By using solid-state nuclear detectors, the air radon concentrations at 87 underground working places were measured during spring, summer and winter, respectively. The survey covered 23 cities whose annual radon conce...By using solid-state nuclear detectors, the air radon concentrations at 87 underground working places were measured during spring, summer and winter, respectively. The survey covered 23 cities whose annual radon concentrations range from {14.9} to {246.4} Bq·m{-3} with an overall arithmetic mean value of {106.7} Bq·m{-3}. The average annual effective dose received by people working in these underground working places was {1.6} mSv, hence the lifetime fatality risk was {1.2}×10{-4}. Fujian Province had the highest radon level during the survey. It is better to reduce the radon concentration heavily in summer because of higher radon concentration than in winter.展开更多
With the continuous development of today's construction industry, the construction of underground engineering is becoming more and more common. Especially with the accelerating process of urbanization, more and mo...With the continuous development of today's construction industry, the construction of underground engineering is becoming more and more common. Especially with the accelerating process of urbanization, more and more construction projects and underground projects begin to develop in the direction of ultra-deep and multi-layered. However, under the influence of groundwater, the underground building engineering structure will be subject to greater buoyancy, and this buoyancy will increase with the buried depth of the underground building structure, which will lead to the floating of the bottom plate and wall. In serious cases, there will even be building overturning, which will bring a great security threat to the construction engineering. In order to ensure the anti- floating ability of underground engineering, the anti-floating design and control measures are analyzed in this paper. It is hoped that this analysis can lay a good foundation for the improvement of anti-floating performance of underground construction engineering, so as to ensure the quality and safety of the overall construction engineering.展开更多
固体充填采煤作为一种兼顾资源回收与生态保护的绿色开采方法,其核心环节煤矸分选是井下采选充一体化技术高效运行的前提,而煤矸识别作为实现煤矸精准分选的关键技术,面临着井下复杂工况中特征提取困难、边界定位模糊等挑战。为此,以固...固体充填采煤作为一种兼顾资源回收与生态保护的绿色开采方法,其核心环节煤矸分选是井下采选充一体化技术高效运行的前提,而煤矸识别作为实现煤矸精准分选的关键技术,面临着井下复杂工况中特征提取困难、边界定位模糊等挑战。为此,以固体充填采煤井下煤矸分选为研究背景,提出一种融合多模态大模型(Multimodal Large Language Model,MLLM)的固体充填采煤井下分选煤矸下落瞬态图像识别方法。首先,自主设计并搭建固体充填采煤井下分选煤矸下落瞬态图像采集实验平台,以模拟井下低照度、高粉尘的复杂工况,利用高速相机采集不同工况下的煤矸下落瞬态图像;对采集的煤矸图像进行预处理,运用优化算法提升低照度图像亮度并改善粉尘环境图像质量,同时进行标注和数据扩充,构建用于煤矸识别模型训练和测试的数据集;其次,针对传统SegFormer模型在煤矸图像边界识别中的缺陷,引入高效通道注意力机制(ECA)并优化损失函数,构建ECSegFormer模型;进一步提出将MLLM融合到ECSegFormer模型,形成MLLM-ECSegFormer煤矸识别模型融合架构,利用多模态大模型Qwen-VL(7B)提取煤矸目标中心坐标,通过高斯热图生成空间注意力掩膜,分阶段融入ECSegFormer编码器,实现多模态先验知识与图像特征的动态交互。试验结果表明,融合多模态大模型后,各经典图像识别模型性能均显著提升。其中,MLLM-ECSegFormer的MIoU提升至95.50%、MPA提升至98.92%、准确率提升至98.87%,在识别精度、模型复杂程度和识别效率方面均显著优于经典图像识别模型,且与其他图像识别模型相比,MLLM-ECSegFormer在复杂工况下的边缘识别连续性更强,尤其在粉尘干扰、煤矸形态不规则场景中,对目标区域的分割精度显著优于传统模型。研究成果为煤矸精准识别提供了新方法,提升了固体充填采煤技术的智能化水平,对煤炭资源的绿色智能开采具有重要意义。展开更多
文摘By using solid-state nuclear detectors, the air radon concentrations at 87 underground working places were measured during spring, summer and winter, respectively. The survey covered 23 cities whose annual radon concentrations range from {14.9} to {246.4} Bq·m{-3} with an overall arithmetic mean value of {106.7} Bq·m{-3}. The average annual effective dose received by people working in these underground working places was {1.6} mSv, hence the lifetime fatality risk was {1.2}×10{-4}. Fujian Province had the highest radon level during the survey. It is better to reduce the radon concentration heavily in summer because of higher radon concentration than in winter.
文摘With the continuous development of today's construction industry, the construction of underground engineering is becoming more and more common. Especially with the accelerating process of urbanization, more and more construction projects and underground projects begin to develop in the direction of ultra-deep and multi-layered. However, under the influence of groundwater, the underground building engineering structure will be subject to greater buoyancy, and this buoyancy will increase with the buried depth of the underground building structure, which will lead to the floating of the bottom plate and wall. In serious cases, there will even be building overturning, which will bring a great security threat to the construction engineering. In order to ensure the anti- floating ability of underground engineering, the anti-floating design and control measures are analyzed in this paper. It is hoped that this analysis can lay a good foundation for the improvement of anti-floating performance of underground construction engineering, so as to ensure the quality and safety of the overall construction engineering.
文摘固体充填采煤作为一种兼顾资源回收与生态保护的绿色开采方法,其核心环节煤矸分选是井下采选充一体化技术高效运行的前提,而煤矸识别作为实现煤矸精准分选的关键技术,面临着井下复杂工况中特征提取困难、边界定位模糊等挑战。为此,以固体充填采煤井下煤矸分选为研究背景,提出一种融合多模态大模型(Multimodal Large Language Model,MLLM)的固体充填采煤井下分选煤矸下落瞬态图像识别方法。首先,自主设计并搭建固体充填采煤井下分选煤矸下落瞬态图像采集实验平台,以模拟井下低照度、高粉尘的复杂工况,利用高速相机采集不同工况下的煤矸下落瞬态图像;对采集的煤矸图像进行预处理,运用优化算法提升低照度图像亮度并改善粉尘环境图像质量,同时进行标注和数据扩充,构建用于煤矸识别模型训练和测试的数据集;其次,针对传统SegFormer模型在煤矸图像边界识别中的缺陷,引入高效通道注意力机制(ECA)并优化损失函数,构建ECSegFormer模型;进一步提出将MLLM融合到ECSegFormer模型,形成MLLM-ECSegFormer煤矸识别模型融合架构,利用多模态大模型Qwen-VL(7B)提取煤矸目标中心坐标,通过高斯热图生成空间注意力掩膜,分阶段融入ECSegFormer编码器,实现多模态先验知识与图像特征的动态交互。试验结果表明,融合多模态大模型后,各经典图像识别模型性能均显著提升。其中,MLLM-ECSegFormer的MIoU提升至95.50%、MPA提升至98.92%、准确率提升至98.87%,在识别精度、模型复杂程度和识别效率方面均显著优于经典图像识别模型,且与其他图像识别模型相比,MLLM-ECSegFormer在复杂工况下的边缘识别连续性更强,尤其在粉尘干扰、煤矸形态不规则场景中,对目标区域的分割精度显著优于传统模型。研究成果为煤矸精准识别提供了新方法,提升了固体充填采煤技术的智能化水平,对煤炭资源的绿色智能开采具有重要意义。
文摘基于Rhino建模软件,根据地下矿山U型钢棚支护系统的结构和受力特点,建立了岩巷U型钢棚和煤巷U型钢棚支护系统的分析模型;以某地下矿山支护项目为例,查阅相关文献,结合实际资料以及经验类比法,估算U型钢棚静载加载载荷,选取600 kN载荷作为主要加载载荷,800 k N作为次要加载载荷,以研究破坏过程,考虑极端危险状态下,临空煤帮在模拟当中受侧压为0,结合实测资料及矿压观测经验,侧压系数选取1.2、1.5和1.8这3种情况,运用FLAC^(3D)软件,模拟2种U型钢棚支护在不同采空情况以及不同侧压系数下,工作载荷对应力集中以及形变的影响规律。结果表明:数值模拟结果真实地反映了U型钢棚支护系统的动力特征。随着埋深和相同时步下侧压系数的增加,侧压系数和钢棚所产生的塑性区破坏面积呈增加趋势。基于对U型钢棚铺设深度围压的分析,推荐U型钢棚静载承载压力应低于600 kN。
