基于核主元分析和支持向量机的结构损伤识别研究

Structural Damage Identification Study Based on Kernel Component Analysis and Support Vector Machine

为了提取有效的损伤特征并提出实用的损伤识别方法,本文利用核主元分析（KPCA）良好的非线性特征提取和支持向量机（SVM）在非线性映射、分类方面的优秀性能,提出了一种基于非线性特征提取的支持向量机损伤识别方法.首先采用粒子群算法（PSO）来优化KPCA的核参数,然后运用优化后的KPCA进行损伤特征提取,最后用SVM进行模式分类并输出识别结果.为了验证所提方法的有效性,通过一个12层钢混框架模型进行损伤识别,并重点研究了KPCA的核参数优化模型及可分性分析、噪声程度、不同特征提取方法、神经网络模型对该方法性能的影响.研究发现：本文所提出的方法不仅能有效地提取损伤特征和降低数据维数,而且具有较高的损伤识别和抗噪能力、泛化能力,且鲁棒性很强.

To study effective damage features and propose practical damage identification methods,this paper utilize the excellent performance of nonlinear feature extraction and support vector machine（ SVM） in nonlinear mapping and classification using kernel principal component analysis（ KPCA） and then propose a support vector machine damage identification method based on nonlinear feature extraction. Firstly the particle swarm optimization（ PSO） is used to optimize kernel parameters of KPCA,next the optimized KPCA is used to extract damage features,finally a SVM classifier is used for identifying damage patterns. To verify the effectiveness of the proposed method,a 12-story reinforced concrete frame model was used for damage identification and the kernel parameter optimization model of KPCA and its separability analysis,noise level,different feature extraction methods and neural network models were studied. The discovery indicates that the method proposed in this paper not only can effectively extract damage characteristics and reduce data dimension,but also has high damage identification and anti noise ability,generalization ability and strong robustness.