基于A- CGAN的深反射地震数据随机噪声压制方法研究

基于深度学习的地震数据噪声压制方法是当前地震数据去噪处理的重要方向。深度学习方法突破了传统滤波处理的局限,在对常规地震数据的噪声压制中表现出效率高、信噪分离效果好的特点。但针对深部弱有效反射数据,当前的深度学习方法特征提取能力有限,难以取得较好的去噪效果。笔者等结合深反射地震数据特点,针对当前深度学习噪声压制方法在特征提取及对数据集依赖上的局限,提出了基于注意力循环生成对抗网络(Attention Cycle- Consistent Generative Adversarial Networks,A- CGAN)的深反射地震数据随机噪声压制方法。借助循环一致生成对抗网络(Cycle- Consistent Generative Adversarial Networks,Cycle- GAN)的域映射思想,降低对数据集的要求。为了构建适用于深反射地震数据的去噪网络,从3个方面对Cycle- GAN进行改进：在Cycle- GAN的生成器（去噪器）中加入残差结构和注意力机制,用于加深网络深度和提高其特征提取能力;在Cycle- GAN的鉴别器中使用块判决,提高鉴别精度和准确度;在损失函数部分加入感知一致性损失函数,提升网络模型恢复纹理细节信息的能力。通过合成地震数据和实际深反射地震数据测试,验证了优化算法的有效性,体现了良好的应用价值。     

主余震作用下高强钢筋混凝土框架的易损性分析

为研究主余震作用下配置高强钢筋的RC框架结构的抗震性能,根据相同设计指标和"等强代换"的原则,设计了三榀六层四跨的钢筋混凝土框架,纵向受力钢筋的强度等级分别为HRB400、HRB500和HRB600。通过OpenSees平台进行有限元建模,将最大层间位移角作为结构损伤指标,反应地震作用的结构响应。选取15条地震加速度时程,通过重复法构造人工主余震序列并进行IDA分析,得到结构在不同强度的主余震序列作用下的地震易损性曲线。结果表明:主震PGA越大,余震的最大层间位移角越大,结构达到各极限状态的概率越高。对比配置不同钢筋强度的三榀框架的计算结果,可知"等强代换"原则下,相同主余震序列作用时,钢筋强度的提高对结构抗倒塌性能有不利影响,但影响有限。     

基础类型对框架结构及场地土地震响应影响试验研究

本文设计独立基础框架和整体箱型基础框架结构模型,基于试验数据的对比分析,探讨基础类型与地震动特性对场地土以及结构自身地震响应的影响。试验结果及分析表明:地表结构的存在总体上是放大了地表加速度响应,放大最大幅度达到了40%,影响范围可达3倍的结构跨度,且具有一定埋深的箱型基础的影响大于浅埋独立基础。由于土体对独立基础的约束相对较弱,导致独立基础结构模型的加速度响应总体上大于箱型基础的;独立基础结构模型可能发生摇摆运动导致结构基础竖向响应的频谱特性含有较多的高频成分。另外,地震动特性对结构响应也较显著,其中脉冲地震动NR波的影响最为显著。     

剪切屈服型可更换耗能梁的抗震性能研究

针对传统结构震后修复能力不足,带可更换构件的混合框架结构体系在地震作用下,可更换耗能构件集中损伤和耗散地震能量,保护其他构件不损伤或轻微损伤,更换损伤的耗能构件,即可实现结构预定功能震后可恢复。通过3个可更换耗能梁试件,研究其抗震性能。在此基础上,通过SAP2000有限元建模,对带可更换构件的混合框架结构进行非线性分析,研究整体结构体系的屈服机制、承载力和可更换耗能构件的可更换性能。结果表明:试件均发生剪切屈服型破坏,破坏特征包括腹板－加劲肋焊缝撕裂、腹板屈曲和腹板撕裂。各试件的滞回曲线非常饱满,具有优异的承载能力、变形能力和耗能能力;在地震作用下,带可更换构件的混合框架结构体系中各构件能够实现良好的有序屈服机制,可更换耗能构件具有较好的可更换性。     

形变监测高分辨率InSAR数据的精度分析

收集江苏高分辨率InSAR地面沉降监测结果及兴化、淮安同期水准测量资料和基岩标同期监测资料,利用79个水准点测量数据、7个基岩标（分别选取不同时段）测量数据共98个数据样本,将InSAR技术与水准测量及基岩标测量技术的监测结果进行对比发现,监测结果一致性率为86.73%,平均差值为2.08 mm,标准差为2.50 mm,最大差值为18.64 mm,表明InSAR可进行半定量形变监测,但在高层建筑、排水管线、大坝、桥梁等需高精度形变监测的工程方面仍不能替代水准测量。     

Temporal variations of reference evapotranspiration and controlling factors: Implications for climatic drought in karst areas

Variations in reference evapotranspiration (ET₀) and drought characteristics play a key role in the effect of climate change on water cycle and associated ecohydrological patterns. The accurate estimation of ET₀ is still a challenge due to the lack of meteorological data and the heterogeneity of hydrological system. Although there is an increasing trend in extreme drought events with global climate change, the relationship between ET₀ and aridity index in karst areas has been poorly studied. In this study, we used the Penman–Monteith method based on a long time series of meteorological data from 1951 to 2015 to calculate ET₀ in a typical karst area, Guilin, Southwest China. The temporal variations in climate variables, ET₀ and aridity index (AI) were analyzed with the Mann–Kendall trend test and linear regression to determine the climatic characteristics, associated controlling factors of ET₀ variations, and further to estimate the relationship between ET₀ and AI. We found that the mean, maximum and minimum temperatures had increased significantly during the 65-year study period, while sunshine duration, wind speed and relative humidity exhibited significant decreasing trends. The annual ET₀ showed a significant decreasing trend at the rate of ?8.02 mm/10a. However, significant increase in air temperature should have contributed to the enhancement of ET₀, indicating an “evaporation paradox”. In comparison, AI showed a slightly declining trend of ?0.0005/a during 1951–2015. The change in sunshine duration was the major factor causing the decrease in ET₀, followed by wind speed. AI had a higher correlation with precipitation amount, indicating that the variations of AI was more dependent on precipitation, but not substantially dependent on the ET₀. Although AI was not directly related to ET₀, ET₀ had a major contribution to seasonal AI changes. The seasonal variations of ET₀ played a critical role in dryness/wetness changes to regulate water and energy supply, which can lead to seasonal droughts or water shortages in karst areas. Overall, these findings provide an important reference for the management of agricultural production and water resources, and have an important implication for drought in karst regions of China.     

泥沙起动概率的高速摄影测量方法

为了通过实验测量泥沙起动概率,基于封闭槽道内开展的泥沙起动实验,系统论述了基于高速摄影的起动概率测量方法所涉及的起动概率定义、相机采样频率、图像处理方法、起动比例与运动比例的关系等关键问题.根据泥沙在水流紊动猝发作用下起动的物理机制,定义实测起动概率为单个紊动猝发周期内,从一定面积床面上冲刷外移的泥沙比例.根据推移质运动概念模型,提出高速图片拍摄间隔应小于推移质运动的中间时间尺度,以捕获泥沙的每一次起动.在高速图像处理时,通过连续两帧图片之间的灰度差实现运动比例的无偏差提取.进一步分析表明,运动泥沙可处于起动、止动或滑翔状态,运动比例等于起动比例、止动比例和滑翔比例之和,起动比例与运动比例之比等于两张图片之间的时间间隔除以推移质运动的中间时间尺度.基于上述关键技术,将高速摄影测量方法用于泥沙起动概率的测量,结果表明,临界起动水流条件时,床沙处于少量起动现象,对应的临界起动概率与窦国仁所推导的2.28%接近.     

Photon in a cavity – a Gedankenexperiment

The inertial and gravitational mass of electromagnetic radiation (i.e., a photon distribution) in a cavity with reflecting walls has been treated by many authors for over a century. After many contending discussions, a consensus has emerged that the mass of such a photon distribution is equal to its total energy divided by the square of the speed of light. Nevertheless, questions remain unsettled on the interaction of the photons with the walls of the box. In order to understand some of the details of this interaction, a simple case of a single photon with an energy $E_{\nu }=h\nu$ bouncing up and down in a static cavity with perfectly reflecting walls in a constant gravitational field $\mathit{g}$, constant in space and time, is studied and its contribution to the weight of the box is determined as a temporal average.     

GRAPES_GFS中三维参考大气的研究：参考态构造和实际预报试验

GRAPES_GFS（Global Regional Assimilation and PrEdiction System, Global Forecast System）采用基于等温大气构造的一维参考大气,前期的研究工作已在GRAPES_GFS的动力框架中引入了三维参考大气,并通过一系列理想试验检验了新方法的正确性以及计算精度。本部分研究工作,主要针对实际资料的预报试验,对比分析了不同三维参考态给定方法的优劣,并采用气候平均法,进行夏季两个月的四维变分循环预报试验。检验结果显示,使用三维参考大气后,模式的综合预报性能得到了提升,对流层高度场、温度场的预报偏差有所减小,长期积分过程中模式质量损失较为严重的问题也得到了明显的缓解。另外,通过动能谱的对比也可以看到使用三维参考大气后,模式在高层的能量耗散明显较小,能谱变化与观测更加吻合。      

小波框架的彩色图像泊松去噪

噪声是影响图像质量的重要因素,噪声的存在导致图像的某些特征细节不能被辨识以及图像的信噪比下降,所以图像去噪是图像处理中的一个重要问题.本文提出了基于小波框架的彩色图像的泊松去噪模型.在该模型中,赋权的l²项作为保真项,包含小波框架的l¹项作为正则项.同时,又提出了解决该模型的重新赋权的split Bregman算法.最后,利用仿真实验以及PSNR（峰值信噪比）指标对该模型的去噪效果进行评估,评估结果表明该算法可行、有效.