The reflection and transmission of Lamb waves at overlap joints are researched by the numerical method and the experiments. The numerical method is used to simulate the reflection and transmission of Lamb waves at wel...The reflection and transmission of Lamb waves at overlap joints are researched by the numerical method and the experiments. The numerical method is used to simulate the reflection and transmission of Lamb waves at welds with different widths on 2 mm-thick stainless steel plates. The reflection and transmission coefficients are calculated as well. When the welds width is less than 4 mm, a quasi-linear correlation is observed between the reflection coefficient of SO mode and the weld width. In contrast to the weld width, both the gap between two plates and the reinforcement height of weld do not have obvious effect on the reflection coefficient. Based on the results, we propose that the weld width could be rapidly measured through examining the echo amplitude of SO mode. Experiments further confirm the quasilinear correlation and the feasibility of our proposed method.展开更多
The effect of through-thickness reinforcement by composite pins （Z-pins） on the static tensile strength and failure mechanisms of the joints made from ceramic matrix composite （CMC） is investigated. Overlap length...The effect of through-thickness reinforcement by composite pins （Z-pins） on the static tensile strength and failure mechanisms of the joints made from ceramic matrix composite （CMC） is investigated. Overlap length of the single lap joint is 15 mm, 20 mm, 23 mm, 37 mm, and 60 mm, respectively. The experimental results indicate that the final failure modes of the joints can be divided into two groups, （a） the bond-line stops debonding until crack encounters Z-pins; and then the adherends break at the location of Z-pins, when overlap length is more than 20 mm; （b） the bond-line detaches entirely and Z-pins are drawn from adherends, when overlap length is equal to 15 mm. A simple efficient computational approach is presented for analyzing the benefit of through-thickness pins for restricting failure in the single lap joints. Here, the mechanics problem is simplified by representing the effect of the pins by tractions acting on the fracture surfaces of the cracked bond-line. The tractions are prescribed as functions of the crack displacement, which are available in simple forms that summarize the complex deformations to a reasonable accuracy. The resulting model can be used to track the evolution of complete failure mechanisms, for example, bond-line initial delamination and ultimate failure associated with Z-pin pullout, ultimate failure of the adherends. The paper simulates connecting performance of the single lap joints with different Z-pins＇ diameter, spacing and overlap length; the numerical results agree with the experimental results; the numerical results indicate enlarging diameter and decreasing spacing of Z-pins are in favor of improving the connecting performance of the joints. By numerical analysis method, the critical overlap length that lies between two final failure modes is between 18 mm and 19 mm, when Z-pins＇ diameter and spacing are 0.4 mm, 5 mm, respectively.展开更多
中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的...中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的中文电子病历实体关系联合抽取模型AMFRel(adversarial learning and multi-feature fusion for relation triple extraction),提取电子病历的文本和词性特征,得到融合词性信息的编码向量;利用编码向量联合对抗训练产生的扰动生成对抗样本,抽取句子主语;利用信息融合模块丰富文本结构特征,并根据特定的关系信息抽取出相应的宾语,得到医疗文本的三元组。采用CHIP2020关系抽取数据集和糖尿病数据集进行实验验证,结果显示:AMFRel在CHIP2020关系抽取数据集上的Precision为63.922%,Recall为57.279%,F1值为60.418%;在糖尿病数据集上的Precision、Recall和F1值分别为83.914%,67.021%和74.522%,证明了该模型的三元组抽取性能优于其他基线模型。展开更多
在现有的实体关系联合抽取任务中,级联解码的方法直接对三元组进行优化,解决了一部分重叠问题,但是在特定关系下解码的实体,造成实体识别不平衡问题。仅用集合预测的方法可以同时解码出实体和关系,虽然解决了三元组的顺序问题,但也导致...在现有的实体关系联合抽取任务中,级联解码的方法直接对三元组进行优化,解决了一部分重叠问题,但是在特定关系下解码的实体,造成实体识别不平衡问题。仅用集合预测的方法可以同时解码出实体和关系,虽然解决了三元组的顺序问题,但也导致实体之间联系性不强、实体和关系之间交互性差的问题。为了进一步提高联合抽取模型的效果,提出一种融合双阶段解码的实体关系联合抽取模型,包括级联策略下的实体解码与集合预测网络阶段的关系解码。该模型分为三个部分:采用Bert进行编码,有效关注到了上下文的信息;采用级联解码的策略先对实体识别,得到不受关系限制的实体信息,充分识别实体;将融合了实体信息的表示嵌入集合预测网络解码出实体-关系三元组,加强实体与关系的联系。在公开数据集纽约时报(The New York Times,NYT)、WebNLG和ACE2005上的实验结果表明,所提出的模型基本优于基线模型,验证了该模型的有效性。展开更多
基金supported by the National Natural Science Foundation of China(51705470)
文摘The reflection and transmission of Lamb waves at overlap joints are researched by the numerical method and the experiments. The numerical method is used to simulate the reflection and transmission of Lamb waves at welds with different widths on 2 mm-thick stainless steel plates. The reflection and transmission coefficients are calculated as well. When the welds width is less than 4 mm, a quasi-linear correlation is observed between the reflection coefficient of SO mode and the weld width. In contrast to the weld width, both the gap between two plates and the reinforcement height of weld do not have obvious effect on the reflection coefficient. Based on the results, we propose that the weld width could be rapidly measured through examining the echo amplitude of SO mode. Experiments further confirm the quasilinear correlation and the feasibility of our proposed method.
基金supported by the National Natural Science Foundation of China (No. 90405015)the Research Fund forthe Doctoral Program of Higher Education (No. 20030699040).
文摘The effect of through-thickness reinforcement by composite pins （Z-pins） on the static tensile strength and failure mechanisms of the joints made from ceramic matrix composite （CMC） is investigated. Overlap length of the single lap joint is 15 mm, 20 mm, 23 mm, 37 mm, and 60 mm, respectively. The experimental results indicate that the final failure modes of the joints can be divided into two groups, （a） the bond-line stops debonding until crack encounters Z-pins; and then the adherends break at the location of Z-pins, when overlap length is more than 20 mm; （b） the bond-line detaches entirely and Z-pins are drawn from adherends, when overlap length is equal to 15 mm. A simple efficient computational approach is presented for analyzing the benefit of through-thickness pins for restricting failure in the single lap joints. Here, the mechanics problem is simplified by representing the effect of the pins by tractions acting on the fracture surfaces of the cracked bond-line. The tractions are prescribed as functions of the crack displacement, which are available in simple forms that summarize the complex deformations to a reasonable accuracy. The resulting model can be used to track the evolution of complete failure mechanisms, for example, bond-line initial delamination and ultimate failure associated with Z-pin pullout, ultimate failure of the adherends. The paper simulates connecting performance of the single lap joints with different Z-pins＇ diameter, spacing and overlap length; the numerical results agree with the experimental results; the numerical results indicate enlarging diameter and decreasing spacing of Z-pins are in favor of improving the connecting performance of the joints. By numerical analysis method, the critical overlap length that lies between two final failure modes is between 18 mm and 19 mm, when Z-pins＇ diameter and spacing are 0.4 mm, 5 mm, respectively.
文摘中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的中文电子病历实体关系联合抽取模型AMFRel(adversarial learning and multi-feature fusion for relation triple extraction),提取电子病历的文本和词性特征,得到融合词性信息的编码向量;利用编码向量联合对抗训练产生的扰动生成对抗样本,抽取句子主语;利用信息融合模块丰富文本结构特征,并根据特定的关系信息抽取出相应的宾语,得到医疗文本的三元组。采用CHIP2020关系抽取数据集和糖尿病数据集进行实验验证,结果显示:AMFRel在CHIP2020关系抽取数据集上的Precision为63.922%,Recall为57.279%,F1值为60.418%;在糖尿病数据集上的Precision、Recall和F1值分别为83.914%,67.021%和74.522%,证明了该模型的三元组抽取性能优于其他基线模型。
文摘在现有的实体关系联合抽取任务中,级联解码的方法直接对三元组进行优化,解决了一部分重叠问题,但是在特定关系下解码的实体,造成实体识别不平衡问题。仅用集合预测的方法可以同时解码出实体和关系,虽然解决了三元组的顺序问题,但也导致实体之间联系性不强、实体和关系之间交互性差的问题。为了进一步提高联合抽取模型的效果,提出一种融合双阶段解码的实体关系联合抽取模型,包括级联策略下的实体解码与集合预测网络阶段的关系解码。该模型分为三个部分:采用Bert进行编码,有效关注到了上下文的信息;采用级联解码的策略先对实体识别,得到不受关系限制的实体信息,充分识别实体;将融合了实体信息的表示嵌入集合预测网络解码出实体-关系三元组,加强实体与关系的联系。在公开数据集纽约时报(The New York Times,NYT)、WebNLG和ACE2005上的实验结果表明,所提出的模型基本优于基线模型,验证了该模型的有效性。
