Machine learning for pore-water pressure time-series prediction: Application of recurrent neural networks     

Xin Wei  Lulu Zhang  Hao-Qing Yang  Limin Zhang  Yang-Ping Yao 《地学前缘(英文版)》2021,12(1):453-467

Knowledge of pore-water pressure(PWP)variation is fundamental for slope stability.A precise prediction of PWP is difficult due to complex physical mechanisms and in situ natural variability.To explore the applicability and advantages of recurrent neural networks(RNNs)on PWP prediction,three variants of RNNs,i.e.,standard RNN,long short-term memory(LSTM)and gated recurrent unit(GRU)are adopted and compared with a traditional static artificial neural network(ANN),i.e.,multi-layer perceptron(MLP).Measurements of rainfall and PWP of representative piezometers from a fully instrumented natural slope in Hong Kong are used to establish the prediction models.The coefficient of determination(R^2)and root mean square error(RMSE)are used for model evaluations.The influence of input time series length on the model performance is investigated.The results reveal that MLP can provide acceptable performance but is not robust.The uncertainty bounds of RMSE of the MLP model range from 0.24 kPa to 1.12 k Pa for the selected two piezometers.The standard RNN can perform better but the robustness is slightly affected when there are significant time lags between PWP changes and rainfall.The GRU and LSTM models can provide more precise and robust predictions than the standard RNN.The effects of the hidden layer structure and the dropout technique are investigated.The single-layer GRU is accurate enough for PWP prediction,whereas a double-layer GRU brings extra time cost with little accuracy improvement.The dropout technique is essential to overfitting prevention and improvement of accuracy.  相似文献   

Late Holocene slip rate and average recurrence interval of great earthquakes on the Shangzhi segment of the Yilan‐Yitong Fault Zone,northeastern China: Constraints from paleo‐earthquakes and historical written records          下载免费PDF全文

Zhongyuan Yu  Peizhen Zhang  Wei Min  Qinghai Wei  Yugang Liu 《Island Arc》2018,27(1)

The Yilan‐Yitong Fault Zone (YYFZ) is considered to be the key branch of the Tancheng‐Lujiang Fault Zone (TLFZ) in northeastern China. Although the Mesozoic and early Cenozoic deformation of the YYFZ has been studied intensively over the past century, few estimates of slip rate and recurrence interval of large earthquakes in the late Quaternary, which are the two most important parameters for understanding the potential seismic hazard of this crucial structure, were obtained. Based on integrated interpretations of high resolution satellite images and detailed geologic and geomorphic mapping, linear landforms were identified, including fault scarps and troughs, along the Shangzhi segment of the YYFZ, which exceeds 25 km in length. Synthesized results of trench excavations and differential GPS measurements of terrace surfaces indicate that two events (E1, E2) occurred along the Shangzhi segment during the late Holocene, which resulted in 3.2 ±0.1 m of total vertical co‐seismic displacement with clear features of thrust motion. ¹⁴C dating of samples suggests that event E1 occurred between 440 ±30 years BP and 180 ±30 years BP and that event E2 occurred between 4 090 ±30 years BP and 3 880 ±30 years BP, which indicates that the minimum vertical slip rate of the Shangzhi segment of the YYFZ has been approximately 0.8 ±0.03 mm/year during the late Holocene. Constraints from paleo events and the slip rate suggest that the average recurrence interval of major earthquakes on the YYFZ is 3 800 ±200 years. Historical documents in Korea show that event E1 possibly corresponds to the earthquake that occurred in AD 1810 (the Qing Dynasty in Chinese history) in the Ningguta area, which had surface‐wave magnitude (Ms) of 6.8–7.5. Studies of kinematics show that the right‐lateral strike‐slip with a reverse component has been dominant along the YYFZ during the late Holocene.  相似文献   

Comparison of bootstrap confidence intervals for an ANN model of a karstic aquifer response     

Ioannis Trichakis  Ioannis Nikolos  George P. Karatzas 《水文研究》2011,25(18):2827-2836

Following many applications artificial neural networks (ANNs) have found in hydrology, a question has been rising for quantification of the output uncertainty. A pre‐optimized ANN simulated the hydraulic head change at two observation wells, having as input hydrological and meteorological parameters. In order to calculate confidence intervals (CI) for the ANN output two bootstrap methods were examined namely bootstrap percentile and BCa (Bias‐Corrected and accelerated). The actual coverage of the CI was compared to the theoretical coverage for different certainty levels as a means of examining the method's reliability. The results of this work support the idea that the bootstrap methods provide a simple tool for confidence interval computation of ANNs. Comparing the two methods, the percentile requires fewer calculations and yields narrower intervals with similar actual coverage to that of BCa. Overall, the actual coverage was proved lower than desired when not modeled points were present in the data subset. Copyright © 2011 John Wiley & Sons, Ltd.  相似文献   

基于深度学习的天气雷达回波序列外推及效果分析     

黄兴友  马玉蓉  胡苏蔓 《气象学报》2021,79(5):817-827

天气雷达探测资料是进行强对流天气临近预报的主要参考数据。针对传统雷达回波外推方法中存在资料信息利用率不足和外推时效有限的问题,文中利用神经网络进行雷达回波的外推、利用预测神经网络模型进行2 h以内的回波变化预报。回波外推问题的关键是回波时、空序列预测问题,该网络具有解决时间记忆问题的长、短时记忆单元（Long Short-Term Memory,LSTM）和提取空间特征的卷积模块。应用福建、江苏和河南多年的雷达探测资料构造训练和测试数据集。为消除降水的不平衡和提高对强回波的预报准确率,网络采用带权重的损失函数进行训练。对光流法和预测神经网络进行测试集检验以及个例分析,结果表明,在相同外推时效和检验反射率阈值的情况下,预测神经网络的临界成功指数、命中率均高于光流法,虚警率低于光流法。不同类型降水预测神经网络的SSIM值（structural similarity）均高于光流法,且层状云降水的SSIM值比对流云降水的大。因此,预测神经网络对强回波的预报能力高于光流法;在预报时效性上,预测神经网络模型具有一定的优越性;预测神经网络对层状云降水预报的准确率比对流云降水的高。   相似文献   

GPS卫星原子钟频率稳定度表征方法分析   总被引：1，自引：0，他引：1  

冯遂亮  宋力杰 《全球定位系统》2009,34(1):19-23

原子钟频率稳定度的表征通常在时域和频域进行，而对于“尺度域”小波方差表征方法的分析研究较少。对该方法进行系统归纳总结的基础上，给出时域尺度域两类不同表征方法的实用计算公式及其置信度计算方法，通过对各种方差特性的比较分析，指出其适用范围。分析表明：时域系列方差计算简单、直接，各具特点，可以满足不同的实际需要；尺度域的小波方差是一种具有独特优越性的新方法，但可以达到时域表征方法的效果，适用于进行原子钟频率稳定度的估计。  相似文献   

Automated threshold selection methods for extreme wave analysis   总被引：2，自引：0，他引：2  

Paul Thompson  Yuzhi Cai  Dominic Reeve  Julian Stander 《Coastal Engineering》2009,56(10):1013-1021

The study of the extreme values of a variable such as wave height is very important in flood risk assessment and coastal design. Often values above a sufficiently large threshold can be modelled using the Generalized Pareto Distribution, the parameters of which are estimated using maximum likelihood. There are several popular empirical techniques for choosing a suitable threshold, but these require the subjective interpretation of plots by the user.In this paper we present a pragmatic automated, simple and computationally inexpensive threshold selection method based on the distribution of the difference of parameter estimates when the threshold is changed, and apply it to a published rainfall and a new wave height data set. We assess the effect of the uncertainty associated with our threshold selection technique on return level estimation by using the bootstrap procedure. We illustrate the effectiveness of our methodology by a simulation study and compare it with the approach used in the JOINSEA software. In addition, we present an extension that allows the threshold selected to depend on the value of a covariate such as the cosine of wave direction.  相似文献   

On the Occurence of Extreme Events in Long-term Correlated and Multifractal Data Sets     

Mikhail I. Bogachev  Jan F. Eichner  Armin Bunde 《Pure and Applied Geophysics》2008,165(6):1195-1207

We review recent studies of the statistics of return intervals (i) in long-term correlated monofractal records and (ii) in multifractal records in the absence (or presence) of linear long-term correlations. We show that for the monofractal records which are long-term power-law correlated with exponent γ, the distribution density of the return intervals follows a stretched exponential with the same exponent γ and the return intervals are long-term correlated, again with the same exponent γ. For the multifractal record, significant differences in scaling behavior both in the distribuiton and correlation behavior of return intervals between large events of different magnitudes are demonstrated. In the absence of linear long-term correlations, the nonlinear correlations contribute strongly to the statistics of the return intervals such that the return intervals become long-term correlated even though the original data are linearly uncorrelated (i.e., the autocorrelation function vanishes). The distribution density of the return intervals is mainly described by a power law.  相似文献   

Stability of motion in the Sitnikov 3-body problem     

P. Soulis  T. Bountis  R. Dvorak 《Celestial Mechanics and Dynamical Astronomy》2007,99(2):129-148

We study the stability of motion in the 3-body Sitnikov problem, with the two equal mass primaries (m ₁ = m ₂ = 0.5) rotating in the x, y plane and vary the mass of the third particle, 0 ≤ m ₃ < 10⁻³, placed initially on the z-axis. We begin by finding for the restricted problem (with m ₃ = 0) an apparently infinite sequence of stability intervals on the z-axis, whose width grows and tends to a fixed non-zero value, as we move away from z = 0. We then estimate the extent of “islands” of bounded motion in x, y, z space about these intervals and show that it also increases as |z| grows. Turning to the so-called extended Sitnikov problem, where the third particle moves only along the z-axis, we find that, as m ₃ increases, the domain of allowed motion grows significantly and chaotic regions in phase space appear through a series of saddle-node bifurcations. Finally, we concentrate on the general 3-body problem and demonstrate that, for very small masses, m ₃ ≈ 10⁻⁶, the “islands” of bounded motion about the z-axis stability intervals are larger than the ones for m ₃ = 0. Furthermore, as m ₃ increases, it is the regions of bounded motion closest to z = 0 that disappear first, while the ones further away “disperse” at larger m ₃ values, thus providing further evidence of an increasing stability of the motion away from the plane of the two primaries, as observed in the m ₃ = 0 case.  相似文献   

Statistical significance of modal parameters of bridge systems identified from strong motion data     

Yalin Arici  Khalid M. Mosalam 《地震工程与结构动力学》2005,34(10):1323-1341

Modal parameters of structural systems have commonly been determined using system identification (SI) methods for damage detection and health monitoring. For determining the deterioration of the integrity of structural systems correctly, modal parameters of a healthy structure have to be obtained with adequate certainty so that these parameters can be used as reliable references for the healthy system to compare with those of the damaged system. In this study, the statistical significance of modal parameters identified using strong motion time histories recorded on two bridge structures is assessed. The confidence intervals of identified modal frequencies and damping ratios are obtained using Monte Carlo simulations and sensitivity analyses in conjunction with eigenrealization algorithm. The dependence of the statistical bounds on model parameters is examined. The effect of using different number of sensors on the statistical significance is evaluated using simulated time history data from a validated finite element model of a bridge. Copyright © 2005 John Wiley & Sons, Ltd.  相似文献   

Accuracy assessment of lidar-derived digital elevation models   总被引：2，自引：0，他引：2  

Fernando J.  Aguilar  Jon. P.  Mills 《The Photogrammetric Record》2008,23(122):148-169

Despite the relatively high cost of airborne lidar-derived digital elevation models (DEMs), such products are usually presented without a satisfactory associated estimate of accuracy. For the most part, DEM accuracy estimates are typically provided by comparing lidar heights against a finite sample of check point coordinates from an independent source of higher accuracy, supposing a normal distribution of the derived height differences or errors. This paper proposes a new methodology to assess the vertical accuracy of lidar DEMs using confidence intervals constructed from a finite sample of errors computed at check points. A non-parametric approach has been tested where no particular error distribution is assumed, making the proposed methodology especially applicable to non-normal error distributions of the type usually found in DEMs derived from lidar. The performance of the proposed model was experimentally validated using Monte Carlo simulation on 18 vertical error data-sets. Fifteen of these data-sets were computed from original lidar data provided by the International Society for Photogrammetry and Remote Sensing Working Group III/3, using their respective filtered reference data as ground truth. The three remaining data-sets were provided by the Natural Environment Research Council's Airborne Research and Survey Facility lidar system, together with check points acquired using high precision kinematic GPS. The results proved promising, the proposed models reproducing the statistical behaviour of vertical errors of lidar using a favourable number of check points, even in the cases of data-sets with non-normally distributed residuals. This research can therefore be considered as a potentially important step towards improving the quality control of lidar-derived DEMs.  相似文献