青藏铁路沙害及其防治研究进展

青藏铁路是中国乃至世界上海拔最高、穿越沙漠冻土的高原铁路。建成以来风沙危害日趋严重,成为危及铁路安全运营的一大隐患,因此,沿线的风沙防治一直备受关注。由于青藏高原风力强劲,沙物质丰富加上人类活动的影响,铁路沙害呈现出分布相对集中,冻融与风力、水力复合侵蚀,不断发展并持续累积,风沙活动稳定性差等特点。沙害分为路基风蚀、道床积沙、磨蚀等类型。累计有轻度、中度、严重沙害路段440 km,主要分布在锡铁山、伏沙梁、红梁河、秀水河-北麓河、沱沱河、通天河、扎加藏布、错那湖等8个路段。目前铁路沙害防治以机械措施为主,在设置初期有一定的防沙效果,但随着时间的推移,最终会被积沙埋没而失效。因此,青藏铁路防沙应以生物措施（恢复植被）为主,机械措施为辅。     

冻结和林黄土力学性质的试验研究

对饱和冻结和林黄土在-2℃的温度条件下进行了单轴压缩试验和围压范围为1~5MPa的三轴压缩试验。试验结果表明：单轴压缩的应力应变曲线为应变软化型曲线,而三轴压缩试验中各围压下的应力应变曲线均为应变硬化型曲线,围压主要影响冻结和林黄土的初始变形行为,而对后期的变形影响不大;各围压下冻结黄土的体积应变都呈现出先体缩后体胀的特征,总体变均较小;围压对冻结和林黄土的屈服极限和强度影响不大;研究了初始切线模量的确定方法,用修正的Duncan-Chang模型对不同应变范围内的试验数据进行拟合,并由此确定了相应的初始切线模量,发现在0~4%应变范围拟合试验数据确定初始切线模量最为合理。     

川西攀枝花—西昌地区结晶基底的划分

长期以来，由于同位素年龄依据不足和没有正确区分晚二叠世热接触变质岩、喜马拉雅期动力变质岩与前震旦纪区域变质岩，攀西地区结晶基底的划分存在许多困难和问题。本文依据1：50000区域地质调查结果，将原仁和群(Pt1R)修改为晚二叠世岩浆岩和喜马拉雅期动力变质岩，将五马箐(岩)组(Pt1w)和顶针杂岩(Pt1D)修改为晚二叠世热接触变质岩和喜马拉雅期动力变质岩，将安宁村组(Ptlα)和纸房沟组(Pt1z)修改为震旦系地层，并依据岩石学、岩石化学和微量元素地球化学特征，将结晶基底划分为变质侵入体、变质表壳岩和TTG套岩，论述了结晶基底的成因和演化。     

冻结粉质黏土-桩基接触面剪切特性试验研究

在寒区工程中, 建筑物的冻拔病害和冻土-桩基接触面间的剪切特性密切相关。借助自制的试验模具, 采用压桩法对冻结粉质黏土中埋置的混凝土桩、 钢桩以及木桩进行了不同负温条件下的剪切试验。结果表明： 在负温下随着剪切位移的增加, 剪切力经历线性增长、 骤降的脆性破坏、 维持恒定三阶段。温度越低, 桩与冻土间的冰胶结力越大, 冻结强度越大, 残余强度越大, 破坏允许位移也越大。在-30 ℃时, 木桩与冻土间的冻结强度最大, 混凝土桩与冻土间的冻结强度次之, 钢桩与冻土间的冻结强度最小。混凝土桩、 钢桩对应的冻结强度及残余强度与温度的关系可用线性拟合, 木桩对应的冻结强度及残余强度与温度的关系可用二次多项式拟合, 三种桩的破坏允许位移与温度的关系均呈现线性规律。研究成果可为寒区结构物抗冻拔病害防治提供参考。     

Interactions between freeze–thaw actions,wind erosion desertification,and permafrost in the Qinghai–Tibet Plateau

The unique natural environment of the Qinghai–Tibet Plateau has led to the development of widespread permafrost and desertification. However, the relationship between desertification and permafrost is rarely explored. Here we study the interaction between desertification and permafrost using a combination of simulations, experiments, and field observations in the Qinghai–Tibet Plateau. Results show the cohesion values of the test samples that experienced 1, 3, and 6 freeze–thaw cycle times decreased by 65.9, 46.0, and 35.5 %, respectively, and the compressive strength of the test samples decreased by 69.6, 39.6, and 34.7 %, respectively, compared to the test samples that did not experience freeze–thaw cycles. The wind erosion rate of the test block eroded by sand-bearing wind was far larger than that by clean wind under the same conditions; the maximum value was 50 times higher than that by clean wind. The wind erosion rate increased with an increasing number of freeze–thaw cycles, water content, and freeze–thaw temperature difference. The ground temperature below the sand layer was decreased, compared to the natural ground surface that without sand layer covering, the drop amplitude of yearly average temperature was roughly maintained at 0.2 °C below the thick sand layer (1.2 m), and the maximum drop of yearly average temperature was 0.7 °C below the thin sand layer (0.1 m). Therefore, with the presence of water, the destruction of surface soil structure caused by repeated and fierce freeze–thaw actions is the main cause of wind erosion desertification in the permafrost region of Qinghai–Tibet Plateau, and sand-bearing wind is the main dynamic force. The development of eolian sand deposits after the desertification emerges. As a result, the properties of the underlying surface are altered. Due to the high reflectivity and poor heat conductivity of the sand layer, the heat exchange of the land–atmosphere system is impeded, causing a drop in the ground temperature of the underlying permafrost that subsequently preserves the permafrost.     

基于数字图像技术的土冻胀试验系统研究

开展土冻胀过程的研究对于认识土冻胀机理具有重要的意义,以往学者多采用建立模型进行数值分析的方法,但是在试验中却不能对模型预测的冷生构造、冰分凝、水分迁移等实时过程进行验证。通过建立包括实时图像采集系统、图像预处理系统以及图像数字化系统为一体的基于数字图像技术的土冻胀试验系统,对青藏粉土的冻胀发展过程和冰分凝过程进行了研究分析。结果显示：试样冻胀后已冻区纵剖面的冷生构造可以分为微薄层状构造区、薄层状构造区以及最暖端厚层状构造区三个部分,其中薄层状构造区下部的冰透镜体的分凝作用是冻胀的主要贡献部分,最暖端厚层状构造区中冰透镜体的分凝作用导致了试样未冻区的固结,试样中冰透镜体的分凝过程与试样的冻胀过程和试后含水率的结果能够相互验证。建立的试验系统可以对土冻胀过程中包括冷生构造发育、冻胀变形、冰分凝等过程进行实时采集和定量分析,为冻胀模型的建立和验证提供了试验方法和技术支撑。     

Late Triassic tectonic inversion in the upper Yangtze Block: Insights from detrital zircon U–Pb geochronology from south‐western Sichuan Basin

The Sichuan Basin and the Songpan‐Ganze terrane, separated by the Longmen Shan fold‐and‐thrust belt (the eastern margin of the Tibetan Plateau), are two main Triassic depositional centres, south of the Qinling‐Dabie orogen. During the Middle–Late Triassic closure of the Paleo‐Tethys Ocean, the Sichuan Basin region, located at the western margin of the Yangtze Block, transitioned from a passive continental margin into a foreland basin. In the meantime, the Songpan‐Granze terrane evolved from a marine turbidite basin into a fold‐and‐thrust belt. To understand if and how the regional sediment routing system adjusted to these tectonic changes, we monitored sediment provenance primarily by using detrital zircon U‐Pb analyses of representative stratigraphic samples from the south‐western edge of the Sichuan Basin. Integration of the results with paleocurrent, sandstone petrology and published detrital zircon data from other parts of the basin identified a marked change in provenance. Early–Middle Triassic samples were dominated by Neoproterozoic (~700–900 Ma) zircons sourced mainly from the northern Kangdian basement, whereas Late Triassic sandstones that contain a more diverse range of zircon ages sourced from the Qinling, Longmen Shan and Songpan‐Ganze terrane. This change reflects a major drainage adjustment in response to the Late Triassic closure of the Paleo‐Tethys Ocean and significant shortening in the Longmen Shan thrust belt and the eastern Songpan‐Ganze terrane. Furthermore, by Late Triassic time, the uplifted northern Kangdian basement had subsided. Considering the eastward paleocurrent and depocenter geometry of the Upper Triassic deposits, subsidence of the northern Kangdian basement probably resulted from eastward shortening and loading of the Songpan‐Ganze terrane over the western margin of the Yangtze Block in response to the Late Triassic collision among Yangtze Block, Yidun arc and Qiangtang terrane along the Ganze‐Litang and Jinshajiang sutures.     

模糊优选神经网络模型及其应用

给出一种能反映宏观决策意图的模糊优选神经网络模型,经用于大连市经济与水资源、环境可持续发展决策,取得了满意结果。     

面实体匹配的集成学习CatBoost方法

现有的面实体多指标几何匹配方法在计算综合相似度和确定最终匹配实体时面临着指标权重和阈值难以科学量化的难题,集成学习算法通过构建并结合多个机器学习器来完成学习任务,在解决分类问题时体现出了较为明显的性能优势。为此,本文提出了一种基于集成学习算法CatBoost的面实体匹配方法,将匹配问题转化为分类问题。选取形状、面积、方向和位置4个几何特征作为模型分类特征;利用过采样与欠采样相结合的混合重采样技术减轻原始训练样本的类别不平衡度;借助贝叶斯优化算法确定CatBoost模型的最优超参数;引入可解释人工智能领域的SHAP解释框架从全局和局部两个角度解释各输入特征对匹配结果的影响。在青藏高原的面状湖泊数据上对本文提出的方法进行了验证,实验结果表明：对模型预测影响最大的特征是位置,然后依次是面积、形状,影响最小的特征是方向。CatBoost匹配方法在实验数据集上的查准率、查全率和F1-score分别达到0.9937、0.9753和0.9844,相比于直接使用样本不均衡的原始样本进行模型训练,分别提高了约5.8%、0.6%和3.3%。与传统的面实体多指标双向匹配方法和逻辑回归、K近邻、决策树、神经网...     

Measuring and modeling the dielectric constant of soil during freezing and thawing processes: an application on silty clay

Acta Geotechnica - Accurately measuring the dielectric constant of soil and modeling it during freezing and thawing processes are important foundations of estimating the physical and chemical...