神经网络的前向生成

前向生成神经网络是一种解决映射问题的神经网络。本文介绍其基本思想、性能和训练方法。该网络的主要优点在于其隐节点数目不很大时训练的快速和输出的精确     

辊型在线超声检测可行性研究􀀁

随着工业技术的发展，人们对所需要钢的质量和性能要求越来越高，作为轧制在生产线上的一个重要工具－轧辊，其性能对轧制钢的影响是不容忽视的，轧辊在轧制过程中始终处于高温、高负荷和高速的状态下，因此随着轧制的进行，轧辊表面逐渐磨损，变得的粗糙，此时就会引起板厚精度，板形及带钢等的恶化，因此，必须对轧辊进行监测和磨新     

田湾核电站循环水泵泵轴腐蚀问题及处理措施

江苏田湾核电站1、2号机组的循环冷却水泵在调试运行阶段出现了比较严重的腐蚀现象,其中泵轴尤为突出。根据泵轴腐蚀的不同特点,将其分为4个不同区域,从泵体结构、选材和工作环境等方面对这4个区域的腐蚀特征进行分析,找出了腐蚀原因,采取了更换密封盘根材料、加装淡水冲洗装置、在泵轴上安装牺牲阳极机构和泵轴腐蚀点的修补处理措施,恢复了泵轴的正常使用,同时尽可能改善其运行工况。运行表明,由于采取的措施行之有效,基本上消除了泵轴的腐蚀。     

Simulated annealing heuristics for the dynamic facility layout problem

In today's economy, manufacturing plants must be able to operate efficiently and respond quickly to changes in product mix and demand. Therefore, this paper considers the problem of arranging and rearranging (when there are changes between the flows of materials between departments) manufacturing facilities such that the sum of the material handling and rearrangement costs is minimized. This problem is known as the dynamic facility layout problem (DFLP). In this paper, two simulated annealing (SA) heuristics are developed for the DFLP. The first SA heuristic (SA I) is a direct adaptation of SA to the DFLP. The second SA heuristic (SA II) is the same as SA I with a look-ahead/look-back strategy added. To test the performance of the heuristics, a data set taken from the literature is used in the analysis. The results obtained show that the proposed heuristics are very effective for the dynamic facility layout problem.     

Refinement of packet size in low carbon bainitic steel by special thermo-mechanical control process

The packet size of bainitic steel can be refined by a special relaxation-precipitation-control phase transformation (RPC) technology, When processed by RPC process, the low carbon bainitic steel composes of two kinds of main intermediate transformation phases. One is ultra-fine lath-like bainitic ferrite and the lath is less than 1 tam in width and about 6 tam in length; the alignment of laths forms a refined packet, and the size of packets is about 5-7 ~am in length and about 3-4 tam in width, The other is acicular structure. The morphology and distribution of these acicular structures are influenced by relaxation process, the thin and short acicular structures cut the prior austenite grain and refine the bainitic packet size, For the optimum relaxation time, the packet size can be refined to the finest. The mechanical properties are influenced by relaxation time and the 800 MPa grade low carbon bainitic steel with excellent toughness can be obtained by RPC process,     

基于神经网络和遗传算法的VBR通信流量预测

论文阐述了利用神经网络预测由连续自动回归(AR)马尔可夫模型所代表的可变位速率通信流量(VBR);在这一理论的基础上,介绍一个BP神经网络模型,它采用拆分,组装方法来构造一个学习结果达到均方根误差全局最小点的BP神经网络,该方法结合遗传算法能有效克服局部极小点,缩短学习时间和减小学习难度,并能够精确地预测VBR通信流量。     

用于汽车尾气监测的新型气体浓度传感器系统

针对汽车尾气新的排放标准和检定方法的实施,设计了用于汽车尾气监测的热电堆传感器系统。采用导电不定型碳多层镀膜技术的电调制红外光源,使其调制频率最高可以达到100 Hz.对传统气室结构进行了优化设计,使气室与外支撑分离,降低了外界环境的影响。实现了当红外光照射到吸收窗时,在输出引脚上产生与入射光功率成正比的输出信号电压,-3 dB衰减频率大约为10 Hz.实验结果表明：该气体浓度传感器系统满足汽车尾气中CO、CO2、HC监测的要求。     

Nonlinear cumulative damage model for multiaxial fatigue

On the basis of the continuum fatigue damage theory, a nonlinear uniaxial fatigue cumulative damage model is first proposed. In order to describe multiaxial fatigue damage characteristics, a nonlinear multiaxial fatigue cumulative damage model is developed based on the critical plane approach. The proposed model can consider the multiaxial fatigue limit, mean hydrostatic pressure and the unseparated characteristic for the damage variables and loading parameters. The recurrence formula of fatigue damage model was derived under multilevel loading, which is used to predict multiaxial fatigue life. The results showed that the proposed nonlinear multiaxial fatigue cumulative damage model is better than Miner’s rule.     

基于磁阻原理非接触式位移测量的非线性方法

针对低功耗和非接触式位移测量的应用需求，利用磁阻传感器，本文在传统的线性测量方法的基础上，提出了一种非线性的测量方法。设计了一个测量系统，实测结果表明该方法测量精度高、测量范围广、简单易用，也节省了成本。     

广西汽配行业ASP服务平台拉动单管理系统的研究

针对采用JIT生产方式的主机厂要求配套厂送货及时，以及配套厂拉动单管理不科学的现状，提出了拉动单管理系统的模型结构，对功能模块进行划分，并设计实现了拉动单管理系统，较好地解决了主机厂与配套厂数据对接问题。