2．5ka BP以来额济纳绿洲沙丘的粒度特征及其反映的沉积过程

IEECAS SKLLQG

BP算法在辽宁观音阁水库枯季地表入库径流实时预报中的应用

根据太子河流域降水和产汇流特点 ,利用流域试点区观音阁水库枯季入库径流量、上游汛期降水量、汛末入库径流量和枯季降水量等实测系列资料 ,对所建立的BP神经网络模型进行训练、检验和对比分析 ,结果表明 ,所建立的枯季径流BP预报模型是合理可行的 ,并且具备较高的精度 ,可在同类地区推广使用

降水量的BP人工神经网络预测模型及其应用

由于影响因素的复杂性 ,预测降水量具有相当的难度。在假设区域长时间内降水量和蒸发量保持平衡的基础上 ,用 BP人工神经网络建立了陕西省汉中市的降水量预测模型 ,根据前 3个月降水量和蒸发量对降水量资料进行了模拟预测 ,结果认为其准确率为 84% ,合格率为 10 0 %。

Cold event at 5 500 a BP recorded in mud sediments on the inner shelf of the East China Sea

A 700-year record (1.0-1.5 a resolution) of the East Asian winter monsoon (EAWM), based on grain-size analysis and AMS(14)C dating of Core EC2005 from the inner-shelf mud wedge of the East China Sea (ECS), was compared with the Dongge stalagmite delta O-18 record during the mid-Holocene. The upper muddy section of Core EC2005 has been formed mainly by suspended sediments derived from the Changjiang (Yangtze) River mouth since 7.3 ka BP. High precipitation and a strengthened EAWM might have played key roles in the high sedimentation rate (1 324-1 986 cm/ka) between 5.9-5.2 ka BP. The EAWM strengthened when the Asian summer monsoon weakened, especially around 5 500 a BP, which corresponded to a worldwide cold event. The EAWM during the mid-Holocene shows statistically significant solar periodicities at 62 and 11 a. The 5 500 a BP cold event might be resulted from orbital forcing and changes in solar activity.

Carbonate cycle and its control factors in Kuroshio source region during the last 190ka BP

Two deep sea cores (Ph05-5, 16.05 degrees N, 124.34 degrees E, water depth 3382m and WP3: 22.15 degrees N, 122.95 degrees E, water depth 2700m) retrieved from the Kuroshio source region of the western Philippine Sea were selected to carry out the CaCO3 and calcareous nannofossil faunas study. Based on AMS(14)C data and comparing tire oxygen isotope curve with SPECMAP delta O-18 (Martinson et al., 1987) a stratigraphy was established. And, combining the changes of primary productivity and dissolution index of carbonate, the carbonate cycle and its control factors were analyzed in this region during the last 190ka BP. The carbonate contents showed higher values in the glacial periods and lower values during the interglacial and Holocene periods, which characteristics was similar to the tendency of "Pacific Type" carbonate cycle. However, there were high carbonate contents in the warm period and low values during the cold interval, which displayed the same tendency with the "Atlantic Type" carbonate cycle during the last glacial period (MIS4-2) in the east of Phillipines. The variations of primary productivity and carbonate dissolution index indicated that the carbonate dissolution was a major factor controlling the carbonate content in tire cast of Philippines, and the variations in carbonate contents were mainly affected by the productivity of calcareous organism in the Southeast of Taiwan. The "Atlantic Type" carbonate cycle in the cast of Phillipines during the last glacial period (MIS4-2) was an effect of the process of dissolution combined with the change of primary productivity.

利用BP神经网络实现三维物体姿态的测定

本文利用BP(Back-Propagation)人工神经网络对三维物体的姿态测定进行了研究。姿态测定一直缺少通用而实际的方法,人工神经网络由于具有强大的自组织、自适应学习能力,迅速的并行信息处理能力,可望解决这个问题。但现有BP算法存在训练慢和易陷入局部最小两个问题.本文提出的级联形式网络结构,使BP网络的训练速度大为提高,陷入局部最小的可能性大为降低。利用这种级联结构对飞机模型姿态测定,取得了较好的实验结果。

基于粗糙集-BP神经网络的发动机故障诊断

由于发动机光谱分析监控数据中磨损微粒种类过多,如果将这些微粒信息直接作为神经网络的输入,则存在输入层神经元过多、网络结构复杂等诸多问题。本文将粗糙集引入到发动机故障诊断中来,利用粗糙集在属性约简方面的优势,删除冗余磨损微粒,提取出重要磨损微粒,并将其作为BP神经网络的输入,建立发动机故障诊断模型。该方法降低输入层的神经元个数,简化了网络结构,缩短网络训练时间,并且由于剔除了冗余磨损微粒,减少了由该部分微粒信息不准确而带来的误差,有效提高了故障诊断的精确度。最后通过算例分析验证了相关算法和诊断模型的准确性和有效性。

基于改进的模糊C均值的BP分类器设计

提出了一种基于改进模糊C均值的BP神经网络分类器的设计,通过改进的模糊C均值算法对大量的数据进行聚类划分,然后设计BP神经网络对划分后的数据进行训练和测试,最后由计算机进行综合判断.试验证明该分类器是有效的,可以对高速公路车辆的车型进行迅速判别.

神经网络在覆盖件模具表面激光硬化虚拟过程中的应用

基于人工神经网络技术对覆盖件模具表面激光硬化虚拟过程的仿真建模,结合几何因素分析了模型的主要影响参数,对BP网络的结构和训练进行了说明。预测了激光表面硬化的加工效果 (表面硬度、硬化层深、相对耐磨性和表面粗糙度),实现激光加工工艺参数的优化,为实际生产和加工提供了依据。并以C语言为开发语言,利于实现各平台间的集成。