基于IGWO-SVM的基桩缺陷分类识别  

Classification and Recognition of Foundation Pile Defects Based on IGWO-SVM

作　　者：李天博[1] 任昊 董德勇 LI Tian-bo;REN Hao;DONG De-yong(School of Electrical and Information Engineering,Jiangsu University,Zhenjiang 212013,China;Maintenance Branch of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210000,China)

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013 [2]国网江苏省电力有限公司检修分公司,江苏南京210000

出　　处：《软件导刊》2022年第10期84-90,共7页Software Guide

基　　金：国家自然科学基金青年基金项目(51805218);江苏大学校企合作项目(8421140006)。

摘　　要：针对目前基桩缺陷类型识别精度不高的问题,提出融合时域、频域与小波包能量的特征提取方法,并建立基于改进灰狼优化算法—支持向量机(IGWO-SVM)的分类模型。该模型中的IGWO采用Sobol序列实现了对种群的初始化,并引入莫兰指数提高灰狼优化算法(GWO)的全局收敛能力,最终确定了模型的两个重要参数C为39.21,g为0.16。试验结果表明,该模型对基桩缺陷类型的识别准确率可达96.97%,相较于BP神经网络、KNN两种分类模型分别高出6.36%、12.36%。IGWO-SVM分类模型在基桩缺陷识别方面具有较显著的有效性和优越性。In view of the current situation that the accuracy of pile defect type recognition is not high,a feature extraction method integrating time domain,frequency domain and wavelet packet energy is proposed,and a classification model based on improved grey wolf optimization algorithm support vector machine(IGWO-SVM)is established. In the model,IGWO uses Sobol sequence to initialize the population and introduces Moran index to improve the global convergence ability of grey wolf optimization algorithm(GWO). Finally,two important parameters of the model,C is 39.21 and g is 0.16,are determined. The test results show that the recognition accuracy of this model for pile defect types can reach 96.97%,which is 6.36% and 12.36% higher than BP neural network and KNN classification models respectively. IGWO-SVM classification model has significant effectiveness and superiority in pile defect identification.

关 键 词：灰狼优化算法 支持向量机 小波包分析 超声透射法 基桩缺陷 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]