基于层次聚类和人工免疫的无监督结构故障分类算法  

Unsupervised Structural Damage Classification Algorithm Based on Hierarchical Clustering and Artificial Immune Pattern Recognition

作　　者：周悦[1] 贾雪松[1] 张东伟[1] 岳林蓓 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2014年第2期374-378,共5页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(61100159);辽宁省自然科学基金项目(201102180);辽宁省教育厅基金项目(L2011093)

摘　　要：目的利用层次聚类与人工免疫模式识别相结合的方法解决无监督结构健康监测中对结构故障识别和分类的问题.方法通过凝聚型层次聚类实现样本数据的分类,通过模仿生物免疫识别和学习机理来训练记忆细胞集合,进而实现对结构故障的识别与分类.结果在benchmark结构模型上的仿真实验测试结果表明在抗原样本数据中采用凝聚型层次聚类和方法能够成功地确定抗原样本数据的模式数目,继而采用人工免疫模式识别算法对实测数据进行模式识别与分类,分类成功率为81%.结论基于层次聚类和人工免疫的无监督结构故障检测与分类算法通过免疫学习和进化产生高质量的记忆细胞,从而有效地识别结构故障模式.This paper incorporates hierarchical clustering with artificial immune pattern recognition method to solve damage patterns recognition and classification problems in unsupervised structure health monitoring. Firstly,sample data,as antigens,are classified by agglomerative hierarchical clustering algorithm. Then,the memory cell sets are trained through imitating the immune recognition and learning mechanism. Finally,the structural data patterns are judged by trained memory cell sets. With the structure model of benchmark proposed by the IASC-ASCE SHM w orking group,the unsupervised structural damage detection and classification algorithm is tested. The simulation results show that agglomerate hierarchical clustering and L method w hich are incorporated can achieve the best number of patterns in the antigen sample data successfully. The measured data can be identified and classified effectively by the proposed algorithm. The unsupervised structural damage classifiction algorithm based on hierarchical clustering and artificial immune pattern recognition can produce the high quality memory cells w hich can identify all kinds of structural damage patterns effectively.

关 键 词：层次聚类 AIPR 无监督 结构健康监测 故障分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]