基于退火BP神经网络的GPS高程转换

阐述模拟退火算法的基本思想和原理,提出并介绍模拟退火算法优化的BP神经网络模型在GPS高程转换中的具体应用,同时编写相应的MATLAB处理程序,结合大量数据进行仿真实验,结果表明文中提出的退火BP神经网络模型具有收敛速度快、精度高、避免陷入局部最小的优良特性。     

ARTIFICIAL NEURAL NETWORK CORRECTION OF ROUGH ERRORS OF OBSERVATIONS*

In the context of tower measured radiation datasets.following the correction principle meeting a diagnostic equation in data quality control and in terms of a technique for model construction on data and ANN (artificial neural network) retrieval for BP correction of radiation measurements with rough errors available,a BP model is presented.Evidence suggests that the developed model works well and is superior to a convenient multivariate linear regression model,indicating its wide applications.     

Sedimentary cycles of trace elements in Salawusu River Valley since 150 ka BP

The paper makes some analyses on 11 trace elements in the Milanggouwan stratigraphical section in the Salawusu River valley, which is regarded as a prototype geology-palaeoclimate record since 150 ka BP. The results show that the content and variation of trace elements has experienced remarkably regular changes in the pace with coarse and fine sedimentary cycles of palaeo-aeolian sands to its overlying fluvio-lacustrine facies or/and palaeosols. The trace elements with chemical properties of relatively active (V, Sr, Cu, Ni, As) and relatively stable (P, Pb, Rb, Mn, Nb, Zr) are a manifestation of the corresponding 27 changeable cycles between peak and valley values, appearing a multi-fiuctuational process line of relative gathering and migration since then. The low numerical value distribution of these two types of trace elements in the aeolian sand facies represents erosion and accumulation under wind force during the cold-dry climate. Whereas their enrichments in both fluvio-lacustrine facies and palaeosols are related to the valley’s special low-lying physiognomic position between the Ordos Plateau and the Loess Plateau under the warm and humid climate conditions. The above relatively migrated and gathered change of the trace elements is the result of 27 climatic cycles of cold-dry and warm-humid, which is probably caused by repeated alternations of winter monsoon and summer monsoon in the Mu Us Sandy Land influenced by the climate vicissitudes in northern hemisphere during glacial and interglacial periods since 150 ka BP.     

基于BP神经网络的溶洞规模预测及应用

复杂岩溶地区的溶洞发育规模受地质构造、地区岩性、地下水动力系统等多种因素的影响,具有高度复杂性和非线性的特征。通过对岩溶区溶洞的赋存规律研究,确定影响溶洞发育规模的控制因素进行定量处理,收集已探明溶洞的样本数据。为克服已有研究对溶洞发育规模定性描述的模糊性,文章利用BP（Back Propagation）神经网络对自组织、自适应特性对数据样本的非线性关系揭示的能力,实现对溶洞发育规模的预测,并基于MATLAB实现BP神经网络结构的设计、训练、预测,其结果表明:BP神经网络模型对溶洞规模预测的精度高、收敛性能好。      

基于神经网络的区域生态环境分类方法研究

如何利用智能化信息提取技术,进行区域生态环境自动分类,一直是一种前沿性研究。该文在分析研究区自然景观特征的基础上,总结了影响区域生态环境的建模要素,基于神经网络技术,并根据生态环境的遥感探测机理,利用TM卫星遥感数据中的可见光、热红外、植被指数(NDVI)以及DEM数据,建立了基于BP神经网络的区域生态环境信息自动提取模型,形成了一种新的生态环境分类方法,其分类结果与实际情况完全一致。     

基坑支护结构水平变形预测的遗传神经网络方法

采用遗传算法和误差反向传播算法相结合的混合算法来训练前馈人工神经网络,先用遗传学习算法进行全局训练,再用BP算法进行精确训练。就遗传算法过程中的选择、变异进行了探索,提出了用BP网络训练产生变异的遗传算法。作为实例,将该方法应用于预测基坑支护结构水平变形中。结果表明,该方法有收敛速度较快、预测精度高等优点。     

改进后灰色神经网络及短期基坑沉降预测研究

在短期基坑沉降监测中,由于数据量少且呈非线性变化,沉降模型很难准确建立。灰色GM(1,1)对数据少、趋势性强、波动小的数据有较高的预测精度,但不能模拟复杂的非线性函数;BP神经网络可以对非线性数据进行学习训练,具有自学习、自适应能力;通过将GM(1,1)与BP神经网络组合,并优化网络部分的学习率、权值和阈值等,建立一种改进的灰色神经网络模型,该模型具有对非线性数据自学习、自适应能力和预测精度更高等优点。通过某基坑沉降监测分析,验证改进的灰色神经网络模型预测精度更高,适合短期建模,具有很好的实用性。     

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。