A series of suction-controlled triaxial tests was conducted on Nanyang expansive clay to investigate the effects of dry density and suction on dilatancy and strength.The suction of the soil samples was controlled usin...A series of suction-controlled triaxial tests was conducted on Nanyang expansive clay to investigate the effects of dry density and suction on dilatancy and strength.The suction of the soil samples was controlled using a vapour equilibrium technique,with four suction levels ranging from 3.29 MPa to 198.14 MPa,where water retention is dominated by adsorption.The experimental results show that the tested soil exhibits a brittle failure mode under high suction,significantly distinguishing the hydro-mechanical behaviour of the soil at high suction from that observed at low suction.This brittle failure mode significantly increases the contribution of suction to peak strength compared to residual strength,causes the soil to fail before reaching the critical state,a phenomenon not observed in soils under high suction,and results in dilatancy caused by damage to the soil particle aggregates rather than particle rearrangement.The dilatancy data obtained from the triaxial tests reveal that significant soil dilatancy occurs during shear after reaching peak strength,with the maximum dilatancy angle increasing with suction and decreasing with confining pressure.However,the initial dry density has a negligible impact on the soil's dilatancy under high suction levels.This observation further supports that,for unsaturated soils under high suction levels,dilatancy is attributed to damage to soil particle aggregates rather than the rearrangement of soil particles.展开更多
This study investigated the hydraulic and mechanical behaviors of unsaturated coarse-grained railway embankment fill materials(CREFMs)using a novel unsaturated large-scale triaxial apparatus equipped with the axis tra...This study investigated the hydraulic and mechanical behaviors of unsaturated coarse-grained railway embankment fill materials(CREFMs)using a novel unsaturated large-scale triaxial apparatus equipped with the axis translation technique(ATT).Comprehensive soil-water retention and constant-suction triaxial compression tests were conducted to evaluate the effects of initial void ratio,matric suction,and confining pressure on the properties of CREFMs.Key findings reveal a primary suction range of 0 e100 kPa characterized by hysteresis,which intensifies with decreasing density.Notably,the air entry value and residual suction are influenced by void ratio,with higher void ratios leading to decreased air entry values and residual suctions,underscoring the critical role of void ratio in hydraulic behavior.Additionally,the critical state line(CSL)in the bi-logarithmic space of void ratio and mean effective stress shifts towards higher void ratios with increasing matric suction,significantly affecting dilatancy and critical states.Furthermore,the study demonstrated that the mobilized friction angle and modulus properties depend on confining pressure and matric suction.A novel modified dilatancy equation was proposed,which enhances the predictability of CREFMs'responses under variable loading,particularly at high stress ratios defined by the deviatoric stress over the mean effective stress.This research advances the understanding of CREFMs'performance,especially under fluctuating environmental conditions that alter suction levels.展开更多
微表情检测旨在视频中定位幅度微弱、时间短暂的表情区间。其难点在于有效提取面部区域间的动态关联特征和多尺度时序特征,进而精准捕捉面部各区域微小动作之间的关联。针对这些问题,提出了一种融合自适应图注意力和多尺度可变空洞卷积...微表情检测旨在视频中定位幅度微弱、时间短暂的表情区间。其难点在于有效提取面部区域间的动态关联特征和多尺度时序特征,进而精准捕捉面部各区域微小动作之间的关联。针对这些问题,提出了一种融合自适应图注意力和多尺度可变空洞卷积的微表情检测网络(AG-DDNet)。通过引入参数可学习矩阵来实现键值对的特征变换,通过计算面部区域特征向量间的相似度得到动态邻接矩阵,并结合图注意力机制计算区域间权重系数,实现特征的动态融合;采用了多尺度可变空洞卷积模块,通过自适应池化与卷积组合的预测器生成动态感受野,从而实现多尺度的特征提取;引入基于Fisher信息矩阵的自然梯度优化机制,通过Fisher Adam优化器有效捕捉参数空间的几何结构信息,实现学习率的精确自适应调整,从而显著增强了模型对微表情和宏表情的协同检测能力。在微表情检测任务中,该算法与同类代表性算法相比,在CAS(ME)2数据集和SAMM Long Videos数据集上的性能分别提升了54.20%和20.11%。与最新算法相比,两个数据集上的提升幅度分别为38.43%和6.81%,有效证明了该方法在长视频微表情检测任务上的优越性能。展开更多
基金support from the China Scholarship Council(CSC)-University of Technology Sydney joint scholarship and the National Key R&D Program of China(Grant No.2016YFC0800200)is gratefully acknowledged.
文摘A series of suction-controlled triaxial tests was conducted on Nanyang expansive clay to investigate the effects of dry density and suction on dilatancy and strength.The suction of the soil samples was controlled using a vapour equilibrium technique,with four suction levels ranging from 3.29 MPa to 198.14 MPa,where water retention is dominated by adsorption.The experimental results show that the tested soil exhibits a brittle failure mode under high suction,significantly distinguishing the hydro-mechanical behaviour of the soil at high suction from that observed at low suction.This brittle failure mode significantly increases the contribution of suction to peak strength compared to residual strength,causes the soil to fail before reaching the critical state,a phenomenon not observed in soils under high suction,and results in dilatancy caused by damage to the soil particle aggregates rather than particle rearrangement.The dilatancy data obtained from the triaxial tests reveal that significant soil dilatancy occurs during shear after reaching peak strength,with the maximum dilatancy angle increasing with suction and decreasing with confining pressure.However,the initial dry density has a negligible impact on the soil's dilatancy under high suction levels.This observation further supports that,for unsaturated soils under high suction levels,dilatancy is attributed to damage to soil particle aggregates rather than the rearrangement of soil particles.
基金jointly supported by the Science Fund for Distinguished Young Scholars of Hunan Province,China(Grant No.2024JJ2073)the National Natural Science Foundation of China(Grant No.52178443)the Fundamental Research Funds for the Central Universities of Central South University,China(Grant No.2022ZZTS0620)。
文摘This study investigated the hydraulic and mechanical behaviors of unsaturated coarse-grained railway embankment fill materials(CREFMs)using a novel unsaturated large-scale triaxial apparatus equipped with the axis translation technique(ATT).Comprehensive soil-water retention and constant-suction triaxial compression tests were conducted to evaluate the effects of initial void ratio,matric suction,and confining pressure on the properties of CREFMs.Key findings reveal a primary suction range of 0 e100 kPa characterized by hysteresis,which intensifies with decreasing density.Notably,the air entry value and residual suction are influenced by void ratio,with higher void ratios leading to decreased air entry values and residual suctions,underscoring the critical role of void ratio in hydraulic behavior.Additionally,the critical state line(CSL)in the bi-logarithmic space of void ratio and mean effective stress shifts towards higher void ratios with increasing matric suction,significantly affecting dilatancy and critical states.Furthermore,the study demonstrated that the mobilized friction angle and modulus properties depend on confining pressure and matric suction.A novel modified dilatancy equation was proposed,which enhances the predictability of CREFMs'responses under variable loading,particularly at high stress ratios defined by the deviatoric stress over the mean effective stress.This research advances the understanding of CREFMs'performance,especially under fluctuating environmental conditions that alter suction levels.
文摘微表情检测旨在视频中定位幅度微弱、时间短暂的表情区间。其难点在于有效提取面部区域间的动态关联特征和多尺度时序特征,进而精准捕捉面部各区域微小动作之间的关联。针对这些问题,提出了一种融合自适应图注意力和多尺度可变空洞卷积的微表情检测网络(AG-DDNet)。通过引入参数可学习矩阵来实现键值对的特征变换,通过计算面部区域特征向量间的相似度得到动态邻接矩阵,并结合图注意力机制计算区域间权重系数,实现特征的动态融合;采用了多尺度可变空洞卷积模块,通过自适应池化与卷积组合的预测器生成动态感受野,从而实现多尺度的特征提取;引入基于Fisher信息矩阵的自然梯度优化机制,通过Fisher Adam优化器有效捕捉参数空间的几何结构信息,实现学习率的精确自适应调整,从而显著增强了模型对微表情和宏表情的协同检测能力。在微表情检测任务中,该算法与同类代表性算法相比,在CAS(ME)2数据集和SAMM Long Videos数据集上的性能分别提升了54.20%和20.11%。与最新算法相比,两个数据集上的提升幅度分别为38.43%和6.81%,有效证明了该方法在长视频微表情检测任务上的优越性能。
