共查询到19条相似文献，搜索用时 140 毫秒
1.
2.
为进一步提高大地电磁非线性反演的稳定性、运算效率及准确度，将遗传神经网络算法引入大地电磁反演.首先针对大地电磁二维地电模型建立BP（Back Propagation）神经网络基本框架进行学习训练，网络输入为已知地电模型的视电阻率参数，输出为该地电模型参数；再利用遗传算法对神经网络学习训练过程进行优化，计算出多种地电模型网络连接权值和阈值的最优解；最后将最优连接权值和阈值对未知模型进行反演测试，网络输入为未知地电模型的视电阻率参数，输出为该地电模型参数.模型实验表明：遗传神经网络算法充分结合了遗传算法的全局寻优性和神经网络的局部寻优性，相比单一神经网络算法，在网络学习训练中提高了解的收敛成功率和计算速度，在反演测试中能更准确地逼近真实模型.将遗传神经网络算法与最小二乘正则化反演进行对比，理论模型和实测数据都验证了遗传神经网络算法在大地电磁反演中的可行性和有效性. 相似文献
3.
4.
神经网络反演双侧向电阻率测井曲线的物理约束 总被引：4，自引：4，他引：4
以一种新型高分辨率双侧向测井仪器在二维轴对称地层模型中的模拟响应为训练集训练BP神经网络，得到了针对该双侧向测井仪的反演网络模型。在训练中，神经网络结构的确定一般采用交叉验证法，但这种验证法经验性的成分偏重，不能完全解决网络结构的范化问题。为此，本文在模型的训练中，在交叉验证的基础上，根据双侧向测井的原理和仪器响应特性，提出了一种新的物理约束方法；反演地层电阻率的误差应随着侵入半径的增加而加大，违反此规律的模型不予采纳。理论研究结果表明，由此得到的神经网络模型具有很好的范化能力。 相似文献
5.
结构网络最小混合型神经元网络油气预测 总被引：1，自引：0，他引：1
基于统计学习理论中的结构风险最小化原理，从理论上给出了神经网络的结构设计方法和实现过程。该方法能自适应地扩展神经网络的容量，从而完成网络的结构设计，并且在有限样本的情况下，阳大限度地提高网络的训练精度和泛化能力，进而提高神经网络预测结果的可靠性。此外，本方法可使神经网络同时具有多种类型的特性函数，增强了网络的信息处理能力。文中给出了该方法在大庆油田某开发区块储层油气检测的应用实例。 相似文献
6.
7.
本文以提高地震数据的成像质量为目标,提出一种智能的卷积神经网络降噪框架,从带有噪声的地震数据中自适应地学习地震信号。为了加速网络训练和避免训练时出现梯度消失现象,我们在网络中加入残差学习和批标准化的方法,并采用了ReLU激活函数和Adam优化算法优化网络。此外,Marmousi和F3数据集被用来对网络进行训练和测试,经过充分训练的网络不仅能在学习中保留地震数据特征,而且能去除随机噪声。首先充分地训练网络,从中提取出随机噪声,并保留学习到的地震数据特征,之后通过重建地震数据估算测试集中的波形特征。合成记录和实际数据的处理结果显示了深度卷积神经网络在随机噪声压制任务中的潜力,并通过实验验证表明了深度卷积神经网络框架有很好的去噪效果。 相似文献
8.
针对随机地震反演中存在的两个主要问题,随机实现含有噪声和难以从大量随机实现中挖掘有效信息,提出了一种基于神经网络的随机地震反演方法.通过对多组随机实现及其正演地震数据的计算,构建了基于序贯高斯模拟的训练集.这也为应用神经网络求解地球物理反问题,提供了一种有效建立训练集的方法.较之传统的神经网络反演,这种训练集不仅保证了学习样本具有多样性,同时还引入了空间相关性.数值模拟结果表明,该方法只需要通过单层前馈神经网络,就可以比较有效的解决一个500个阻抗参数的反演问题. 相似文献
9.
本文试图解释用ＢＰ神经网络解界面反问题时效果不佳的原因。文中首先从信息量的角度提出了ＢＰ神经网络训练本集容量的概念，给出了它的定义及组织训练样本集时应遵循的原则和方法。对于如何用ＢＰ神经网络解界面反问题，给出了其基本步骤，并根据上述训练样本集容量的概念及界面反总理的特殊性，给出了组织界面反问题训练样本集的方法。 相似文献
10.
为实现自动检测地震噪声波形是否异常,提出应用BP神经网络技术进行地震噪声波形检测.选取福建地震台网88个测震台站2018-2019年的地震噪声原始波形,计算波形的加速度功率谱密度(PSD)值作为神经网络的输入特征值,在MATLAB中构建BP神经网络进行学习训练和仿真测试.测试验证了训练后的BP神经网络模型具备了可靠的地震噪声波形是否异常的检测能力.应用BP神经网络检测地震噪声波形免去了人工判断的工作,实现全自动处理,提高了检测效率,为今后地震噪声波形质量自动监控提供了新的技术方法. 相似文献
11.
作为深度学习方法的一种,长短时记忆神经网络(LSTM)是一种信号处理的重要方法.本文基于实际观测地电场数据来合成训练集,对特定结构的长短时记忆神经网络进行训练,将训练所得网络对测试集数据进行测试后,将网络应用至实际观测数据.结果显示,经过训练的网络很好地学到了训练集样本的特征,对测试集数据的信噪比压制了约20 dB,并过滤了人为添加的特定频率的干扰成分,对实际观测数据处理后得到明显的日变、半日变以及半月变、月变、半年变、年变等潮汐响应,表明长短时记忆神经网络可以有效应用于地电场数据处理研究. 相似文献
12.
Ｔｈｅｍｅｄｉｕｍ┐ａｎｄｓｈｏｒｔ┐ｔｅｒｍｐｒｅｄｉｃｔｉｏｎｍｅｔｈｏｄｓｏｆｓｔｒｏｎｇｅａｒｔｈｑｕａｋｅｓｂａｓｅｄｏｎｎｅｕ┐ｒａｌｎｅｔｗｏｒｋＺＨＩ－ＱＩＡＮＧＨＡＮ（韩志强）ＢＩ－ＱＵＡＮＷＡＮＧ（王碧泉）Ｉｎｓｔｉｔｕｔｅｏｆ... 相似文献
13.
当前震后建筑经济损失评估模型得到的震后建筑经济损失评估精确度、效率低,针对单一神经网络易产生局部极值等问题,对神经网络方法进行了改进,提出LM-BP神经网络在震后建筑损失评估模型中的应用。输入样本要素为影响震后建筑经济损失的5项因素,输出样本是震后建筑经济损失评估结果,在此基础上采用LM-BP神经网络将训练转化成最小二乘问题,结合LM算法重新定义隐含层节点数量,构建基于LM-BP的神经网络震后经济损失评估模型,采用该模型获取最优震后建筑经济损失评估结果。仿真实验结果表明,所设计的评估模型最小评估误差为0.1%,相比同类模型具有高精确度的优势,是一种可靠的震后建筑经济损失评估模型。 相似文献
14.
One of the most important problems in hydrology is the establishment of rating curves. The statistical tools that are commonly used for river stage‐discharge relationships are regression and curve fitting. However, these techniques are not adequate in view of the complexity of the problems involved. Three different neural network techniques, i. e., multi‐layer perceptron neural network with Levenberg‐Marquardt and quasi‐Newton algorithms and radial basis neural networks, are used for the development of river stage‐discharge relationships by constructing nonlinear relationships between stage and discharge. Daily stage and flow data from three stations, Yamula, Tuzkoy and Sogutluhan, on the Kizilirmak River in Turkey were used. Regression techniques are also applied to the same data. Different input combinations including the previous stages and discharges are used. The models' results are compared using three criteria, i. e., root mean square errors, mean absolute error and the determination coefficient. The results of the comparison reveal that the neural network techniques are much more suitable for setting up stage‐discharge relationships than the regression techniques. Among the neural network methods, the radial basis neural network is found to be slightly better than the others. 相似文献
15.
《水文科学杂志》2013,58(5):896-916
Abstract The performances of three artificial neural network (NN) methods for combining simulated river flows, based on three different neural network structures, are compared. These network structures are: the simple neural network (SNN), the radial basis function neural network (RBFNN) and the multi-layer perceptron neural network (MLPNN). Daily data of eight catchments, located in different parts of the world, and having different hydrological and climatic conditions, are used to enable comparisons of the performances of these three methods to be made. In the case of each catchment, each neural network combination method synchronously uses the simulated river flows of four rainfall—runoff models operating in design non-updating mode to produce the combined river flows. Two of these four models are black-box, the other two being conceptual models. The results of the study show that the performances of all three combination methods are, on average, better than that of the best individual rainfall—runoff model utilized in the combination, i.e. that the combination concept works. In terms of the Nash-Sutcliffe model efficiency index, the MLPNN combination method generally performs better than the other two combination methods tested. For most of the catchments, the differences in the efficiency index values of the SNN and the RBFNN combination methods are not significant but, on average, the SNN form performs marginally better than the more complex RBFNN alternative. Based on the results obtained for the three NN combination methods, the use of the multi-layer perceptron neural network (MLPNN) is recommended as the appropriate NN form for use in the context of combining simulated river flows. 相似文献
16.
Many novel techniques for reconstructing rainfall‐runoff processes require hydrometeorologic and geomorphologic information for modelling. However, certain information is not always measurable. In this paper, we employ a special recurrent neural network to reconstruct the rainfall‐runoff process by using collected rainfall data. In addition, we propose an indirect system identification to overcome the drawback of a traditional, time‐consuming trial‐and‐error search. The indirect system identification is an efficient method to recognize the structure of a recurrent neural network. The unit hydrograph can be derived directly from the weights of the network due to a state‐space form embedded in the recurrent neural network. This improves the link between the weights of the network and the physical concepts that most neural networks fail to connect. The case studies of 41 events from 1966 to 1997 have been implemented in Taiwan's Wu‐Tu watershed, where the runoff path‐lines are short and steep. Two recurrent neural networks and one state‐space model are utilized to simulate the rainfall‐runoff processes for comparison. The results are validated by four criteria: coefficient of efficiency; peak discharge error; time to peak arrival error; total discharge volume error. The resulting data from the recurrent neural network reveal that the neural network proposed herein is appropriate for hydrological systems. Copyright © 2005 John Wiley & Sons, Ltd. 相似文献
17.
Artificial neural networks for removal of couplings in airborne transient electromagnetic data 
下载免费PDF全文
下载免费PDF全文 Kristoffer K. Andersen Casper Kirkegaard Nikolaj Foged Anders V. Christiansen Esben Auken 《Geophysical Prospecting》2016,64(3):741-752
Modern airborne transient electromagnetic surveys typically produce datasets of thousands of line kilometres, requiring careful data processing in order to extract as much and as reliable information as possible. When surveys are flown in populated areas, data processing becomes particularly time consuming since the acquired data are contaminated by couplings to man‐made conductors (power lines, fences, pipes, etc.). Coupled soundings must be removed from the dataset prior to inversion, and this is a process that is difficult to automate. The signature of couplings can be both subtle and difficult to describe in mathematical terms, rendering removal of couplings mostly an expensive manual task for an experienced geophysicist. Here, we try to automate the process of removing couplings by means of an artificial neural network. We train an artificial neural network to recognize coupled soundings in manually processed reference data, and we use this network to identify couplings in other data. The approach provides a significant reduction in the time required for data processing since one can directly apply the network to the raw data. We describe the neural network put to use and present the inputs and normalizations required for maximizing its effectiveness. We further demonstrate and assess the training state and performance of the network before finally comparing inversions based on unprocessed data, manually processed data, and artificial neural network automatically processed data. The results show that a well‐trained network can produce high‐quality processing of airborne transient electromagnetic data, which is either ready for inversion or in need of minimal manual processing. We conclude that the use of artificial neural network scan significantly reduce the processing time and its costs by as much as 50%. 相似文献
18.
A back-propagation neural network is successfully applied to pick first arrivals (first breaks) in a background of noise. Network output is a decision whether each half-cycle on the trace is a first or not. 3D plots of the input attributes allow evaluation of the attributes for use in a neural network. Clustering and separation of first break from non-break data on the plots indicate that a neural network solution is possible, and therefore the attributes are suitable as network input. Application of the trained network to actual seismic data (Vibroseis and Poulter sources) demonstrates successful automated first-break selection for the following four attributes used as neural network input: (1) peak amplitude of a half-cycle; (2) amplitude difference between the peak value of the half-cycle and the previous (or following) half-cycle; (3) rms amplitude ratio for a data window (0.3 s) before and after the half-cycle; (4) rms amplitude ratio for a data window (0.06 s) on adjacent traces. The contribution of the attributes based on adjacent traces (4) was considered significant and future work will emphasize this aspect. 相似文献