洪水预报中特征值预报的若干数学方法比较

讨论研究了水文特征值预报的数学方法,统计回归模型、神经网络模型和模糊回归模型。三个计处实例表明如果系统的线性关系较好,统计回归模型的结果最好;如果系统的线民生关系差,神经网络模型的结果最好;如果用于率定模型的资料太短,任何一个模型都不可靠。     

Artificial ground motion compatible with specified peak velocity and target spectrum

In this paper, a method, which synthesizes the artificial ground motion compatible with the specified peak velocity as well as the target acceleration response spectrum, was proposed. In this method, firstly, an initial acceleration time history α8^（0） （t）, which satisfies the prescribed peak ground acceleration, the target spectral acceleration ST（ω, ζ）,and the specified intensity envelope, is generated by the traditional method that generates the requency domain; secondly,α8^（0） （t）is further modulated by superimposing narrow-band time histories upon it in the time domain to make its peak velocity, approach the target peak ground velocity, and at the same time to improve its fitting precision to the target spectrum. Numerical examples show that this algorithm boasts high calculation precisions.     

Evolutionary properties of stochastic models of earthquake accelerograms: Their dependence on magnitude and distance

An approach to generate artificial earthquakeaccelerograms on hard soil sites is presented. Eachtime-history of accelerations is considered as arealization of a non-stationary gaussian stochasticprocess, with statistical parameters depending onmagnitude and source-to-site distance. In order tolink the values of these parameters for each groundmotion record with the corresponding magnitude andsource-to-site distance, semi-empirical functionalrelations called generalized attenuationfunctions are determined. The set of realground-motion time histories used to obtain thesefunctions correspond to shocks generated at differentsources and recorded at different sites in thevicinity of the southern coast of Mexico. The resultsshow significant dispersion in the parameters of themodel adopted, which reflect that associated with thereal earthquakes included in the sample employed.The problem of conditional simulation of artificialacceleration time histories for prescribed intensitiesis briefly presented, but its detailed study is leftfor a companion paper. The criteria and modelsproposed are applied to generate two families ofartificial acceleration records for recurrenceintervals of 100 and 200 years at a specific sitelocated in the region under study. The results shownin this article correspond to acceleration timehistories recorded on firm ground for earthquakesgenerated at the subduction zone that runs along thesouthern coast of Mexico, and cannot be generalized tocases of earthquakes generated at other sources orrecorded at other types of local conditions. Thismeans that the methods and functional forms presentedhere are applicable to these other cases, but thevalues of the parameters that characterize thosefunctions may differ from those presented here.     

基于小波分析和神经网络的框架结构损伤诊断方法

本文提出了基于“能量—损伤”原理，综合利用小波包分析和神经网络的框架结构损伤诊断方法。在结构损伤诊断中，时域信号经小波分析后其缺损特征会更加明显，把分布在不同频带的结点能量作为神经网络的训练样本能够很好地反映结构缺损特征。以美国土木工程师学会提出的基准结构为例，阐述了结构损伤的发生、位置和程度诊断过程，取得了令人满意的结果。     

Simulation of artificial random field considering the temporal-spatial variation

Introduction For the seismic design of special structures such as nuclear power station, marine platform, long-span bridge and dam, generally the time-history response analysis of the structure under seismic excitation is imperative, which was coded in most seismic design codes. The earthquake records suitable for the seismic situation and site condition are necessary to be used as the seismic input in the dynamic analysis of structures. As a result of the limited observational condition of st…     

Application of genetic BP network to discriminating earthquakes and explosions

Introduction Artificial Neural Network (ANN) is an important branch of artificial intelligence. It is proposed on the foundation of the study on modern neural science, is a man-made network that can implement some functions based on the mans comprehensive understanding for cerebral neural network (HAN, WANG, 1997). ANN is a mathematical model of simplified human brain neural network and is used to simulate the structures and functions of human brain neural network. ANN is a complex netw…     

Time domain method for calculating free field motion of a layered half-space subjected to obliquely incident body waves

In numerical simulation of wave scattering under oblique incident body waves using the finite element method, the free field motion at the incident lateral boundary induced by the background layered half-space complicates the computational area. In order to replace the complex frequency domain method, a time-domain method to calculate the free field motion of a layered half-space subjected to oblique incident body waves is developed in this paper. The new method decouples the equations of motion used in the finite element method and offers an interpolation formula of the free field motion. This formula is based on the fact that the apparent horizontal velocity of the free field motion is constant and can be calculated exactly. Both the theoretical analysis and numerical results demonstrate that the proposed method offers a high degree of accuracy.     

沙坡头人工植被固沙区天然降水的入渗和分配研究

土壤、植被冠层与大气界面间(SVAT)物质传输过程日益成为水文学研究最感兴趣的领域,降水量的迁移与转换是非灌溉区SVAT主要的物质传输过程.干旱半干旱区稀疏灌丛蒸散量占降水量的90%以上,因此对降水入渗与水分在土壤内植物根际区再分配的研究显得十分重要.试验于2001年8月17日至9月30日在中国科学院沙坡头试验站进行,主要观测人工植被荒漠灌木柠条(Caragana korshinskii) 灌丛固定沙丘降水入渗与再分配过程.结果表明:在7次不连续降水过程中,土壤入渗深度与降水强度呈简单线性相关关系,土壤入渗速率约为降水强度的10倍.当次降水过程中降水强度小于0.46 mmh-1时,土壤入渗速率约为0 cm*h-1,此时的降水对沙区土壤基本上没有水分补给作用.受荒漠灌木柠条根系吸水作用的影响,其根系密集剖面深度40～140 cm内降水水分入渗积累不明显.降水入渗速率及入渗深度受土壤剖面初始含水率多寡而变化,干燥土壤剖面有助于提高瞬时入渗速率.降水以后随着时间的推移,人工固沙区微环境内空气温度、湿度等气象条件适宜,柠条生长进入相对旺盛阶段,其根系密集层140 cm深度处土壤含水率在总体上下降的过程中,表现出昼消夜长的趋势,试验期间翌日 8:00时土壤含水率值略高于前一日20:00时水分值 0.1%～0.3%.     

BP模型双样本训练方法

根据气象预报应用情况，分析ＢＰ模型训练人工神经元网络中存在的问题，并提出双样本训练方案。试验表明，采用双样本训练的网络，实际预报能力较高。     

基于Householder矩阵和Butterworth滤波器组的反馈延迟网络人工混响

针对现有人工混响方法中反馈延迟网络方法存在延迟线的个数选择受限和无法设定子带混响时间的不足,本文设计了一种基于Householder反馈矩阵和Butterworth滤波器组的反馈延迟网络人工混响器.Householder反馈矩阵有助于最大化回声密度,延迟线个数的选择相对自由,Butterworth滤波器组便于设定子带混响时间.仿真实验中分别利用语谱图、语音质量感知评价和主观评价验证了本文方法的有效性.