve many of the current problems of biodegradability and functionality of biopolymers for potential applications in areas such as intelligent bioactive food packaging.

Composite interfaces are commonly simulated by cohesive zone models with the key challenge being the calibration of interfacial parameters. A novel inverse identification framework is presented in this paper to determine the interface parameters of cohesive zone models. This approach employs the multi-island genetic algorithm to obtain the key parameters of cohesive zone model which can reproduce the experimental observations. Utilizing two independent metrics, namely the load-displacement response and the debonding length history, an objective function is formulated to give the framework inherent robustness. The interface debonding length is used to uniquely determine the debonding properties of the specimen. To demonstrate the feasibility of the proposed framework, the strength-based cohesive zone model is taken as an example. The inverse algorithm is adopted to identify the interface parameters of both the double cantilever beam (DCB) experiments and the fixed-ratio mixed-mode (FRMM) tests. The robustness, accuracy and sensitivity of the framework are validated through the double cantilever beam test. The findings indicate that the numerical results align closely with the experimental data, confirming that the interface parameters identified by the proposed framework can reproduce the experimental results.

A thermo-mechanical strength based cohesive zone model is proposed to simulate the behaviour of an imperfect interface. In this model, the load transfer behaviour is described by the strength model rather than the traditional traction-separation law, and the heat transfer behaviour is delineated by the interface conductance including the bonding conductance, the air conductance and the contact conductance. A new damage state variable is introduced in the derivation, which needs to be updated even during the elastic stage. A reasonable assumption is made to reduce the dimension of the failure surface to simplify the model. The proposed model is validated by the comparison of the simulating results and experimental data. It is concluded that the thermo-mechanical strength based cohesive zone model provides the interaction mechanism between interface load and heat transfer and can effectively simulate the interface behaviour at various temperatures.

针对三维探地雷达传统计算方法中直接取芯法和振幅全反射法存在代表性差、计算效率低等问题,文中对直接取芯和振幅全反射两种计算方法的优点与缺陷进行分析,结合三维探地雷达的电磁波在路面内部传播的几何关系,提出一种基于频率步进法(SFL)的计算模型.选取杭绍台高速公路K91+600～K101+700段开展沥青路面厚度和介电常数测试,比较不同方法的计算精度,以验证频率步进法计算模型的准确性.研究结果表明:频率步进法对于路面面层厚度和介电常数的测量误差分别为3.85%和4.69%,相对于振幅全反射法,路面厚度和介电常数测量误差分别降低6.37%和10.59%.频率步进法具有较高的检测精度,能为沥青路面厚度和介电常数测量提供新的计算途径.

在对基于强度理论的内聚力模型(ST-CZM)进行二维整理及三维拓展的基础上,使用Abaqus用户单元子程序(UEL)对该模型进行有限元实现.通过经典界面破坏算例验证ST-CZM有限元模型的有效性和准确性,建立纤维增强复合材料(FRP)加固混凝土模型,采用ST-CZM模拟FRP与混凝土界面间的黏结破坏过程,实现ST-CZM在复杂工况下的应用.相较于传统基于牵引力准则的内聚力模型,ST-CZM具有更灵活的混合模态耦合方式,且其切向和法向的强度模型相互独立,ST-CZM的收敛性更好,对强度的预测更准确.所有算例表明,相较于传统基于牵引力准则的内聚力模型,ST-CZM有限元模型能够更好地实现黏结界面峰值应力的预测和损伤阶段的模拟.

This study develops a sequential multiscale model to investigate the interface failure of graphene-reinforced epoxy nanocomposites. The molecular dynamics simulations are carried out to track the graphene-epoxy interface behavior through normal opening and tangential sliding modes modeling at nanoscale. Four kinds of functional groups on the graphene including –OH, –NH2, –CH3 and –COOH are considered. The traction-separation law in the normal direction and the shear-lag model in the tangential direction are used to mimic the interface behavior. Furthermore, the scaled boundary finite element method in collaboration with a new hybrid quadtree algorithm is used to discretize the nanocomposite representative volume element at microscale. In addition to the functional group types, the volume fraction and aspect ratio of the graphene nanoplatelets are taken into account in the microscale simulation. The results show that the addition of functional groups to graphene can greatly improve the load-bearing capacity of the interface between nanoplatelets and epoxy resin. Meanwhile, the reinforcement effect of nanoplatelets is related to their aspect ratio and volume fraction. The efficiency and accuracy of the proposed model are validated by comparing the results with traditional finite element method. This study provides a theoretical support and guidance for the design and fabrication of graphene reinforced nanocomposites.

Two supervised machine learning methods, logistic regression and classification and regression trees (CART) are used to predict pavement roughness level using the data collected from the Long Term Pavement Performance (LTPP) database. Analysis results showed that the most significant factors for treatment roughness are pre-treatment roughness, overlay thickness, followed by overlay age, milling treatment and structural number are two marginal significant factors. In addition, high pre-treatment roughness and long service time increase the probability of high roughness; whereas thick overlay, deep milling and strong pavement structural capacity tend to reduce the roughness level. According to the machine learning performance measures including precision, recall, accuracy, and F1_score, the CART decision tree obtain much better classification results than that of the logistic regression. The decision tree uses multiple decision rules and therefore can also generate complex classification. Logistic regression is a linear classification but can provide explicit model form and parameter estimates.

为实现沥青路面裂缝长度和面积等几何信息的智能提取,提出了一种基于改进两步式卷积神经网络的沥青路面裂缝几何信息提取法.在完成数据的采集后,第1步主要通过人工筛选出失真图像、无病害图像和可能存在病害图像各600张,训练基于卷积注意力模块改进的图像分类算法ResNet50,以此作为本模型的清洗算法,完成对410000张原始路面图像的清洗工作,筛除其中的失真图像和无病害图像,构建裂缝病害图像数据集;第2步基于卷积注意力模块对图像语义分割算法U-Net进行改进,使用上一步筛选的数据进行训练与测试,训练集与测试集的样本比例为10:1,以此实现裂缝图像的分割,并对裂缝的长度和面积进行提取.试验结果表明,本研究提出的改进ResNet50算法对于总体样本清洗结果的精确率、召回率和F1值均已超过95％,其中F1值已经达到了96.8％;两步式沥青路面裂缝几何信息提取法的均交并比为0.4967,其中横向裂缝、纵向裂缝、龟裂、块状裂缝的交并比分别为0.4951,0.5467,0.6085和0.3366;在裂缝长度信息提取上,横向裂缝的误差为12.18％,纵向裂缝的误差仅为3.88％;在裂缝面积的信息提取中,横向裂缝、纵向裂缝、龟裂、块状裂缝的误差分别为0.58％,12.14％,12.27％,11.44％.

为了研究集料粒径和体积分数对水稳碎石材料抗裂性能的影响,借助离散元法构建了具有相同集料轮廓与位置、不同集料粒径与含量的圆形劈裂数值试件.通过虚拟间接拉伸试验模拟了水稳碎石细观开裂过程,并进一步研究了集料粒径与体积分数对结构细观开裂行为的影响.结果 表明:集料粒径和体积分数对数值试件的力学性能有着显著影响,降低集料体积分数可以增强结构的变形协调能力,提高试件抗裂性能.对于水稳碎石材料,集料的强度骨架作用对整体抗裂性能的贡献弱于界面的衰减效应,充盈的集料最终会损害结构的整体抗裂强度.集料对结构整体强度贡献不如砂浆基体大,对于集料体积分数较高的水稳材料,改善砂浆-集料界面性能更有益于提高整体结构强度.

为充分利用三维探地雷达采集到的多视图雷达数据以提高其病害分类准确性,提出一种可同时利用多视图雷达图谱数据的端到端双塔模型.将多视图数据作为输入分别训练单视图的特征提取器,其中采用通道间自注意力机制自适应地提取单视图内重要特征;采用视图间注意力机制将双视图特征带权融合,以学习不同病害对于各视图的偏好;将融合特征经过多层感知机和Softmax归一化进行道路病害分类.实验结果表明,提出算法在真实道路数据中精度较高,达到94.62％,采用数据增强后精度进一步提升至95.78％.

Titanium dioxide (TiO2) was recently employed to apply onto road surfaces to degrade the harmful compounds from vehicle emissions. However, it remains a challenging task to find a highly compatible pavement type for TiO2 application to achieve durable and efficient air-purifying performance. This study proposed to coat TiO2 particles onto semi-flexible pavement surface and tried to investigate an optimum coating method. Three coating methods, including direct mixing TiO2 (MT) with asphalt mixture, spraying dry TiO2 (ST) coating and water-solution-based TiO2 (WT) coating on semi-flexible pavement surface. To achieve this objective, semi-flexible samples were prepared to evaluate and compare the performances of three coating methods by employing resistance to wearing, NO removal efficiency tests and residual texture depth tests. It was found that the ST method not only provided better NO degrading efficiency but also improved the resistance to wearing than the other two methods.