\r王佳俊，钟登华，吴斌平，刘明辉，张宗亮\r
\r
AuthorsHTML:\r王佳俊，钟登华，吴斌平，刘明辉，张宗亮\r
\r
AuthorsListE:\rWang Jiajun，Zhong Denghua，Wu Binping，Liu Minghui，Zhang Zongliang\r
\r
AuthorsHTMLE:\rWang Jiajun，Zhong Denghua，Wu Binping，Liu Minghui，Zhang Zongliang\r
\r
Unit:\r天津大学水利工程仿真与安全国家重点实验室，天津 300350\r
\r
Unit_EngLish:\rState Key Laboratory of Civil Engineering Simulation and Safety，Tianjin University，Tianjin 300350，China\r
\r
Abstract_Chinese:\r土石坝压实质量评价模型的更新对保证其长期高精度评价压实质量具有重要的意义，然而目前对于压实质量模型的更新还缺乏相应的研究．借鉴流数据中概念漂移检测的思想，同时针对碾压施工流数据具有不平衡数据、含有噪声且流速缓慢的特点，本文提出了一种基于概念漂移检测的土石坝压实质量评价模型更新方法．首先提出基于K-means 的下抽样技术处理不平衡数据；其次提出基于增强概率神经网络(enhanced probabilistic neural network，EPNN)和可变窗口技术(variable window technique，VWT)的碾压施工流数据概念漂移检测方法；最后若检测到有概念漂移则进行压实质量评价模型的更新．工程应用表明：基于K-means 的下抽样技术能保证分类器具有较高的一致性；基于EPNN 与VWT 的方法能有效地检测出碾压施工流数据概念漂移；同时以出现概念漂移为条件而更新的压实质量评价模型能够长期高精度评价压实质量．\r
\r
Abstract_English:\rUpdating the compaction quality assessment model of earth-rock dams is important to ensure long-term and high-precision evaluation of the compaction quality. However，there is a lack of research on the update of the compaction quality model.In this study，based on the idea of concept drift detection in stream data，as well as the characteristics of construction stream data such as slow velocity，existing noise data，and unbalanced data，a method of detecting concept drift and updating the compaction quality assessment model is proposed. First，a down sampling technology based on K-means is designed to address the unbalanced data. Second，a concept drift detection method based on enhanced probabilistic neural network(EPNN)and variable window technique(VWT)is proposed. The compaction quality assessment model is updated if a concept drift is detected. The engineering application shows that the down sampling method based on K-means ensures high consistency of classifier. The method based on EPNN and VWT can effectively detect the concept drift of compaction stream data.\r
\r
Keyword_Chinese:压实质量评价模型；概念漂移检测；碾压施工流数据；增强概率神经网络；可变窗口技术；模型更新\r

Keywords_English:compaction quality assessment model；compaction data stream；concept drift detection；enhanced probabilistic neural network；variable window technology；model updating\r


PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6228