为了抑制连续体结构拓扑优化结果中的灰度单元,在实体各向同性材料惩罚密度法(Solid isotropic microstructures with penalization,SIMP)的基础上提出一种双重SIMP方法。首先利用含有灵敏度过滤的SIMP方法对设计变量进行迭代和优化,将...为了抑制连续体结构拓扑优化结果中的灰度单元,在实体各向同性材料惩罚密度法(Solid isotropic microstructures with penalization,SIMP)的基础上提出一种双重SIMP方法。首先利用含有灵敏度过滤的SIMP方法对设计变量进行迭代和优化,将优化结果作为中间变量,然后用不含有灵敏度过滤的SIMP方法对中间变量进行优化迭代,得到最终优化结果。以简支梁的结构柔顺度最小为例,将双重SIMP方法用于柔性结构的优化设计中并与标准SIMP算法相比;结果表明,双重SIMP方法得到清晰的拓扑分布结构,有效抑制了灰度单元,并且双重SIMP方法可以得到更小的结构柔顺度,证明了该方法的有效性;双重SIMP算法的灰度单元对惩罚因子的依赖性很小,可以选用较小的惩罚因子;针对不同的网格划分模型,双重SIMP算法得到的优化目标和灰度单元数更稳定,体现出了很好的网格依赖性。说明该方法具有很好的适用性和稳定性。展开更多
近年来,材料基因组计划(material genome initiative,MGI)已成为全球热点。数据源的缺乏和数据存储方法的不规范导致材料领域缺乏可用于机器学习模型训练的结构化数据,这成为了研究人员对材料性能进行预测的瓶颈。随着材料科学的不断发...近年来,材料基因组计划(material genome initiative,MGI)已成为全球热点。数据源的缺乏和数据存储方法的不规范导致材料领域缺乏可用于机器学习模型训练的结构化数据,这成为了研究人员对材料性能进行预测的瓶颈。随着材料科学的不断发展,材料领域文本中包含的大量信息,已成为材料领域研究人员应用机器学习的主要数据来源,如何获取大量有效的材料数据是成为现阶段的一项具有挑战意义的工作。本论文采用自然语言处理技术从铝硅合金材料文献中获取有效数据。命名实体识别是自然语言处理中一项重要的子任务,旨在识别文本中具有特定意义的实体。具体研究方法是从材料科学文献中选择五类实体,手工标注构建了铝硅合金材料实体识别数据集,包括5347个句子,2835个实体。为了减少自然语言处理任务对标注语料的依赖,利用迁移学习将语言模型预训练后应用到特定领域任务中;结合实体特征,基于ALBERT(A Lite BERT)预训练语言模型与条件随机场(conditional random fields,CRF)进行联合建模,并将预训练模型基于主动学习应用于合金材料实体识别。在基于少量标注的训练集样本下,结合主动学习,使得模型的F1值、精确率、召回率分别提高了0.61%,2.68%,0.29%。实验证明结合预训练和主动学习能够进一步减少实体识别任务模型对标注数据的依赖及人工标注的成本。论文研究成果可解决材料数据孤岛问题,改善材料基因组机器学习一直处于小规模数据集的困境,将促进铝硅合金的研发进程,为材料基因组新材料设计提供科学依据。展开更多
Numerical exercises are presented on the thermally induced motion of internally heated beams under various heat transfer and structural boundary conditions. The dynamic displacement and dynamic thermal moment of the b...Numerical exercises are presented on the thermally induced motion of internally heated beams under various heat transfer and structural boundary conditions. The dynamic displacement and dynamic thermal moment of the beam are analyzed taking into consideration that the temperature gradient is independent as well as dependent on the beam displacement. The effect of length to thickness ratio of the beam on the thermally induced vibration is also investigated. The type of boundary conditions has its influence on the magnitude of dynamic displacement and dynamic thermal moment. A sustained thermally induced motion is observed with progress of time when the temperature gradient being evaluated is dependent on the forced convection generated due to beam motion. A finite element method (FEM) is used to solve the structural equation of motion as well as the heat transfer equation.展开更多
Underground contamination by non-aqueous phase liquids (NAPLs) becomes increasingly serious. Rapid and reliable detection of contaminated zone and degree is the first step to site remediation. In this paper, diesel an...Underground contamination by non-aqueous phase liquids (NAPLs) becomes increasingly serious. Rapid and reliable detection of contaminated zone and degree is the first step to site remediation. In this paper, diesel and fine sand are used as experiment materials to investigate the applicability of using time-domain reflectometry (TDR) to detect LNAPLs contamination. The major work includes: measurement of dielectric constant and electrical conductivity for the diesel-water-air-sand mixtures; measurement of reflection waveform and dielectric constant for specimens with a diesel contaminated layer being sandwiched in sand. The experimental results show the followings: A significant decrease in both dielectric constant and electrical conductivity is observed for the diesel-water-air-sand mixtures when diesel displaces the pore water, and the content of diesel can be calculated by the model; insignificant change in dielectric properties is measured when diesel only displaces the pore gas; when the diesel contaminated sand is sandwiched between two saturated sand layers, the interfaces of the diesel contaminated layer can be identified by analyzing the reflection waveform; for field application, TDR method is valid for the case that LNAPLs seep into saturated sand layer, and the applicability of TDR method in vadose zone depends on the initial saturation of the sand layer. The findings obtained in this paper provide a guidance for the use of TDR for the field investigation of NAPLs contaminated site.展开更多
Magnesium alloys show promise for application in orthopedic implants, owing to their biodegradability and biocompatibility. In the present study, ternary Mg-(3.5,6.5 wt%) Li-(0.2, 0.5, 1.0 wt%) Ca alloys were develope...Magnesium alloys show promise for application in orthopedic implants, owing to their biodegradability and biocompatibility. In the present study, ternary Mg-(3.5,6.5 wt%) Li-(0.2, 0.5, 1.0 wt%) Ca alloys were developed. Their mechanical strength, corrosion behavior and cytocompatibility were studied. These alloys showed improved mechanical strength than pure Mg and exhibited suitable corrosion resistance. Furthermore, Mg-3.5Li-0.5Ca alloys with the best in vitro performance were implanted intramedullary into the femurs of mice for 2 and 8 weeks. In vivo results revealed a significant increase in cortical bone thickness around the Mg-3.5Li-0.5Ca alloy rods, without causing any adverse effects.Western blotting and immunofluorescence staining of β-catenin illustrated that Mg-3.5Li-0.5Ca alloy extracts induced osteogenic differentiation of human bone marrow-derived mesenchymal stem cells(hBMMSCs) through the canonical Wnt/β-catenin pathway. Our studies demonstrate that Mg-3.5Li-0.5Ca alloys hold much promise as candidates for the facilitation of bone implant application.展开更多
文摘为了抑制连续体结构拓扑优化结果中的灰度单元,在实体各向同性材料惩罚密度法(Solid isotropic microstructures with penalization,SIMP)的基础上提出一种双重SIMP方法。首先利用含有灵敏度过滤的SIMP方法对设计变量进行迭代和优化,将优化结果作为中间变量,然后用不含有灵敏度过滤的SIMP方法对中间变量进行优化迭代,得到最终优化结果。以简支梁的结构柔顺度最小为例,将双重SIMP方法用于柔性结构的优化设计中并与标准SIMP算法相比;结果表明,双重SIMP方法得到清晰的拓扑分布结构,有效抑制了灰度单元,并且双重SIMP方法可以得到更小的结构柔顺度,证明了该方法的有效性;双重SIMP算法的灰度单元对惩罚因子的依赖性很小,可以选用较小的惩罚因子;针对不同的网格划分模型,双重SIMP算法得到的优化目标和灰度单元数更稳定,体现出了很好的网格依赖性。说明该方法具有很好的适用性和稳定性。
文摘近年来,材料基因组计划(material genome initiative,MGI)已成为全球热点。数据源的缺乏和数据存储方法的不规范导致材料领域缺乏可用于机器学习模型训练的结构化数据,这成为了研究人员对材料性能进行预测的瓶颈。随着材料科学的不断发展,材料领域文本中包含的大量信息,已成为材料领域研究人员应用机器学习的主要数据来源,如何获取大量有效的材料数据是成为现阶段的一项具有挑战意义的工作。本论文采用自然语言处理技术从铝硅合金材料文献中获取有效数据。命名实体识别是自然语言处理中一项重要的子任务,旨在识别文本中具有特定意义的实体。具体研究方法是从材料科学文献中选择五类实体,手工标注构建了铝硅合金材料实体识别数据集,包括5347个句子,2835个实体。为了减少自然语言处理任务对标注语料的依赖,利用迁移学习将语言模型预训练后应用到特定领域任务中;结合实体特征,基于ALBERT(A Lite BERT)预训练语言模型与条件随机场(conditional random fields,CRF)进行联合建模,并将预训练模型基于主动学习应用于合金材料实体识别。在基于少量标注的训练集样本下,结合主动学习,使得模型的F1值、精确率、召回率分别提高了0.61%,2.68%,0.29%。实验证明结合预训练和主动学习能够进一步减少实体识别任务模型对标注数据的依赖及人工标注的成本。论文研究成果可解决材料数据孤岛问题,改善材料基因组机器学习一直处于小规模数据集的困境,将促进铝硅合金的研发进程,为材料基因组新材料设计提供科学依据。
文摘Numerical exercises are presented on the thermally induced motion of internally heated beams under various heat transfer and structural boundary conditions. The dynamic displacement and dynamic thermal moment of the beam are analyzed taking into consideration that the temperature gradient is independent as well as dependent on the beam displacement. The effect of length to thickness ratio of the beam on the thermally induced vibration is also investigated. The type of boundary conditions has its influence on the magnitude of dynamic displacement and dynamic thermal moment. A sustained thermally induced motion is observed with progress of time when the temperature gradient being evaluated is dependent on the forced convection generated due to beam motion. A finite element method (FEM) is used to solve the structural equation of motion as well as the heat transfer equation.
基金supported by the National High Technology Research and Development Program of China ("863" Project) (Grant No.2012AA062601)the National Natural Science Foundation of Major International Cooperation Projects (Grant No. 51010008)
文摘Underground contamination by non-aqueous phase liquids (NAPLs) becomes increasingly serious. Rapid and reliable detection of contaminated zone and degree is the first step to site remediation. In this paper, diesel and fine sand are used as experiment materials to investigate the applicability of using time-domain reflectometry (TDR) to detect LNAPLs contamination. The major work includes: measurement of dielectric constant and electrical conductivity for the diesel-water-air-sand mixtures; measurement of reflection waveform and dielectric constant for specimens with a diesel contaminated layer being sandwiched in sand. The experimental results show the followings: A significant decrease in both dielectric constant and electrical conductivity is observed for the diesel-water-air-sand mixtures when diesel displaces the pore water, and the content of diesel can be calculated by the model; insignificant change in dielectric properties is measured when diesel only displaces the pore gas; when the diesel contaminated sand is sandwiched between two saturated sand layers, the interfaces of the diesel contaminated layer can be identified by analyzing the reflection waveform; for field application, TDR method is valid for the case that LNAPLs seep into saturated sand layer, and the applicability of TDR method in vadose zone depends on the initial saturation of the sand layer. The findings obtained in this paper provide a guidance for the use of TDR for the field investigation of NAPLs contaminated site.
基金supported by the National Key Research and Development Program of China (2016YFC1102900 and 2016YFC1102402)the National Natural Science Foundation of China (81771039, 81470769 and 51431002)+2 种基金the Project for Culturing Leading Talents in Scientific and Technological Innovation of Beijing, China (Z171100001117169)the NSFC-RFBR Cooperation Project (51611130054)the NSFC/RGC Joint Research Scheme (51361165101 and 5161101031)
文摘Magnesium alloys show promise for application in orthopedic implants, owing to their biodegradability and biocompatibility. In the present study, ternary Mg-(3.5,6.5 wt%) Li-(0.2, 0.5, 1.0 wt%) Ca alloys were developed. Their mechanical strength, corrosion behavior and cytocompatibility were studied. These alloys showed improved mechanical strength than pure Mg and exhibited suitable corrosion resistance. Furthermore, Mg-3.5Li-0.5Ca alloys with the best in vitro performance were implanted intramedullary into the femurs of mice for 2 and 8 weeks. In vivo results revealed a significant increase in cortical bone thickness around the Mg-3.5Li-0.5Ca alloy rods, without causing any adverse effects.Western blotting and immunofluorescence staining of β-catenin illustrated that Mg-3.5Li-0.5Ca alloy extracts induced osteogenic differentiation of human bone marrow-derived mesenchymal stem cells(hBMMSCs) through the canonical Wnt/β-catenin pathway. Our studies demonstrate that Mg-3.5Li-0.5Ca alloys hold much promise as candidates for the facilitation of bone implant application.
