Closure of main arch rib is the key process of steel box girder arch bridge construction. The way to achieve quantitative and accurate prediction of closure error risk remains to be studied in depth. Taking the construction risk model of closure of main arch rib of Xiangxi Yangtze River Bridge as an example, firstly, the causal relationship between prior data and closure error of main arch rib is considered to construct Bayesian network topology; Further, the central limit theorem (CLT) is used to update the prior distribution of the parent nodes in the network using dynamic multi-source monitoring data; Then the finite element analysis and BP neural network are combined to reproduce the corresponding relationship between the arch rib closure error and the influencing factors. Finally, the probability distribution is determined by the Monte Carlo method, and the quantitative prediction of the joint construction risk error is realized. The analysis data shows that the risk probability of arch rib vertical error of Xiangxi Yangtze River Bridge is transmissible, and the risk level of arch rib closure error reaches Grade IV when it is 75 similar to 125mm. Therefore, the prestress of anchor cable, tower buckle offset and local temperature should be closely monitored during construction. The calculated value of the joint prediction model of construction risk is consistent with the actual closure control error trend of the project, which can provide reference for the prediction and regulation of related bridge construction risks.
In response to the increasing demand for urban low-altitude Unmanned Aerial Vehicle (UAV) operation, this paper comprehensively considers the constraints of urban area traffic demand, airspace environment and UAV performance, with the goal of minimizing network nodes and maximizing coverage demand points, this paper proposes an urban low-altitude UAV route network construction method. First, the network node generation model and the route network construction model are established; secondly, the improved greedy algorithm is used to generate network nodes; then, k-Nearest Neighbor(K-NN) and Prim's algorithm are combined to construct the UAV route network. Finally, simulation verification is carried out based on the demand data of logistics business in Shanghai and the low-altitude operating environment, and the performance of the constructed route network is analyzed by using the complex network-related indicators. The results show that the urban low-altitude UAV route network construction method proposed in this paper can realize the construction of trunk and branch route networks at the same time, and the trunk route network realizes full network connectivity. In addition, the network nodes can be set according to the UAV mission requirements to achieve flexible planning of the route network.
《武汉理工大学学报(交通科学与工程版)》,(ISSN2095-3844,CN42-1824/U),刊期为双月刊,创刊于1959年.是由武汉理工大学主办的自然科学综合性学术期刊. 办刊宗旨:《武汉理工大学学报(交通科学与工程版)》是交通运输领域高水平的学术理论性刊物,是我国交通科学与工程领域的学科前沿阵地和学术交流的平台.学报坚持理论联系实际,推崇科技创新,注重科学引领,用优质的论文彰显我国交通领域最新科学发展与研究成果,促进交通运输科技成果转化,为交通运输工程及相关学科建设服务,为发现和培养高水平交通科技人才服务,为促进我国交通科学与工程学术研究与国际交流服务.
文中基于混合交通组队策略,针对合流区车辆合流特性,构建混合交通流环境下的合流区多车协同决策机制.搭建混合交通合流仿真环境,验证多车协同方法有效性.结果表明:与匝道自由合流控制相比,文中提出的方法在当网联车辆渗透率高于40％～60％时,车辆平均行驶时间的优化率增幅加快,随着渗透率增高,效果最高改善了 22.42％,且总平均延误、总冲突数分别减少了14.5％、21.1％,匝道延误平均减少了 36.3％.
文中通过在深度卷积神经网络的特征提取模块中引入注意力机制,实现不同通道之间的信息交互,缓解船名位置分布多样导致漏检的问题.通过特征金字塔增强模块融合低高级信息来加深不同尺度的特征,基于特征融合模块将不同深度的特征金字塔增强模块产生的特征聚集成最终的特征进行分割,在提高检测精度的同时降低模型复杂度.利用 smooth L1 损失函数解决船名预设框的位置回归问题.水面船舶船名检测结果表明:改进后的算法在自建船名数据集上检测精确率达到 88.1%,相较于现有优势算法DBNet提高了 3.4%.
