脱空和不密实是隧道衬砌最常见的两种病害。在这两种病害长期作用下会导致隧道出现破裂、渗漏水、钢筋锈蚀,最终造成隧道塌方等问题,严重威胁行车安全。采用探地雷达对隧道进行无损探测是发现这些病害或缺陷的常见方式,但大量雷达数据...脱空和不密实是隧道衬砌最常见的两种病害。在这两种病害长期作用下会导致隧道出现破裂、渗漏水、钢筋锈蚀,最终造成隧道塌方等问题,严重威胁行车安全。采用探地雷达对隧道进行无损探测是发现这些病害或缺陷的常见方式,但大量雷达数据的人工识别存在着工作量大、效率低、强烈依赖人员的专业素养等问题。本文提出一种基于深度学习的隧道衬砌缺陷的自动检测方法——自监督多尺度池化区域卷积神经网络方法(Self-monitoring Multi-scale ROI Align Region Convolutional Neural Network,SMR-RCNN),以提高缺陷识别的效率,并减少主观因素的影响。在雷达探测隧道衬砌的实践中,数据量巨大,但缺陷样本却很少,这对训练神经网络是一个相当大的挑战。为此,设计了一种数据增强的方法来增加缺陷的样本数量,且使用一种自监督对比学习的网络模型来提取雷达数据的特征,然后将其迁移到改进后的Faster-RCNN网络模型中;最后,使用有标签的样本对改进的Faster-RCNN网络进行细调训练。实验结果表明,相较于传统的Faster-RCNN方法,本文提出的算法增强了神经网络对脱空和不密实两类缺陷的自动识别能力,在检测精度上得到了显著提高,mAP值提升了12%。展开更多
Delays in the construction of nuclear reactors due to licensing issues have been a problem across the world, affecting projects in Finland, France, and the United States. Small Modular Reactors (SMRs) emerge as a tran...Delays in the construction of nuclear reactors due to licensing issues have been a problem across the world, affecting projects in Finland, France, and the United States. Small Modular Reactors (SMRs) emerge as a transition between Generations III+ and IV in order to make nuclear energy more competitive with other energy sources, including renewables. In this study, the SMR NuScale, one of the most promising projects today, is investigated for its conversion into a U-233-producing reactor through the Radkowsky seed-blanket fuel element concept, applied in the Shippingport reactor, in a parametric study. Initially, a validation of the reference reactor (NuScale) was carried out with data from technical documents and papers, thus demonstrating the agreement of the computational model carried out with the SERPENT code. Then, a parametric study is carried out to define the area of the seed and blanket region, proportions of enrichment and pitch length. Finally, a comparison is made between the production of U-233, TRU reduction, burn-up extension and neutronic and thermohydraulic safety parameters. This study demonstrates an improvement in the conversion factor and a considerable reduction in the production of TRU, in addition to the production of U-233 with a low proportion of other uranium isotopes that can lead to the beginning of the thorium cycle with already consolidated technologies.展开更多
文摘脱空和不密实是隧道衬砌最常见的两种病害。在这两种病害长期作用下会导致隧道出现破裂、渗漏水、钢筋锈蚀,最终造成隧道塌方等问题,严重威胁行车安全。采用探地雷达对隧道进行无损探测是发现这些病害或缺陷的常见方式,但大量雷达数据的人工识别存在着工作量大、效率低、强烈依赖人员的专业素养等问题。本文提出一种基于深度学习的隧道衬砌缺陷的自动检测方法——自监督多尺度池化区域卷积神经网络方法(Self-monitoring Multi-scale ROI Align Region Convolutional Neural Network,SMR-RCNN),以提高缺陷识别的效率,并减少主观因素的影响。在雷达探测隧道衬砌的实践中,数据量巨大,但缺陷样本却很少,这对训练神经网络是一个相当大的挑战。为此,设计了一种数据增强的方法来增加缺陷的样本数量,且使用一种自监督对比学习的网络模型来提取雷达数据的特征,然后将其迁移到改进后的Faster-RCNN网络模型中;最后,使用有标签的样本对改进的Faster-RCNN网络进行细调训练。实验结果表明,相较于传统的Faster-RCNN方法,本文提出的算法增强了神经网络对脱空和不密实两类缺陷的自动识别能力,在检测精度上得到了显著提高,mAP值提升了12%。
文摘Delays in the construction of nuclear reactors due to licensing issues have been a problem across the world, affecting projects in Finland, France, and the United States. Small Modular Reactors (SMRs) emerge as a transition between Generations III+ and IV in order to make nuclear energy more competitive with other energy sources, including renewables. In this study, the SMR NuScale, one of the most promising projects today, is investigated for its conversion into a U-233-producing reactor through the Radkowsky seed-blanket fuel element concept, applied in the Shippingport reactor, in a parametric study. Initially, a validation of the reference reactor (NuScale) was carried out with data from technical documents and papers, thus demonstrating the agreement of the computational model carried out with the SERPENT code. Then, a parametric study is carried out to define the area of the seed and blanket region, proportions of enrichment and pitch length. Finally, a comparison is made between the production of U-233, TRU reduction, burn-up extension and neutronic and thermohydraulic safety parameters. This study demonstrates an improvement in the conversion factor and a considerable reduction in the production of TRU, in addition to the production of U-233 with a low proportion of other uranium isotopes that can lead to the beginning of the thorium cycle with already consolidated technologies.
