电力运维安全是备受社会关注的课题。为了避免因运维人员的操作失误而产生严重后果,提出了一种基于长短期记忆网络LSTM(Long Short Term Memory)的、面向电力运维作业的动作识别方法,该方法涵盖了从数据采集、数据处理到分类识别的整个...电力运维安全是备受社会关注的课题。为了避免因运维人员的操作失误而产生严重后果,提出了一种基于长短期记忆网络LSTM(Long Short Term Memory)的、面向电力运维作业的动作识别方法,该方法涵盖了从数据采集、数据处理到分类识别的整个过程,可对人员工作过程中的操作行为进行识别和监督。基于新构建的电力运维作业数据集将方法中用到的深度学习算法LSTM与传统机器学习算法KNN进行仿真对比实验,结果表明,LSTM的表现更佳,在时间窗口为120帧时,LSTM的准确率达到91.32%,比KNN高出约2个百分点。展开更多
This paper aims at investigating the efficacy of different state-of-art damage detection methods when applied to real worm structures subjected to ground motion excitations, for which the literature contributions are,...This paper aims at investigating the efficacy of different state-of-art damage detection methods when applied to real worm structures subjected to ground motion excitations, for which the literature contributions are, at present, still not fully comprehensive. To this purpose the paper analyses two test structures: (1) a four-story scaled steel frame tested on a shake table in a controlled laboratory conditions, and (2) a seven-story reinforced concrete building monitored during the seismic excitations of the 1999 Chi-Chi (Taiwan) Earthquake main shock and numerous fore and afiershocks. Some model based damage approaches and statistics based damage indexes are reviewed. The different methodologies and indexes are, then, applied to the two test structures with the final aim of analysing their performance and validity within the case of a laboratory scaled model and a real world structure subjected to input ground motion.展开更多
文摘电力运维安全是备受社会关注的课题。为了避免因运维人员的操作失误而产生严重后果,提出了一种基于长短期记忆网络LSTM(Long Short Term Memory)的、面向电力运维作业的动作识别方法,该方法涵盖了从数据采集、数据处理到分类识别的整个过程,可对人员工作过程中的操作行为进行识别和监督。基于新构建的电力运维作业数据集将方法中用到的深度学习算法LSTM与传统机器学习算法KNN进行仿真对比实验,结果表明,LSTM的表现更佳,在时间窗口为120帧时,LSTM的准确率达到91.32%,比KNN高出约2个百分点。
文摘This paper aims at investigating the efficacy of different state-of-art damage detection methods when applied to real worm structures subjected to ground motion excitations, for which the literature contributions are, at present, still not fully comprehensive. To this purpose the paper analyses two test structures: (1) a four-story scaled steel frame tested on a shake table in a controlled laboratory conditions, and (2) a seven-story reinforced concrete building monitored during the seismic excitations of the 1999 Chi-Chi (Taiwan) Earthquake main shock and numerous fore and afiershocks. Some model based damage approaches and statistics based damage indexes are reviewed. The different methodologies and indexes are, then, applied to the two test structures with the final aim of analysing their performance and validity within the case of a laboratory scaled model and a real world structure subjected to input ground motion.