小波变换和BP神经网络模型在沉降变形监测中的应用研究

变形预测在预报工程险情方面起着关键性的作用,针对施工中需及时、准确地预测变形的问题,本文利用小波变换原理对监测数据进行降噪处理,并采用BP神经网络分析不同训练样本下的预测效果和精度水平。实验结果表明:基于小波消噪后的BP网络模型,以连续的近期观测数据作为训练样本,对下期变形预测精度高,效果好,相对误差很小。因此,小波变换和BP神经网络模型在沉降变形监测工程中能作为预测研究与应用的参考。     

基于改进BP神经网络的海底底质分类

通过采用遗传算法优化神经网络初始权值的方法,将GA算法与BP神经网络有机结合,应用于海底底质分类。基于多波束测深系统获取的反向散射强度数据,应用改进的BP神经网络分类方法,实现对海底基岩、砾石、砂、细砂和泥等底质类型的快速、准确识别。通过实验比较,GA-BP神经网络分类精度明显高于BP神经网络,证明了该方法的有效性和可靠性。     

基于贝叶斯正则化BP神经网络的DEM趋势面逼近

趋势面从宏观上揭示了研究对象的特性,在各领域发挥着重要作用。BP神经网络可以对复杂系统进行无限逼近,进而进行预测。建立了基于贝叶斯正则化BP神经网络的数字高程模型趋势面,与二次多项式建立的数字高程模型趋势面进行比较分析,证明了该方法的可行性和有效性。     

GA-BP网络模型在GPS似大地水准面精化中的应用

针对误差反向传播（BP）算法训练速度慢和易于陷入局部最小值的缺点,提出了利用遗传算法（GA）的全局寻优性,结合GA和BP的各自优点,分析和建立了进化神经网络（GA-BP）模型,并将该模型应用于似大地水准面模型精化。最后以南方某市E级GPS控制网高程数据为例,进行BP和GA-BP模型的对比实验,通过对内、外符合精度及MAPE（平均绝对误差百分比）指标分析,验证了该方法的可行性。     

Biogeochemical Variation in Dimethylsulfide, Phytoplankton Pigments and Heterotrophic Bacterial Production in the Subarctic North Pacific during Summer

Dimethylsulfide (DMS), chlorophyll a (Chl-a), accessory pigments (fucoxanthin, peridinin and 19-hexanoyloxyfucoxanthin), and bacterial production (BP) were measured in the surface layer (0–100 m) of the subarctic North Pacific, including the Bering Sea, during summer (14 July–5 September, 1997). In surface sewater, the concentrations of DMS and Chl-a varied widely from 1.3 to 13.2 nM (5.1 ± 3.0 nM, mean ± S.D., n = 48) and from 0.1 to 2.4 µg L^–1 (0.6 ± 0.6 µg L^–1, n = 24), respectively. In the subarctic North Pacific, DMS to Chl-a ratios (DMS/Chl-a) were higher on the eastern side than the western side (p < 0.0001). Below the euphotic zone, DMS/Chl-a ratios were law and the correlation between DMS and Chl-a was relatively strong (r ² = 0.700, n = 27, p < 0.0001). In the euphotic zone, DMS/Chl-a ratios were higher and the correlation between DMS and Chl-a was weak (r ² = 0.128, n = 50, p = 0.01). The wide variation in DMS/Chl-a ratios would be at least partially explained by the geographic variation in the taxonomic composition of phytoplankton, because of the negative correlation between DMS/Chl-a and fucoxanthin-to-Chl-a ratios (Fuc/Chl-a) (r ² = 0.476, n = 26, p = 0.0001). Furthermore, there was a positive correlation between DMS and BP (r ² = 0.380, n = 19, p = 0.005). This suggests that BP did not represent DMS and dimethylsulfoniopropionate (DMSP) removal by bacterial consumption but rather DMSP degradation to DMS by bacterial enzyme.     

山东沂沭河流域2000BC前后古文化兴衰的环境考古

Neolithic culture series in the Yishu River Basin developed in the order of Beixin culture-Dawenkou culture-Longshan culture-Yueshi culture. During the early and middle stage of Longshan culture （4600-4300 cal. yr BP）, the climate in the Yishu River Basin was warm and wet. Paddy-oriented agriculture planted paddy was very developed. The society was flourishing with great amount of archaeological sites. The cooling starting in 4200 cal. yr BP made the paddies shortfall in output or even no seeds were gathered. This situation intensified the discrepancy between population and resource. The scarcity in natural resource led to substantial decrease in population and subsequent drop in archaeological sites. About 4000 cal. yr BP Longshan culture was displaced by Yueshi culture which was relatively un- derdeveloped, simple and unsophisticated.     

古尔班通古特沙漠灌木冠幅预测模型

古尔班通古特沙漠是中国第二大沙漠,也是中国固定和半固定沙丘主要分布区,固沙灌木种较多。冠幅不仅是反映固沙灌木可视化的重要参数,也是反映沙漠植被生长情况的重要变量。以3种沙丘（固定沙丘、半固定沙丘和流动沙丘）上主要固沙灌木为研究对象,利用12种基础模型、BP(Backpropagation Neural Network)神经网络和支持向量机（Support Vector Machine,SVM）机器学习算法建立了基于固沙灌木株高和冠长率的冠幅预测模型,同时将两种机器学习算法拟合结果与基础模型进行比较,最终选出了适合研究区的冠幅预测模型。结果表明：（1）不同沙丘类型和不同灌木种类的最优冠幅预测模型不同,且固定沙丘和半固定沙丘模型优于流动沙丘。3种沙丘类型最优拟合为M2(Quadratic Model)模型;（2）白梭梭（Haloxylon persicum）在半固定沙丘和流动沙丘上拟合的最优模型分别为M2、M7(Gompertz),沙拐枣（Calligonum mongolicum）最优模型为M2,蛇麻黄（Ephedra distachya）和油蒿（Artemisia ordosica）在...     

BP人工神经网络在青藏铁路南段地壳稳定性定量评价中的应用

将BP人工神经网络方法引入区域构造活动性、区域地壳稳定性研究领域,对青藏铁路南段沿线的构造活动性进行定量分析.选用断层运动速率、地震震级、温泉温度及剪切应变4个关键影响因子作为BP人工神经网络的输入向量,构造活动强度(α)作为输出向量,以α为定量判据,将全区划分为相对稳定区(α＜0.22)、较不稳定区(α≈0.22～0.38)、不稳定区(α≈0.38～0.69)、极不稳定区或强烈构造活动区(α≥0.69).在青藏铁路南段沿线划分出格仁错、崩错、当雄-羊八井、错那湖、唐古拉山口南、聂荣东北、聂荣西北、雅鲁藏布江断裂沿线、萨迦等不稳定区,在不稳定区内部进一步划分出申扎、蓬错、尼木、桑雄、羊八井5个极不稳定区.     

神经网络方法在地震预报研究中的初步应用

本文把神经网络方法引进地震预报研究当中。使用地震频次，最大震级，平均震级，等价地震次数等多项地震活动性指标作为神经网络的输入，未来时段内的最大地震震级作为其输出，可以对某一固定地区的最大地震震级作出中近期预报。选用的神经网络模型为含两个中间层的前向模型，并采用ＢＰ算法。所得结果表明，用神经网络方法可以在一定精度范围内使震级预报的内检符合率达到１００％，在本文的例子中，外推预报准确率达到６０％以上。