太原大气能见度影响因子分析及能见度预报

利用2017年1月—2019年12月太原地区逐时气象资料,分析了能见度及其主要影响因子的变化特征,并对两次低能见度过程进行深入分析,构建了能见度预报模型并进行检验,结果表明:(1)从空间分布看,太原北部能见度明显高于南部地区。从时间分布看,太原地区平均能见度最大值出现在5月,最小值出现在1月;日间最低值出现在06:00(北京时,下同),冬季略向后推移,最高值出现在15:00前后。(2)2017—2019年太原地区低能见度分别出现93、84、79 d;低能见度发生时,干霾、湿霾发生频率分别为59.27%、40.73%;湿霾发生时,能见度降低更加明显。(3)所选个例中,能见度均随各影响因子有所起伏,干霾、湿霾过程中能见度分别与颗粒物浓度、相对湿度变化一致。(4)采用神经网络方法构建太原地区能见度预报模型,预报模型相关系数为0.81,均方根为4.43 km,平均绝对误差为17.39%,轻微级能见度的TS评分为87%。神经网络方法对太原地区能见度预报具有较高的参考价值。     

盆地基底岩性的综合地球物理预测方法——以松辽盆地滨北地区基底岩性预测为例

预测盆地基岩岩性不仅对于研究盆地的深部地质结构及盆地的形成演化具有重要的意义,而且也对基岩风化壳油气藏的勘探具有一定的指导作用.本文通过对盆地重、磁异常成因的综合分析,提出了一系列盆地基底岩性综合预测研究的综合地球物理资料处理解释方法技术.指出在地震构造界面的约束下采用重力剥皮技术可以较为可靠地获取基底岩性重力异常并分析了界面密度差对剥皮后基底岩性重力异常的影响,给出了等效密度差的求取方法.分析了基底起伏对基岩岩性磁异常的影响,指出采用"平化曲"将磁异常归化到与基底同一高度,可以有效地提高对基底岩性体的刻画能力.通过综合分析认为:应用基底的相对视密度、相对视磁化率及两者的相关系数可以有效地刻画基底岩性的特征.神经网络是基底岩性判别与分类的有效方法技术.通过对松辽盆地北部滨北地区的基底岩性的综合预测显示了本文系列预测基底岩性方法的有效性,预测结果反映了松辽盆地基底岩性的分布特征.该系列方法技术可为其他盆地的基底地质填图提供了可借鉴的综合预测方法技术.     

区域降水数值预报产品人工神经网络释用预报研究

利用T213、日本细网格降水预报等数值预报产品,采用人工神经网络方法进行预报释用。通过聚类分析方法对广西自治区测站进行分类,简化预报对象,对数量众多的T213数值预报产品采用自然正交分解(EOF)方法,浓缩大量因子的有效信息,并结合日本降水预报因子建立广西5~6月区域降水量级的逐日人工神经网络预报模型。运用与实际业务预报相同的方法进行逐日预报试验。结果表明,用这种数值预报产品释用方法建立广西3个预报区域的B-P人工神经网络预报模型对中雨以上降水量级预报的TS评分分别为0.55、0.5和0.26,比目前业务预报中参考使用的T213和日本数值预报产品降水预报具有更好的预报效果。     

电磁成像正演仿真与传感器参数优化设计

应用有限元方法对电磁层析成像测量进行正演模拟,计算不同测量组合下各种典型流型的电势实部和虚部分布,考察电极个数、电极张角以及电极厚度对模拟测量响应的影响,并采用神经网络方法对传感器各结构参数进行优化设计.研究结果表明,介质分布对电势影响起主要作用,其中虚部分布特征更加明显;传感器各参数中电极数目对测量值影响较大,电极张角和电极厚度有一定影响.     

走向21世纪的实用地震学

实用地震学包括地震勘探、深部地震测深、工程地基与质量勘查以及浅部地质灾害预测等,在国民经济各方面起着重要作用,特别是在我国油气资源勘探与开发中起着并将继续起着不可替代的作用．在即将迈向２１世纪的今天,我们探讨一下实用地震学未来发展的道路是很必要的．实用地震学在２１世纪仍将在地学研究和国家建设中发挥它应有的作用,而且对它的要求也将越来越高,因而它在理论和应用方面都有诸多疑难问题要研究．本文就其发展现状、未来的理论研究和新技术的应用问题做些探讨．     

一种多层神经网络的学习算法

提出一种基于序贯预测误差方法（ＳＰＥ）的多层神经网络（ＭＮＮ）的学习算法。经模拟计算，它比传统的基于最陡下降方法的误差反传（ＳＤＢＥＰ）算法具有更好的收敛性能。并对这两种算法进行了模拟计算的比较．     

Bed changes at toe of inclined seawalls

This study was carried out to investigate the scour phenomenon at the toe of seawalls and the different parameters that affected it. Experiments were achieved using different wave steepnesses, bed material grain sizes, wall positions and inclinations. Based on experimental results, the parametric plots of toe scour for smooth impermeable inclined seawalls were prepared. Also, this paper presents the bed changes prediction at seawalls toe using artificial neural networks on the basis of experimental data to widen the range of application. Suitability of using a neural network model was developed, and a model was validated. It is proposed that this model can be used in coastal engineering applications.     

Neural networks in the dynamic response analysis of slender marine structures

Risers and anchor lines play important roles in offshore oil exploitation activities nowadays. For this reason the proper analysis and design of such slender structures has been of a paramount interest. The principal characteristics to be accounted for in riser and mooring line analysis are the severe nonlinearities involved and the random dynamic effects associated. The Finite Element Method (FEM) is an essential step to cope with this kind of analysis. But the use of the FEM can be computationally very expensive for the solution of the resultant nonlinear differential equations of motion, because the time-domain integration should produce sufficiently long response time-histories using small time-steps in order to obtain reliable time-series statistics of any structural response parameter, e.g., top tension in an anchor line or stresses occurring at a critical section in a steel catenary riser (SCR). This paper presents a very efficient hybrid Artificial Neural Network (ANN)–Finite Element Method (FEM) procedure to perform a nonlinear mapping of the current and past system excitations (inputs) to produce subsequent system response (output) for the random dynamic analysis of mooring lines and risers. Firstly, a quite short FEM-based time-domain response simulation is generated. Then, an ANN is used to predict the remaining structural response time-history simulation. The hybrid ANN–FEM approach can be very efficient for predicting long response time-histories. It has been observed that a 3 h response time-history can accurately be obtained with approximately the computational cost of a 500 s one, i.e., 20 times faster than a complete simulation using finite element-based solution. Roughly, this can represent a reduction of about a dozen of hours of computer time for a single mooring line analysis and about two dozens of hours (or more) for a single SCR analysis, both belonging to a deep-water floating unit.     

Derivation of coastal wind and wave parameters from offshore measurements of TOPEX satellite using ANN

The paper discusses an artificial neural network (ANN) approach to project information on wind speed and waves collected by the TOPEX satellite at deeper locations to a specified coastal site. The observations of significant wave heights, average wave period and wind speed at a number of locations over a satellite track parallel to a coastline are used to estimate corresponding values of these three parameters at the coastal site of interest. A combined network involving an input and output of all the three parameters, viz., wave height, period and wind speed instead of separate networks for each one of these variables was found to be necessary in order to train the network with sufficient flexibility. It was also found that network training based on statistical homogeneity of data sets is essential to obtain accurate results. The problem of modeling wind speeds that are always associated with very high variations in their magnitudes was tackled in this study by imparting training in an innovated manner.     

Prediction of breaking waves with neural networks

The height of a wave at the time of its breaking, as well as the depth of water in which it breaks, are the two basic parameters that are required as input in design exercises involving wave breaking. Currently the designers obtain these values with the help of graphical procedures and empirical equations. An alternative to this in the form of a neural network is presented in this paper. The networks were trained by combining the existing deterministic relations with a random component. The trained network was validated with the help of fresh laboratory observations. The validation results confirmed usefulness of the neural network approach for this application. The predicted breaking height and water depth were more accurate than those obtained traditionally through empirical schemes. Introduction of a random component in network training was found to yield better forecasts in some validation cases.