To address the issue of extreme thermal-induced arching in CRTS II slab tracks due to joint damage,an optimized joint repair model was proposed.First,the formula for calculating the safe temperature rise of the track ...To address the issue of extreme thermal-induced arching in CRTS II slab tracks due to joint damage,an optimized joint repair model was proposed.First,the formula for calculating the safe temperature rise of the track was derived based on the principle of stationary potential energy.Considering interlayer evolution and structural crack propagation,an optimized joint repair model for the track was established and validated.Subsequently,the impact of joint repair on track damage and arch stability under extreme temperatures was studied,and a comprehensive evaluation of the feasibility of joint repair and the evolution of damage after repair was conducted.The results show that after the joint repair,the temperature rise of the initial damage of the track structure can be increased by 11℃.Under the most unfavorable heating load with a superimposed temperature gradient,the maximum stiffness degradation index SDEG in the track structure is reduced by about 81.16%following joint repair.The joint repair process could effectively reduce the deformation of the slab arching under high temperatures,resulting in a reduction of 93.96%in upward arching deformation.After repair,with the damage to interfacing shear strength,the track arch increases by 2.616 mm.展开更多
多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问...多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.展开更多
基金Project(K2022G038)supported by the Science Technology Research and Development Program of China State Railway Group Co.,LtdProject(52178405)supported by the National Natural Science Foundation of China。
文摘To address the issue of extreme thermal-induced arching in CRTS II slab tracks due to joint damage,an optimized joint repair model was proposed.First,the formula for calculating the safe temperature rise of the track was derived based on the principle of stationary potential energy.Considering interlayer evolution and structural crack propagation,an optimized joint repair model for the track was established and validated.Subsequently,the impact of joint repair on track damage and arch stability under extreme temperatures was studied,and a comprehensive evaluation of the feasibility of joint repair and the evolution of damage after repair was conducted.The results show that after the joint repair,the temperature rise of the initial damage of the track structure can be increased by 11℃.Under the most unfavorable heating load with a superimposed temperature gradient,the maximum stiffness degradation index SDEG in the track structure is reduced by about 81.16%following joint repair.The joint repair process could effectively reduce the deformation of the slab arching under high temperatures,resulting in a reduction of 93.96%in upward arching deformation.After repair,with the damage to interfacing shear strength,the track arch increases by 2.616 mm.
文摘多轮对话是人工智能领域的一个重要分支.如何从多轮对话上下文中正确提取与问题相关的核心内容是多轮对话任务的关键问题.现有模型存在辅助任务低效,对全局与局部信息的筛选不够充分,对较短的多轮对话数据学习能力不足等问题.针对上述问题,本文提出了一种局部信息增强且能够感知对话结构的多轮对话模型(Structure-aware Dialogue Model with Fine-grained Local Information,SAFL).针对子任务训练代价大的问题,提出了随机滑动窗口回复预测任务,在多轮对话上下文中的不同位置与大小的窗口内进行回复预测,充分学习细粒度的局部对话语义.针对信息筛选不够充分的问题,提出了重点局部信息蒸馏机制,借助多门控融合方法从全局和局部信息之中蒸馏出重点信息,提升模型融合效果.针对模型对较短的多轮对话上下文学习能力不足的问题,提出阶段信息学习机制,在微调前加强预训练语言模型对短多轮对话数据的领域学习,降低微调阶段中对短多轮对话的学习难度.此外,SAFL设计了对话结构感知任务在对话结构方面进一步加强模型对对话上下文的理解能力.Ubuntu和E-commerce数据集上的实验结果表明,SAFL模型的总体性能优于对比模型.
