神经网络敏感性分析及其在遥感影像分类中的应用

针对基于连接权的神经网络敏感性分析方法中求取敏感性系数的不稳定性,提出一种优化连接权的神经网络敏感性分析方法。首先采用遗传算法根据误差最小化原则对神经网络进行优化,在优化的神经网络模型上进行基于连接权的敏感性分析。以1个数值模拟实例和华盛顿广场地区的遥感图像地物分类为例,验证所提方法的有效性。实验结果表明,所提方法求取输入变量的敏感性系数是稳定有效的,能有效筛选出遥感图像中对分类贡献较大的特征波段,达到降维的同时提高分类精度。     

紫色土经济林栽培技术

为了合理开发利用紫色土 ,保护生态环境 ,为紫色土经济林栽培提供实践指导 ,对紫色土经济林栽培技术进行了研究 测定了全垦、带垦、水平梯土、水平梯土撩壕、鱼鳞坑等 5种整地方法的土壤以及对照地 (不整地 )的土壤侵蚀量 ,观测了各种整地方法对经济林木生长发育的影响 ,同时 ,对栽植在紫色土和板页岩红壤上的经济林木的生长发育情况进行了对比 结果表明 :不同整地方法土壤侵蚀量按从大到小的顺序排列为全垦 (81 4t·hm-2a-1,以下单位均为t·h-2 a-1) >带垦 (63 6) >水平梯土 (3 8 7) >水平梯土撩壕 (3 1 8) >鱼鳞坑 (2 4 3 ) >对照(2 1 1) ;不同整地方法对经济林木生长发育的影响按由好到差的顺序排列为水平梯土撩壕、水平梯土、全垦、带垦、鱼鳞坑、对照 ;在一般抚育管理条件下 ,在紫色土坡地生长发育较好的树种或品种有板栗、枣树、杨梅、枇杷、萘李、甜柿、水蜜桃 ,生长发育一般的树种有茶树、无核蜜柑、柚子 ,生长发育较差的树种有脐橙、柑、葡萄 因此 ,在坡度不大 (<2 5°)的紫色土坡地营造经济林时 ,整地方法采用水平梯土撩壕为宜 ,栽培树种可选择板栗、枣树、杨梅、枇杷、萘李、甜柿、水蜜桃 ,还可选择无核蜜柑、茶树、柚子等     

投影寻踪回归与BP神经网络方法在前汛期降水预测中的比较研究

以1962—2006年粤北地区7个站4—6月前汛期降水量资料为基础,将前汛期降水量与74项环流指数资料进行灰色关联度分析,确定了影响粤北地区前汛期降水量的16个关键环流指数因子,分别应用投影寻踪回归、BP神经网络和逐步回归方法,建立前汛期降水趋势预测模型,对粤北地区前汛期降水趋势进行预测。结果表明:投影寻踪回归和BP神经网络方法的预测能力均优于传统的逐步回归模型。其中,PPR模型比BP神经网络方法的预测效果更好。     

Role of a rheophyte in bench development on a sand‐bed river in southeast Australia

Casuarina cunninghamiana Miq. is an important rheophytic tree in New South Wales, Australia because it is fast growing and can tolerate flood disturbance. Widden Brook is an active sand‐bed stream that has widened substantially since initial European settlement in the early 1800s and is characterized by high flood variability and multi‐decadal periods of alternating high and low flood frequency, called flood‐ and drought‐dominated regimes. Channel contraction by bench formation is currently occurring. Conversion of coarse‐grained point bars to benches is an important process of channel contraction. When point bars grow to a height where suspended sediment is first deposited to thicknesses of at least 50 mm by sub‐bankfull floods, rapid establishment of C. cunninghamiana occurs. As the trees grow they partially block bankside flows, thereby locally reducing flow velocity and inducing further deposition on the benches. Such synergistic relationships between bar height and inundation, fine‐grained sediment deposition, tree establishment and the development of a bankside low current velocity zone are fundamental for bench development. Size‐class frequency data demonstrate that C. cunninghamiana on the benches consists of pure even‐aged stands with most trees clustering near the average diameter. Two benches have similar size class frequency distributions but a third has significantly smaller trees. Recruitment on benches is episodic, may occur in areas open to grazing and is dependent on favourable conditions that allow tree survival. These favourable conditions include high seed availability, low levels of competition, deposition of fine sediments and adequate moisture for tree growth. Although C. cunninghamiana germinates on bars, seedlings are eliminated by prolonged inundation or flood scour and do not reach maturity. Recurring catastrophic floods or a sequence of large floods in rapid succession episodically destroy benches by substantial channel widening and initiate a new phase of bar and bench development. A conceptual model of the conversion of point bars to benches by thick mud deposition and C. cunninghamiana recruitment has been developed for sand‐bed streams draining similar sandstone catchments. Copyright © 2009 John Wiley & Sons, Ltd.     

基于麻雀搜索算法优化的BP神经网络卫星钟差预报

使用麻雀搜索算法(sparrow search algorithm,SSA)对BP神经网络的初始权值和阈值进行优化和调整,以提高神经网络模型短期预报的精度和稳定性.采用IGS产品中的卫星钟差数据,对SSA-BP神经网络模型、PSO-BP神经网络模型、传统BP神经网络模型及传统二次多项式模型(QP模型)进行实验对比,结果...     

Reconstruction and analysis of sub-plinian tephra dispersal during the 1530 A.D. Soufrière (Guadeloupe) eruption: Implications for scenario definition and hazards assessment

The last magmatic eruption of Soufrière of Guadeloupe dated at 1530 A.D. (Soufrière eruption) is characterized by an onset with a partial flank-collapse and emplacement of a debris-avalanche that was followed by a sub-plinian VEI 2–3 explosive short-lived eruption (Phase-1) with a column that reached a height between 9 and 12 km producing about 3.9 × 10⁶ m³ DRE (16.3 × 10⁶ m³ bulk) of juvenile products. The column recurrently collapsed generating scoriaceous pyroclastic flows in radiating valleys up to a distance of 5–6 km with a maximum interpolated bulk deposit volume of 11.7 × 10⁶ m³ (5 × 10⁶ m³ DRE). We have used HAZMAP, a numerical simple first-order model of tephra dispersal [Macedonio, G., Costa, A., Longo, A., 2005. A computer model for volcanic ash fallout and assessment of subsequent hazard. Comput. Geosci. 31, 837–845] to reconstruct to a first approximation the potential dispersal of tephra and associated tephra mass loadings generated by the sub-plinian Phase 1 of the 1530 A.D. eruption. We have tested our model on a deterministic average dry season wind profile that best-fits the available data as well as on a set of randomly selected wind profiles over a 5 year interval that allows the elaboration of probabilistic maps for the exceedance of specific tephra mass load thresholds. Results show that in the hypothesis of a future 1530 A.D. scenario, populated areas to a distance of 3–4 km west–southwest of the vent could be subjected to a static load pressure between 2 and 10 kPa in case of wet tephra, susceptible to cause variable degrees of roof damage. Our results provide volcanological input parameters for scenario and event-tree definition, for assessing volcanic risks and evaluating their impact in case of a future sub-plinian eruption which could affect up to 70 000 people in southern Basse-Terre island and the region. They also provide a framework to aid decision-making concerning land management and development. A sub-plinian eruption is the most likely magmatic scenario in case of a future eruption of this volcano which has shown, since 1992, increasing signs of low-energy seismic, thermal, and acid degassing unrest without significant deformation.     

Dam overtopping risk assessment considering inspection program

Safety inspection of large dams in Taiwan is conducted every 5 years. The practice does not take into consideration uncertainty of dam conditions. The goal of this study is to determine the optimal dam inspection interval under the consideration of overtopping risk incorporating uncertainty gate availability. In earlier studies, assessment of overtopping risk only considered the uncertainties in reservoir properties and natural randomness of hydrologic events without giving much thought to the availability of spillway gates. As a result, the overtopping risk could be underestimated. In this study, an innovative concept is proposed to evaluate dam overtopping by taking into account spillway gate availability. The framework consists of three parts: (1) evaluation of conditional overtopping risk for different numbers of malfunctioning spillway gates; (2) evaluation of spillway gate availability; and (3) dam inspection scheduling. Furthermore, considerations are given to overtopping risk, inspection cost, and dam break cost for determining the optimal inspection schedule. The methodology is applied to the Shihmen Reservoir in Taiwan and to evaluate its time-dependent overtopping risk. Results show that overtopping risk considering the availability of the spillway gates is higher than the one without considering the availability of the spillway gates.     

基于SOFM神经网络的边坡稳定性评价

针对边坡工程稳定性分析中参数的不确定性,在分析自组织特征映射神经网络（SOFM）基本学习算法的基础上,从提高算法收敛速度和性能出发,将自组织特征映射神经网络基本学习算法加以改进,据此建立了评价边坡稳定状态的SOFM神经网络模型。然后用收集到的边坡稳定工程实例作为样本,对该模型进行训练和检验,并与BP神经网络判别结果对比。结果表明,SOFM神经网络性能良好、预测精度高,是边坡稳定性评价的一种有效方法。     

人工神经网络在爆破块度预测中的应用研究

利用人工神经网络模型对爆破块度进行预测,实验结果表明,该方法是完全可行的。通过对实验样本数据进行归一化处理后再对人工神经网络模型进行训练和预测,其预测精度会得到大大提高。