Pipe laying simulation with an active reel drive

Dynamic and quasi-static models for analysis of a pipe lay spread are presented in the paper. Depending on the type of a pipe, spooling it on a big drum (reel) may cause plastic deformations. When the pipe is reeled out at sea, again it undergoes plastic deformations. In order to model such a nonlinear behaviour, a model, which includes both elasto-plastic material characteristics and large deformations, is required. Discretisation of the pipe is performed by means of the Rigid Finite Element Method (RFEM), which allows us to study static and dynamic problems, taking into account required properties of the model. Different sea conditions are simulated and analysed. It is shown how operational conditions can limit the ability of laying operation by a dedicated vessel, due to an unstable system response. The second part of the paper presents an upgrade of a passive reel drive by application of an active drive. In the model used for control applications, the pipe is modelled by a neural network. This allows us to perform a real-time calculations. In this fast-response calculation model, nonlinear aspects of the system are taken into account. The improvements in a pipe laying vessel performance are presented by some results of numerical simulations. The conclusions are also formulated.     

Remote sensing of water depths in shallow waters via artificial neural networks

Determination of the water depths in coastal zones is a common requirement for the majority of coastal engineering and coastal science applications. However, production of high quality bathymetric maps requires expensive field survey, high technology equipment and expert personnel. Remotely sensed images can be conveniently used to reduce the cost and labor needed for bathymetric measurements and to overcome the difficulties in spatial and temporal depth provision. An Artificial Neural Network (ANN) methodology is introduced in this study to derive bathymetric maps in shallow waters via remote sensing images and sample depth measurements. This methodology provides fast and practical solution for depth estimation in shallow waters, coupling temporal and spatial capabilities of remote sensing imagery with modeling flexibility of ANN. Its main advantage in practice is that it enables to directly use image reflectance values in depth estimations, without refining depth-caused scatterings from other environmental factors (e.g. bottom material and vegetation). Its function-free structure allows evaluating nonlinear relationships between multi-band images and in-situ depth measurements, therefore leads more reliable depth estimations than classical regressive approaches. The west coast of the Foca, Izmir/Turkey was used as a test bed. Aster first three band images and Quickbird pan-sharpened images were used to derive ANN based bathymetric maps of this study area. In-situ depth measurements were supplied from the General Command of Mapping, Turkey (HGK). Two models were set, one for Aster and one for Quickbird image inputs. Bathymetric maps relying solely on in-situ depth measurements were used to evaluate resultant derived bathymetric maps. The efficiency of the methodology was discussed at the end of the paper. It is concluded that the proposed methodology could decrease spatial and repetitive depth measurement requirements in bathymetric mapping especially for preliminary engineering application.     

基于人工神经网络的沿海风速多步预测研究

基于气象历史观测资料,将长短期记忆网络LSTM方法和Transformer模型结合提出了混合短期风速多步预测模型BLSTM TRA。以山东半岛南部沿海6个台站为研究区域,通过气象台站观测数据构建数据集。经与2018年ECMWF模式6 h预报结果对比分析,得出如下结论：构建的BLSTM TRA多步预测模型可大幅度降低风速误差,BLSTM TRA的1 h单步预测结果和ECMWF预报模式结果对比,其RMSE平均降低了58.9%,MAE平均降低了63.2%;风速误差和大风统计过程分析发现,BLSTM TRA模型具有一定的抗干扰能力,可以抓住短时大风等敏感信息,对于大风预报结果明显优于ECWMF模式和传统LSTM模型。     

基于雷达反射率因子和雷电定位数据的深度学习雷电预报模型

利用卷积神经网络和门控循环单元（Gated recurrent units ）神经网络,基于雷达反射率因子和雷电定位数据开展了雷电预报研究。首先构建了引用注意力机制的基于卷积神经网络和门控循环单元神经网络的深度学习模型（Attention ConvGRU）;然后将雷达反射率因子数据和对应时间段（6 min）的雷电定位数据处理成图像数据后输入深度学习模型,训练出可预报雷电的模型,包括3种模型：单雷电数据模型、单雷达数据模型和雷电〖CD*2〗雷达双数据模型;最后开展了预报试验和定量评估。综合评估表明,本文建立的雷电预报模型综合预报准确率达到96.74%,虚警率35.83%,关键成功指数（Critical Success Index, CSI）为0.2072。个例分析表明,预报模型对于具有明显移动趋势的雷暴过程（A类雷暴）的预报效果优于不具有明显移动趋势的雷暴过程（B类雷暴）,且随着B类雷暴强度减弱模型预报能力逐渐减弱。     

利用光学遥感数据、GIS及人工神经网络模型分析区域滑坡灾害

用光学遥感数据和地理信息系统(GIS)分析了马来西亚Selangor地区的滑坡灾害。通过遥感图像解译和野外调查,在研究区内确定出滑坡发生区。通过GIS和图像处理,建立了一个集地形、地质和遥感图像等多种信息的空间数据库。滑坡发生的因素主要为:地形坡度、地形方位、地形曲率及与排水设备距离;岩性及与线性构造距离;TM图像解译得到的植被覆盖情况;Landsat图像解译得到的植被指数;降水量。通过建立人工神经网络模型对这些因素进行分析后得到滑坡灾害图:由反向传播训练方法确定每个因素的权重值,然后用该权重值计算出滑坡灾害指数,最后用GIS工具生成滑坡灾害图。用遥感解译和野外观测确定出的滑坡位置资料验证了滑坡灾害图,准确率为82.92%。结果表明推测的滑坡灾害图与滑坡实际发生区域足够吻合。     

BP人工神经网络在青藏铁路南段地壳稳定性定量评价中的应用

将BP人工神经网络方法引入区域构造活动性、区域地壳稳定性研究领域,对青藏铁路南段沿线的构造活动性进行定量分析.选用断层运动速率、地震震级、温泉温度及剪切应变4个关键影响因子作为BP人工神经网络的输入向量,构造活动强度(α)作为输出向量,以α为定量判据,将全区划分为相对稳定区(α＜0.22)、较不稳定区(α≈0.22～0.38)、不稳定区(α≈0.38～0.69)、极不稳定区或强烈构造活动区(α≥0.69).在青藏铁路南段沿线划分出格仁错、崩错、当雄-羊八井、错那湖、唐古拉山口南、聂荣东北、聂荣西北、雅鲁藏布江断裂沿线、萨迦等不稳定区,在不稳定区内部进一步划分出申扎、蓬错、尼木、桑雄、羊八井5个极不稳定区.     

Elman神经网络在低渗储层敏感性预测中的应用

油气储层敏感性是制约低渗储层有效开发的重要因素之一,但在勘探开发早期储层敏感性预测及分布规律的研究受到资料不足的限制.从实验敏感性分析结果入手,对Elman神经网络预测储层敏感性的过程及方法进行了系统的分析,并提出了通过设置虚拟井来预测储层敏感性平面分布的方法.结果表明:①影响低渗储层敏感性的主要因素有蒙皂石、伊利石、高岭石、绿泥石、伊-蒙混层、石英、长石的体积分数,孔隙度和渗透率;②Elman神经网络能够很好地预测低渗储层的敏感性,预测结果与实测结果绝对误差的平均值均低于0.04,水敏性样品的预测结果与实测结果绝对误差的平均值为0.001;③通过设置虚拟井位,能够很好地预测储层敏感性的平面分布规律.     

饱和砂土液化研究进展

根据国内外文献资料,从三方面总结了饱和砂土的最新进展:饱和砂土液化判别方法、砂土液化的试验研究以及液化后分析,特别是探讨了液化对上部结构的影响。最后指出了存在的问题和今后的研究方向。     

基于神经网络的滑移隔震结构智能半主动控制

考虑上部结构的刚度和阻尼,使用神经网络控制算法计算基底摩擦力的大小,研究了滑移隔震结构的半主动控制。对计算实例的分析表明,通过半主动控制的滑移隔震结构不但具有较好的隔震效果,且能有效地减小基底的最大滑移量及残余位移。为对比各种控制方法的控制效果,文中还利用Bang-Bang控制和瞬时最优控制算法对滑移隔震结构进行了半主动控制。对比分析表明,基于神经网络控制算法的控制效果优于其它控制算法,具有反馈量少,稳健性强等特点。     

基于深度学习的海温观测数据质量控制应用研究

基于国家海洋信息中心质量控制后的西太平洋10度方区约100万站次温盐实测历史调查资料,对经过26种严格质量控制方法的综合海温质量符进行分类分析,首次将深度学习技术应用于海洋数据质量控制多分类(multiclass classification)算法与应用研究.通过人工合成少数类样本和加权损失函数方法减少多数类的频率来降...