基于卷积神经网络的目标检测

为了完成视频图像中人体检测,将卷积神经网络和梯度算法结合.将YOLO网络模型结构以及算法函数进行了优化调整.为了保留视频时间信息,引入了卷积神经网络反馈时间维度.采用长短记忆卷积神经网络(LSTM)进行训练,通过将迭代过程的最优解保留,加快了梯度算法中的搜索速度,并且保证收敛于全局最优解.     

一起发电机转子绕组动态匝间短路故障的诊断与分析

转子绕组是汽轮发电机的重要组成部分,运行状态下受机、电、热等多种因素影响,故障判断相对复杂.针对一起660 MW汽轮发电机运行中转子绕组动态匝间短路引发的振动异常,从不同角度及状态进行分析,确定了机组振动与转子绕组动态匝间短路的相关性,并对转子绕组动态匝间短路的诊断方法提出了建议.     

THERMAL DEMAGNETIZATION AND RANDOMNESS OF DOMAIN NUCLEATION OF SINGLE PHASE Nd_2 Fe_(14) B

Based on the observation of temperature variation of both domain structure and magnetic con-trast.the thermal demagnetization and randomness of domain nucleation was discussed.     

铝合金凝固过程的声发射及氢气析出过程

采用声发射测试技术,对Al-Si和Al-Cu系合金凝固过程进行监测,证实合金在凝固过程中存在声发射现象,表现为两个声发射信号计数率峰,两个峰的总计数都随着合金中气体含量增加而增高。本文对Al-Si,Al-Cu合金系产生声发射现象的机理进行了探讨,同时,借助其它实验手段证实了声发射信号与合金凝固中气体析出有很好的对应性。第一个计数率峰与气泡在枝晶臂间形核相吻合。第二个计数率峰与凝固末期的枝晶间拉裂现象相对应,可能是氢在枝晶间析出和聚集产生高压氢导致的拉裂。     

新型手机用PCB板焊点的失效分析

采用多种宏微观分析手段,对新型手机用PCB板的失效焊点(BGA)与散落焊珠的组分进行了系统的表征分析.结果表明,表面贴装技术(SMT)回流焊工艺参数控制不当是主要的失效起因;焊膏质量不佳及回流焊工艺与其不匹配是导致焊点焊块失效的另一重要原因.针对上述问题,就回流焊工艺参数的调整以及焊膏质量的改善提出了具体建议,以确保该PCB板的结构完整性和安全可靠性.     

��������еĶഫ������Ϣ�ںϼ���

 多传感器信息融合是一个新的研究领域，它应用到故障诊断领域还是近几年的事。文章主要论述了多传感器信息融合技术的基本概念、工作原理及常用方法，同时还介绍了其在故障诊断中的应用实例。     

Magnetic Flux Controllers for Induction Heating Applications

Application of magnetic flux controllers/concentrators to induction heating coils can drastically improve the process efficiency and heat pattern control. Presentation includes: benefits provided by flux controllers, materials available for controllers, application techniques, computer assisted design of induction coils with concentrators, examples of applications. Depending on induction system design, magnetic flux controllers can concentrate heating in a specified area,change heat source distribution and shield a particular part zone or external area preventing unintended eddy current heating.Besides of the coil efficiency improvement and optimal power distribution, magnetic flux controllers reduce the coil current demand from a supplying circuitry thus strongly reducing losses in busswork, cables, transformers and inverter components.Improvement that can be achieved due to magnetic flux controllers is case dependable. 2D and 3D computer simulation allows the designer to predict accurately effect of controllers on the coil parameters and temperature distribution and optimize the whole electromagnetic system. Special attention in presentation is paid to new magnetodielectric materials optimized for induction heating conditions. These materials have high magnetic permeability and saturation flux density,excellent machinability, good chemical and temperature resistance. Concentrators from these materials can work in a wide range of frequencies and specific powers. Examples of magnetic flux controller application include surface hardening of shafts and gears, induction surface hardfacing and brazing.     

无油迷宫式压缩机在含尘气体中的应用与改进

针对迷宫式压缩机在含尘气体中使用所出现的使用周期短等问题 ,在压缩机进口增设水洗塔、丝网除沫器 ,并对压缩机气阀、气缸盖等相关部件进行改进 ,取得了较好的效果     

基于电容层析成像和模糊模式识别的油气两相流流型辨识新方法的研究

基于电容层析成像和模糊模式识别技术别提出了一种油气两相流流型辨识的新方法。建立了12电极电容层析成像流型自动识别系统，该系统利用Tikhonov正则化原理并结合SIRT(Simultaneous Reconstruction Techniques)算法进行图像重建。Tikhonov正则化原理用于克服图像重建过程中的不适定问题，SIRT算法用于提高最终重建图像的质量。根据流型的随机和模糊特性，提出了一种根据管截面重建图像进行流型辨识的模糊流型判别方法。研究结果表明，提出的流型辨识新方法是有效的。对于层状流、核心流、环状流、均相流等流型，流型辨识的准确率高于95％，辨识一个流型所用的时间小于0.3秒。对于塞状流，流型辨识的准确率高于90％。