摘要
在桥梁结构健康监测与状态评估过程中所获得的桥梁结构数据库常存在间断性异常或缺损,且不同样本分类数据不均匀,难以在信息缺失、数据分布失衡的情况下完成对桥梁结构健康的监测与状态评估.针对这一问题,在改进相似性度量函数的SOM聚类算法和非平衡集成迁移学习算法的基础上,提出了一种改进的迁移学习模型.通过对实际监测数据的分析,该迁移学习模型的分类精度随着目标数据集所占比例的不断增加而提高,验证了该模型的有效性和科学性.
The examination of bridge structural data obtained in the bridge structural health monitoring and condition assessment process had the problem of intermittent abnormalities or defects in the past. However, the classification of different samples of data is seen to be uneven, thus, making it difficult to complete structural health monitoring and condition assessment of the bridge under the condition of the absence of information and data distribution imbal- ance. In order to solve the problem mentioned above, this paper proposes an improved transfer learning model based on self-organizing map （SOM） clustering algorithm to improve the similarity measure function and unbal- anced integration transfer learning algorithm. According to the analysis of actual monitoring data, the classification accuracy of the proposed transfer learning model increased as the increasing of the proportion of the target data set, validating the efficiency and scientificity of the proposed model.
出处
《智能系统学报》
CSCD
北大核心
2013年第1期46-51,共6页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金资助项目(61070182)
北京市组织部优秀人才培养资助项目(2010D005003000008)
北京市学科建设项目(PXM2012_014213_0000_74)
关键词
非平衡集成迁移学习算法
SOM算法
迁移学习模型
桥梁结构健康监测
unbalanced integrated transfer learning algorithm
self-organizing map algorithm
transfer learning el
bridge structural health monitoring
