基于人工免疫模式识别的结构损伤检测与分类算法

Structural Damage Detection and Classification Algorithm Based on Artificial Immune Pattern Recognition

目的研究人工免疫系统的自治性、主动性、自适应及学习和记忆的仿生机理,来解决结构健康监测中的结构损伤识别和分类问题.方法通过模仿免疫识别和学习机理,提出一种基于Diagonal距离的人工免疫模式识别的结构损伤分类算法,并在IASC-ASCE SHM工作小组所提出的benchmark模型上对结构模式分类进行了实验测试.结果仿真实验表明基于Diagonal距离所得到的分类成功率要略高于Euclidean距离和Mahalanobis距离所得到的分类成功率;基于Diagonal距离研究了克隆率和记忆细胞替代阈值对分类成功率的影响,只要选取合适的参数值,就能获得较高的分类成功率.结论基于Diagonal距离的人工免疫模式识别的结构损伤检测和分类算法通过免疫学习和进化,产生高质量的记忆细胞,从而有效识别各种结构损伤模式.

For the structure health monitoring, this paper studies the structural damage detection and classification problems using the artificial immune system which has the extremely powerful capabilities of autonomy, initiative, adaptive and the bionic principle between learning and memory. An artificial immune pattern recognition and structural detection classification algorithm based on diagonal distance is proposed through imitating the immune recognition and learning mechanism. With the structure of benchmark proposed by the IASC-ASCE SHM working group as the platform, the damage detection and classification are tested. The simulation results show the classification rate based on the diagonal distance is better than Euclidean and Ma- halanobis. The relationship between the classification rate and the parameters which are clone rate and memory cell replacement threshold value is tested based on the diagonal distance, which show that the cloning rate should try to choose suitable parameter values in order to get a better classification success rate. The algorithm based on the immune learning and evolution can produce the high quality memory cells which effectively identify all kinds of structural damage model.