Based on a shallow roadway with weakly cemented soft strata in western China, this paper studies the range and degree of plastic zones in soft strata roadways with weak cementation. Geological radars were used to moni...Based on a shallow roadway with weakly cemented soft strata in western China, this paper studies the range and degree of plastic zones in soft strata roadways with weak cementation. Geological radars were used to monitor the loose range and level of surrounding rocks. A mechanical model of weakly cemented roadway was established, including granular material based on the measured results. The model was then used to determine the plastic zone radium. The predicted results agree well with measured results which provide valuable theoretical references for the analysis of surrounding rock stability and support reinforcing design of weakly cemented roadways. Finally, a combined supporting scheme of whole section bolting and grouting was proposed based on the original supporting scheme. It is proved that this support plan can effectively control the deformation and plastic zone expansion of the roadway surrounding rock and thus ensure the long-term stable and safe mining.展开更多
针对低信噪比条件下雷达信号识别率低,以及分类网络不具备识别样本库新添加信号类型的局限,提出了一种基于深度残差收缩注意力网络的雷达信号识别方法。通过网络将一维雷达信号映射到32维向量空间。网络中的残差连接能有效强化特征的传...针对低信噪比条件下雷达信号识别率低,以及分类网络不具备识别样本库新添加信号类型的局限,提出了一种基于深度残差收缩注意力网络的雷达信号识别方法。通过网络将一维雷达信号映射到32维向量空间。网络中的残差连接能有效强化特征的传播能力,解决网络过深无法训练的问题;注意力机制的引入,不仅构建掩码支路充当主干支路的特征选择器,还能够帮助网络自适应地选择合适的阈值进行软阈值化,从而减少网络中噪声或者冗余信息的影响,提高网络对噪声的鲁棒性。训练过程中排序表损失(ranked list loss,RLL)和分类损失函数共同指导网络训练。RLL能够有效克服传统度量学习损失函数忽略类内特征的问题,分类损失函数能够弥补度量损失优化下对样本整体分布不敏感的问题。实验表明,该方法在提高低信噪比雷达信号识别准确率的同时仍具有识别样本库新添加信号类型的能力。展开更多
基金provided by the National 973 Programs(No.2014CB046905)the National Natural Science Foundation of China(Nos.51274191 and 51404245)+1 种基金the Doctoral Fund of Ministry of Education(No.20130095110018)China Postdoctoral Science Foundation(No.2014M551699)
文摘Based on a shallow roadway with weakly cemented soft strata in western China, this paper studies the range and degree of plastic zones in soft strata roadways with weak cementation. Geological radars were used to monitor the loose range and level of surrounding rocks. A mechanical model of weakly cemented roadway was established, including granular material based on the measured results. The model was then used to determine the plastic zone radium. The predicted results agree well with measured results which provide valuable theoretical references for the analysis of surrounding rock stability and support reinforcing design of weakly cemented roadways. Finally, a combined supporting scheme of whole section bolting and grouting was proposed based on the original supporting scheme. It is proved that this support plan can effectively control the deformation and plastic zone expansion of the roadway surrounding rock and thus ensure the long-term stable and safe mining.
文摘针对低信噪比条件下雷达信号识别率低,以及分类网络不具备识别样本库新添加信号类型的局限,提出了一种基于深度残差收缩注意力网络的雷达信号识别方法。通过网络将一维雷达信号映射到32维向量空间。网络中的残差连接能有效强化特征的传播能力,解决网络过深无法训练的问题;注意力机制的引入,不仅构建掩码支路充当主干支路的特征选择器,还能够帮助网络自适应地选择合适的阈值进行软阈值化,从而减少网络中噪声或者冗余信息的影响,提高网络对噪声的鲁棒性。训练过程中排序表损失(ranked list loss,RLL)和分类损失函数共同指导网络训练。RLL能够有效克服传统度量学习损失函数忽略类内特征的问题,分类损失函数能够弥补度量损失优化下对样本整体分布不敏感的问题。实验表明,该方法在提高低信噪比雷达信号识别准确率的同时仍具有识别样本库新添加信号类型的能力。
