共查询到20条相似文献，搜索用时 46 毫秒
1.
Tae-Woong Kim Hosung Ahn 《Stochastic Environmental Research and Risk Assessment (SERRA)》2009,23(3):367-376
Missing data in daily rainfall records are very common in water engineering practice. However, they must be replaced by proper
estimates to be reliably used in hydrologic models. Presented herein is an effort to develop a new spatial daily rainfall
model that is specifically intended to fill in gaps in a daily rainfall dataset. The proposed model is different from a convectional
daily rainfall generation scheme in that it takes advantage of concurrent measurements at the nearby sites to increase the
accuracy of estimation. The model is based on a two-step approach to handle the occurrence and the amount of daily rainfalls
separately. This study tested four neural network classifiers for a rainfall occurrence processor, and two regression techniques
for a rainfall amount processor. The test results revealed that a probabilistic neural network approach is preferred for determining
the occurrence of daily rainfalls, and a stepwise regression with a log-transformation is recommended for estimating daily
rainfall amounts. 相似文献
2.
土木工程结构健康诊断中的统计识别方法综述 总被引：11，自引：1，他引：11
本文对土木工程结构健康诊断中的统计识别方法进行了综述。对统计识别中的统计系统识别方法(Bayes模型修正、随机有限元模型修正)、统计模式识别方法和概率神经网络方法的基本理论及其在土木工程结构健康诊断中的研究现状进行了论述，在此基础上提出了土木工程结构健康诊断中统计识别方法需要解决的关键问题和研究发展方向。 相似文献
3.
Pijush Samui 《地震工程与工程振动(英文版)》2007,6(4):331-336
Determining the liquefaction potential of soil is important in earthquake engineering. This study proposes the use of the Relevance Vector Machine (RVM) to determine the liquefaction potential of soil by using actual cone penetration test (CPT) data. RVM is based on a Bayesian formulation of a linear model with an appropriate prior that results in a sparse representation. The results are compared with a widely used artifi cial neural network (ANN) model. Overall, the RVM shows good performance and is proven to be more accurate than the ANN model. It also provides probabilistic output. The model provides a viable tool for earthquake engineers to assess seismic conditions for sites that are susceptible to liquefaction. 相似文献
4.
The work develops the approximation approach to solving the inverse MTS problem with the use of neural networks. The inverse
problem is considered in model classes of parametrized geoelectric structures, whose electric conductivity is controlled by
a few hundreds of macroparameters (N ∼ 300). An approximate inverse operator of the problem is constructed for each model class as a neural network, whose coefficients
are determined in the process of training on a representative sample of standard examples of forward problem solutions. The
problem of determination of the model class of geolectric structures corresponding to the presented input MT data is solved
with the use of the neural network classifier constructed for the available set of model classes of structures. Regularizing
factors and errors of the neural network method are analyzed. The operation of the algorithm is illustrated by examples of
the 2-D inversion of synthetic MT data. 相似文献
5.
Ground motion models (GMMs) are traditionally developed from a frequentist approach. The Bayesian framework has received recent attention in developing nonergodic models, measuring uncertainty, or updating the model with additional data. However, no neural networks are developed to date in this framework to predict ground motion parameters or spectra. Hence, the present work develops a probabilistic Bayesian neural network (PBNN) to next-generation attenuation – West2 and Subduction databases using variational inference with mean-field assumption. Network inputs are magnitude, rupture distance, hypocentral depth, shear wave velocity, style of faulting, and region flags; outputs are peak ground values and response spectra. Both models have two hidden layers with seven neurons in each hidden layer. The models are verified for potential overfit, and their performance is validated through the parametric study by varying inputs. The output of a deterministic model is a point estimate. Considering probabilistic layers in hidden and output layers enables the model to capture within-model epistemic uncertainty and aleatory variability. Obtained aleatory standard deviations are consistent with other models. Mean epistemic uncertainty and aleatory variability are in the range 0.07–0.10 and 0.62–0.78 (ln units) for NGA-West2 and 0.09–0.16 and 0.67–0.95 for NGA-Sub models, respectively. The correlation coefficients between recorded and overall mean predictions ranged from 0.94 to 0.97 for NGA-the West2 model and from 0.91 to 0.95 for the NGA-Sub models. Network performance for out-of-training inputs showed increased epistemic deviations with no effect on aleatory deviations. 相似文献
6.
7.
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
8.
Hydrodynamic theory allows us to use the speed of a shock wave front to determine the yield of an explosion. On the basis of this theory we developed a neural network to estimate a yield of underground explosions from the shock wave radius versus time (RVT) data, as measured by continuous reflectometry for radius versus time experiments (CORRTEX). The proposed method not only replaces the subjective elements of conventional algorithms, but produces significantly improved yield estimates. The network was trained with real hydrodynamic data and its performance on unseen test events was studied. A backpropagation network was employed; the architecture consisted of ten input units, a hidden layer with eleven hidden units, and one output unit. The network was trained with thousands of input-output measurement vectors, the feasible input set, derived from the RVT data from only four other training or standard events (selected on the basis of the given RVT data from the unknown event). The feasible input vectors were propagated through the trained network and the network output was used to select the optimum yield estimate. Elements of the input vector were: center of energy (COE) offsets, shock front radii, and time onset and interval of analysis for both the standard and unknown events. We studied the performance of the proposed system using 24 Nevada Test Site (NTS) events that were located in the geologic medium tuff. Sensitivity analysis of the proposed method to the assumed nominal COE offset is discussed. Variations of the proposed system that might lead to further improvements in performance are suggested. 相似文献
9.
This paper examines the sensitivity of seismic hazard analyses to various site response analysis procedures. Site effects are incorporated in the hazard calculations using a probabilistic approach and specifically the methodology of Bazzurro and Cornell [1] for the transformation of a generic ground-motion prediction equation to a site-specific one. The paper explores the sensitivity of the median amplification function, its standard deviation and the resulting surface hazard curve, to different methods of site response analysis and model input parameters. The computed site-specific surface hazard curves are also compared with those obtained from a generic soil ground-motion prediction equation. For the two sites investigated, it is shown that the choice of equivalent linear or nonlinear analysis with different constitutive model parameters has a significant impact on the hazard results. The sandy site was seen to be more sensitive to the site response analysis approach employed than the clayey site. 相似文献
10.
Ghassem Habibagahi 《Soil Dynamics and Earthquake Engineering》1998,17(1):53-56
Estimation of the magnitude of reservoir induced seismicity is essential for seismic risk analysis of dam sites. Different geological and empirical methods dealing with the mechanism or magnitude of such earthquakes are available in the literature. In this study, a method based on an artificial neural network utilizing radial basis functions (RBF network) was employed to analyze the problem. The network has only two input neurons, one representing the maximum depth of the reservoir and the other being a comprehensive parameter representing reservoir geometry. Magnitudes of the induced earthquakes predicted using the RBF network were compared with the actual recorded data. Compared with the conventional statistical approach, the proposed method gives a better prediction, both in terms of coefficients of correlation and error rates. 相似文献
11.
Alex J. Cannon 《水文研究》2010,24(6):673-685
Parameters in a generalized extreme value (GEV) distribution are specified as a function of covariates using a conditional density network (CDN), which is a probabilistic extension of the multilayer perceptron neural network. If the covariate is time or is dependent on time, then the GEV‐CDN model can be used to perform nonlinear, nonstationary GEV analysis of hydrological or climatological time series. Owing to the flexibility of the neural network architecture, the model is capable of representing a wide range of nonstationary relationships. Model parameters are estimated by generalized maximum likelihood, an approach that is tailored to the estimation of GEV parameters from geophysical time series. Model complexity is identified using the Bayesian information criterion and the Akaike information criterion with small sample size correction. Monte Carlo simulations are used to validate GEV‐CDN performance on four simple synthetic problems. The model is then demonstrated on precipitation data from southern California, a series that exhibits nonstationarity due to interannual/interdecadal climatic variability. Copyright © 2009 Her Majesty the Queen in right of Canada. Published by John Wiley & Sons, Ltd. 相似文献
12.
A neural network is employed to select earthquake waves in a time history approach for structural dynamics. The neural network is a preferable alternative to an expert system because knowledge can easily be renewed. It involves a back propagation model having three layers (one input, one hidden and one output layer) and is used to avoid inappropriate earthquake input prior to practical numerical computations. Knowledge to categorize the earthquake waves is acquired through network training with earthquake response spectra and structural responses. The trained network is tested by categorizing the responses of three types of unknown structures caused by 50 previously recorded earthquakes. Comparisons are made with analogous data from the traditional site dominant period method. Results demonstrate that, unlike the latter method, a neural network is generally more successful as the number of training patterns increases. 相似文献
13.
Giovanna Cultrera Valerio De Rubeis Nikos Theodoulidis Héloïse Cadet Pierre-Yves Bard 《Bulletin of Earthquake Engineering》2014,12(4):1493-1514
The Horizontal-to-Vertical Spectral Ratio from earthquake (HVSR) and from ambient noise (HVN) recordings realistically indicate the fundamental frequency of soil response but, for the majority of the worldwide examined sites, they do not provide reliable amplification curves as calculated by the earthquake standard Spectral Ratio (SSR). Given the fact that HVSR and especially HVN can be easily obtained, it is challenging to search for a meaningful correlation with SSR amplification functions for the entire frequency band and to use the results for the SSR estimate at a further site where only noise measurements are available. To this aim we used recordings from 75 sites worldwide and we applied a multivariate statistical approach (canonical correlation analysis) to investigate and quantify any correlation among spectral ratios. The canonical correlation between SSR and HVN is then used to estimate the expected SSR at each site by a weighted average of the SSR values measured at the other sites; the weights are properly set to account more for sites with similar behaviour in terms of the canonical correlation results between HVN and SSR. This procedure, repeated for all sites in turn, constitutes the basis of a cross validation. The comparison between the inferred and the original SSR highlights the improvements of site response estimation with respect to the use of ambient noise techniques. The goodness and limitations of the reconstruction procedure are explained by specific geological settings. 相似文献
14.
Mansour A. Al-Garni 《Acta Geophysica》2010,58(1):143-162
A new approach is proposed in order to interpret spontaneous potential (self-potential) anomalies related to simple geometric-shaped
models such as sphere, horizontal cylinder, and vertical cylinder. This approach is mainly based on using neural network inversion
of SP anomalies, particularly modular algorithm, for estimating the parameters of different simple geometrical bodies. However,
Hilbert transforms are involved to determine the origin location in order to reduce the parameters which minimize the ambiguity
in the inverted models. The inversion has been tested first on synthetic data from different models, using only one well-trained
network. The results of inversion show that the parameter values derived by the inversion are identical to the true values
of parameters. Noise analysis has been also examined, where the results of the inversion produce acceptable results up to
10% of white Gaussian noise. The validity of the neural network inversion is demonstrated through published real field SP
taken from southern Bavarian Woods, Germany. A comparable and acceptable agreement is shown between the results of inversion
derived by the neural network and those from the real field data. 相似文献
15.
A systematic comparison of two basic types of neural network, static and dynamic, is presented in this study. Two back-propagation (BP) learning optimization algorithms, the standard BP and conjugate gradient (CG) method, are used for the static network, and the real-time recurrent learning (RTRL) algorithm is used for the dynamic-feedback network. Twenty-three storm-events, about 1632 rainfall and runoff data sets, of the Lan-Yang River in Taiwan are used to demonstrate the efficiency and practicability of the neural networks for one hour ahead streamflow forecasting. In a comparison of searching algorithms for a static network, the results show that the CG method is superior to the standard BP method in terms of the efficiency and effectiveness of the constructed network's performance. For a comparison of the static neural network using the CG algorithm with the dynamic neural network using RTRL, the results show that (1) the static-feedforward neural network could produce satisfactory results only when there is a sufficient and adequate training data set, (2) the dynamic neural network generally could produce better and more stable flow forecasting than the static network, and (3) the RTRL algorithm helps to continually update the dynamic network for learning—this feature is especially important for the extraordinary time-varying characteristics of rainfall–runoff processes. 相似文献
16.
多属性融合技术在苏14井区的应用 总被引：2，自引：1，他引：1
In this study area the geological conditions are complicated and the effective sandstone is very heterogeneous. The sandstones
are thin and lateral and vertical variations are large. We introduce multi-attribute fusion technology based on pre-stack
seismic data, pre-stack P- and S-wave inversion results, and post-stack attributes. This method not only can keep the fluid
information contained in pre-stack seismic data but also make use of the high SNR characteristics of post-stack data. First,
we use a one-step recursive method to get the optimal attribute combination from a number of attributes. Second, we use a
probabilistic neural network method to train the nonlinear relationship between log curves and seismic attributes and then
use the trained samples to find the natural gamma ray distribution in the Su-14 well block and improve the resolution of seismic
data. Finally, we predict the effective reservoir distribution in the Su-14 well block. 相似文献
17.
Mitsuru Nakamura Sami F. Masri Anastassios G. Chassiakos Thomas K. Caughey 《地震工程与结构动力学》1998,27(9):997-1010
A neural network-based approach is presented for the detection of changes in the characteristics of structure-unknown systems. The approach relies on the use of vibration measurements from a ‘healthy’ system to train a neural network for identification purposes. Subsequently, the trained network is fed comparable vibration measurements from the same structure under different episodes of response in order to monitor the health of the structure. The methodology is applied to actual data obtained from ambient vibration measurements on a steel building structure that was damaged under strong seismic motion during the Hyogo-Ken Nanbu Earthquake of 17 January 1995. The measurements were done before and after repairs to the damaged frame were made. A neural network is trained with data after the repairs, which represents ‘healthy’ condition of the building. The trained network, which is subsequently fed data before the repairs, successfully identified the difference between the damaged storey and the undamaged storey. Through this study, it is shown that the proposed approach has the potential of being a practical tool for a damage detection methodology applied to smart civil structures. © 1998 John Wiley & Sons, Ltd. 相似文献
18.
A. W. P. Thomson 《Pure and Applied Geophysics》1996,146(1):163-193
Two neural network algorithms are applied to the short-term,1 to 3 days, prediction of theAp geomagnetic index. A multi-layer, back-propagation (MBP) network is used to implement a self-prediction filter forAp and this provides a forecast of the numerical value of the index. A probabilistic neural network (PNN) is used to estimate the probability distribution of theAp index, in six activity classes, and to provide a forecast of the single most likely activity class for each day. BothAp and an index of solar activity, based on the daily reports issued by the Space Environment Services Centre (Boulder), are input to the probabilistic net. It is found that the numerical forecasts of the MBP filter are most accurate at low, non-storm, levels of activity. This non-linear method provides quantitatively better estimates of activity than are produced by an existing linear prediction filter, particularly with increasing forward forecasting lag. At high levels of the solar activity index the PNN is found to anticipate storm classAp with around 60% accuracy in 1992 and 1993. Some details of the algorithms and implementation issues are described. It is concluded that interplanetary field and solar wind data will be significant components of any of the possible future developments which are discussed. 相似文献
19.
概率神经网络(PNN)以贝叶斯概率的方法描述测量数据,因而PNN在有噪声条件下的结构损伤检测方面具有巨大潜力。与此同时,PNN的网络规模随着训练样本的增加而增大,这极大地降低了网络运行速度。基于此,本文提出了基于主组分分析(PCA)的PNN损伤定位方法,分别用传统PNN(TPNN)、主组分分析PNN(PCAPNN)和自适应PNN(APNN)三种模型进行了悬索桥的损伤定位研究。研究发现,APNN的识别精度最好,PCAPNN次之,TPNN最差。但APNN需要很长的训练时间,网络规模较大;其他两个网络几乎不需要训练时间,且PCAPNN网络规模较其他两个网络减少了1／3～1／4。在低噪声情况下,PCAPNN的识别效果基本上等同于APNN。 相似文献
20.
A new paradigm called self-recurrent neural network (SRNN) is proposed. Two SRNNs are utilized in a control system, one as an emulator and the other as a controller. To guarantee convergence and for faster learning, an approach using adaptive learning rate is developed by Lyapunov function. Finally, the neural network control algorithm is developed for on-line control of structural seismic response in real time. Simulation-results have shown that it can effectively control structural seismic response and make it consist with the desired response. © 1998 John Wiley & Sons, Ltd. 相似文献