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

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.     

Repacking practice for G-3 engineering driller power unit with drive shaft of dual channel reverse circulation

In order to satisfy operating requirements for constant core drilling technology in reverse circulation with hollow-through DTH,the power unit of G-3 engineering driller was ameliorated. The new one with dual channel drive shaft, achieved the perfect assemble with transmission structure of the original power unit. It could interconvert according to need by using two sets of drive shafts with direct and reverse circulation. The repacked G-3 engineering driller carried on experiment in the field test in Luanchuan molybdenum mine of Henan, whose effect was very good.     

对地勘单位贯彻ISO9000《质量管理和质量保证》系列标准的几点认识

ISO9000系列标准是一套国际统一的质量管理标准,应用的主体内容是质量管理模式(质量体系)的建立与运行。其对地勘单位走向市场,到社会投标、承揽地勘项目,保证地勘工作成果质量,提高单位在地质市场中的信誉度具有十分重要的意义。     

广西若干重大基础地质特征

文章根据华南地区物探、化探、遥感编图广西综合解译成果,研究广西若干重大基础地质特征,着重讨论地壳结构、深部构造、主要隐伏深断裂带、基底性质等,以板块观点划分大地构造单元,提出广西地壳属陆壳型性质和扬子板块在广西的边界划在三江-大新一线的新认识.     

近50 a来辽河口水下三角洲演变特征研究

本文基于近50 a来的辽河口海域的水下地形资料,采用GIS技术手段研究了辽河口水下三角洲的演变特征。研究结果表明:近50 a来辽河水下三角洲外缘(蛤蜊岗子滩西缘和南缘、西滩西缘和南缘)出现较大幅度的侵蚀后退;蛤蜊岗子滩东缘淤积外移,南部滩尾东部有北移东偏趋势;近岸浅滩由淤涨变为基本稳定,局部有所侵蚀后退;受蛤蜊岗子滩向东淤进的影响,蛤蜊岗子滩东侧深槽有所缩窄;蛤蜊岗子滩和西滩之间深槽水域出现拓宽变浅的变化趋势。流域来水来沙减少是近年来辽河口水下三角洲外缘侵蚀后退的主要原因,外缘侵蚀泥沙在波流作用下向河口搬运是蛤蜊岗子滩东侧淤积、深槽缩窄的主要原因,近岸边滩人工围垦对水下三角洲演变的影响值得关注。     

北部湾灯光罩网渔场时空分布与海洋环境关系分析 *

文章根据大型灯光罩网渔船调查数据和卫星遥感海面风场(Sea Surface Wind, SSW)、海表温度(Sea Surface Temperature, SST)和叶绿素a浓度(Chlorophyll a concentration, Chl a)资料, 基于广义线性模型(Generalized Linear Model, GLM)对北部湾渔业资源单位捕捞努力量渔获量(Catch Per Unit Effort, CPUE)进行标准化, 应用多元线性回归等方法, 对北部湾灯光罩网渔场的时空分布及其与海洋环境的关系进行了分析。结果表明, 北部湾灯光罩网渔场适宜SST为27~29℃, Chl a为0.5~1.5mg·m ^-3。较高资源量出现在10月份中上旬, 分布在18°—19°N及20°—21°N海域。北部湾灯光罩网渔场的时空分布与季风、19°N附近的暖水池和Chl a等环境因素有关。     

江汉平原古生界构造结构特征及油气勘探方向

江汉平原古生界具有北西—南东向分为三块、北东—南西向分为三带及纵向上多层楼的构造结构特征。三个区块为当阳—京山构造区、荆州—仙桃构造区及鄂城—大冶构造区,每个区块均可划分出三个次级构造带,并在全区形成三个构造变形分带。这三个分带在变形强度、构造样式等方面具有特征性差异:北东侧的构造分带受东秦岭—大别山造山运动产生的挤压应力的直接作用,显示以逆冲推覆结构为主要特征;南西侧的构造分带在三个构造区有差别,在当阳—京山构造区形成斜坡稳定带,在另两个构造区则以推覆叠瓦构造为特征;中部构造分带一般显示前缘断褶或复合叠加构造。当阳—京山构造区东部和鄂城—大冶构造区应以逆冲推覆体为勘探对象;荆州—仙桃构造区的仙桃地区应以上古生界油气藏为勘探目标;当阳—京山构造区西部宜昌稳定带和荆州—仙桃构造区的珂理—簰洲地区则应以下古生界油气藏为勘探目标。     

超参数对GRU-CNN混合深度学习弹性阻抗反演影响研究

CNN-GRU混合深度学习反演弹性阻抗取得了较好的反演效果.但是,基于深度学习的叠前反演参数众多,包括内部深度学习网络可学习参数和外部超参数等,目前超参数选取对网络性能及计算速度影响尚缺乏系统性研究,这直接影响到了该方法的进一步推广应用.因此,本文在混合深度学习反演弹性阻抗基础上,探讨学习率、Ep-och、batch_...     

UNIT HYDROGRAPH DERIVED FROM A SPATIALLY DISTRIBUTED VELOCITY FIELD

A unit hydrograph model is proposed in which the watershed is decomposed into subareas which are individual cells or zones of neighbouring cells. The unit hydrograph is found for each subarea and the response at the outlet to excess rainfall on each subarea is summed to produce the watershed runoff hydrograph. The cell to cell flow path to the watershed outlet is determined from a digital elevation model. A constant flow velocity is assigned to each cell and the time lag between subarea input and response at the watershed outlet is found by integrating the flow time along the path from the subarea to the outlet. The response function for a subarea is modelled as a lagged linear reservoir in which the flow time is equal to the sum of a time of translation and an average residence time in the reservoir. It is shown that the assumption of a spatially varying, but time-invariant, velocity field underlying this model produces a linear system model for all subareas whose outputs can be summed in the manner indicated. An example application is presented for the 8.70 km² Severn watershed at Plynlimon in Wales using a 50 m digital elevation model in which the cell velocity is calculated by modifying an average watershed velocity according to the terrain slope and the drainage area of each cell. The resulting model reasonably reproduces the observed unit hydrograph.     

APPLICATION OF UNIT HYDROGRAPH TECHNIQUES TO SOLUTE TRANSPORT IN CATCHMENTS

Current models of solute movement in catchments are based on rainfall–runoff models and are consequently biased towards processes which determine the magnitude and timing of water flux. It is shown here that the instantaneous unit hydrograph (IUH), or runoff response function, obtained from a hydrograph is fundamentally different from the residence time distribution which governs the response to solutes/tracers. Using hydrometric and tracer data obtained from a small (25 ha) catchment in the humid tropics a modification of the IUH technique is demonstrated which also allows approximate modelling of the tracer data. New features of the modified conceptual model are identified with known hillslope processes.