Preparation and tribological behavior of polytetrafluoroethylene/metal mesh composites with sandwich structure

Polytetrafluoroethylene (PTFE) owns an excellent self-lubricating performance, but its wear rate is very high due to the large-scale spalling of the matrix in the friction. In this paper, A new kind of PTFE composites with sandwich structure was prepared by layer-press technology, whose middle layer is filled with metal mesh. The influence of the mesh structure and mesh density of middle metal layer on tribological properties of composites were researched in detail. The results revealled that the metal mesh located in the composites can efficiently prevent the large-scale spalling of PTFE, which induces the sample of PTFE/500# plain woven dutch metal mesh (PTFE-500#PWD) to have a lower wear rate (9 × 10⁻⁵ mm³/Nm) and COF (0.106) under the fixed experimental condition. The prepared PTFE/metal mesh composites reveal excellent anti-friction and anti-wear performance, which can be used to fabricate a new kind of self-lubricating materials.     

Effect of welding heat input on HAZ character in ultra-fine grain steel welding

In this essay, we studied how heat input affected the microstructure, hardness, grain size and heat-affected zone (HAZ) dimension of WCX355 ultra-fine grain steel which was welded respectively by the ultra narrow-gap welding (UNGW) process and the overlaying process with CO2 as protective atmosphere and laser welding process. The experimental results show when the heat input changed from 1.65 kJ/cm to 5.93 kJ/cm, the width of its HAZ ranged from O. 6 mm to 2. 1 ram. The average grain size grew up from 2 ~ 5 pln of base metal to 20～70 um and found no obvious soften phenomenon in overheatedzone. The width of normalized zone was generally wide as 2/3 as that of the whole HAZ, and the grain size in this zone is smaller than that in base metal. Under the circumstance of equal heat input, the HAZ width of UNGW is narrower than that of the laser welding.     

水基流延工艺制备氧化铝生带材料研究

利用水基流延法,选用纯丙乳液作为粘结剂,成功制备出表面光滑、结构均匀、柔韧性良好、强度较高的Al2O3生带材料,并发现当分散剂聚丙烯酸铵(PAA-NH4)含量为2.5%(质量分数)、粘结剂纯丙乳液用量为27%～28%(体积分数)、pH值为9.5～10时可制备出稳定性良好、流动性适宜的α-Al2O3的水基流延浆料.对纯丙乳液进行了DTA和TGA热分析,在此基础上确定流延素片的排胶温度在500℃;将该工艺制备的α-Al2O3流延素片在1650℃并保温2 h的烧结条件下,获得了强度和致密度都较高的烧结体,其中烧结较好的流延片的相对密度达到94.5%.     

非线性过程控制系统传感器故障检测新方法

针对电厂热力系统故障检测和定位准确性低的问题,提出了基于鲁棒输入训练网络的传感器故障检测模型;采用带参数限制项的目标函数对网络进行训练,并在测试目标函数中引入影响因子,增加了模型训练精度,抑制了网络计算过程故障数据对正常值的影响,减小了残差污染,提高了模型准确性;以某300 MW电厂热力系统20组测点为对象进行算例分析,通过反复的实验,结果表明,该模型能够更加准确地对非线性系统故障点进行检测和分离,并更加精确重构各变量真实值,验证了该模型用于非线性过程传感器故障检测的有效性和可靠性.     

亭子口水利枢纽左岸天然料砂石加工系统设计

亭子口水利枢纽选用天然砂砾石料场作为混凝土骨料的料源,其骨料加工系统的生产工艺流程较人工山场料加工系统工艺流程更为复杂。详细介绍了该电站左岸天然砂石料加工系统的设计要点,对系统的工艺流程做了深入的研究,选择了适合其生产能力的加工设备,并首次在水电工程中采用梯形磨磨制石粉新工艺。实践证明,该系统的设计是成功的,其生产能力和骨料加工质量满足工程需求。     

New insights on survivability in multi-domain optical networks

With the development of intelligent optical networks and the general multi-protocol label switching (GMPLS) technique, the seamless convergence between IP network and optical network is no longer be a dream but a practical reality. Similar to the Internet, current optical networks have been divided into multiple domains each of which has its own network provider and management policy. Therefore, the development of multi-domain optical networks will be the trend of new-generation intelligent optical networks, and GMPLS-based survivability for multi-domain optical networks will become a hot topic of research in the future. This paper provides a comprehensive review of the existing survivable schemes in multi-domain optical networks and analyzes the shortcomings of current research. Based on previous studies, we present possible challenges and propose new ideas to design efficient survivable schemes to guide the future work of researchers in multi-domain optical networks.     

基于贝叶斯压缩感知的子空间拟合离格DOA估计

针对传统的基于稀疏贝叶斯学习（Sparse Bayesian Learning,SBL）的波达方向估计算法对噪声鲁棒性不高的问题,提出了一种基于SBL的子空间拟合离格波达方向(Direction of Arrival,DOA)估计方法。首先对接收数据的协方差矩阵进行特征分解,获得信号的加权子空间,构造等价信号的稀疏表示模型并利用贝叶斯学习算法进行参数求解。同时对于网格划分带来的建模误差问题,采用了离格贝叶斯推导（Sparse Bayesian Inference,SBI）算法进行求解,利用期望最大化算法迭代更新相应的参数。仿真结果表明,相对于传统的DOA方法,该方法具有更好的估计精度。     

基于自组织增量-图卷积神经网络的金相图半监督学习

采用深度学习对钢铁材料显微组织图像分类,需要大量带标注信息的训练集.针对训练集人工标注效率低下问题,该文提出一种新的融合自组织增量神经网络和图卷积神经网络的半监督学习方法.首先,采用迁移学习获取图像数据样本的特征向量集合;其次,通过引入连接权重策略的自组织增量神经网络(WSOINN)对特征数据进行学习,获得其拓扑图结构,并引入胜利次数进行少量人工节点标注;然后,搭建图卷积网络(GCN)挖掘图中节点的潜在联系,利用Dropout手段提高网络的泛化能力,对剩余节点进行自动标注进而获得所有金相图的分类结果.针对从某国家重点实验室收集到的金相图数据,比较了在不同人工标注比例下的自动分类精度,结果表明:在图片标注量仅为传统模型12％时,新模型的分类准确度可达到91％.     

基于Sinkhorn距离特征缩放的多约束非负矩阵分解算法

为了减少原始特征对非负矩阵分解(NMF)算法的共适应性干扰,并提高NMF的子空间学习能力与聚类性能,该文提出一种基于Sinkhorn距离特征缩放的多约束半监督非负矩阵分解算法。首先该算法通过Sinkhorn距离对原始输入矩阵进行特征缩放,提高空间内同类数据特征之间的关联性,然后结合样本标签信息的双图流形结构与范数稀疏约束作为双正则项,使分解后的基矩阵具有稀疏特性和较强的空间表达能力,最后,通过KKT条件对所提算法目标函数的进行优化推导,得到有效的乘法更新规则。通过在多个图像数据集以及平移噪声数据上的聚类实验结果对比分析,该文所提算法具有较强的子空间学习能力,且对平移噪声有更强的鲁棒性。     

从测角后交最佳点位规律看规范规定之不足

本文分析了文献[1]所总结的测角后交两类公式在取得最佳点位时的规律,指出当前规范对此问题规定之不足.