Abstract: In order to accurately detect and evaluate the corrosion of steel bars, 40 groups of corrosion grading evaluation tests of HRB400 steel bar with different diameters were carried out, which was based on the spontaneous magnetic flux leakage detection technology. The relationship between different corrosion degrees α and magnetic induction intensity By, Bz was analyzed. The quantitative relationship between the corrosion degree α and the magnetic characteristic indexes β and γ of the steel bar specimens was mainly explored, and the evaluation system between the corrosion grade of the steel bars and the magnetic characteristic indexes β and γ was established. And the credibility of the proposed evaluation system was verified. The research results show that when the lift-off height is constant, the magnetic characteristic index values β and γ have a strong correlation with the corrosion degree α of the steel bar specimen. And with the increase of the rust degree α, the magnetic characteristic index values β and γ show an overall increasing trend, but both have a certain degree of dispersion. Based on this, the steel corrosion degree α is divided into 4 grades. Through the establishment of different evaluation systems, it is found that the comprehensive discrimination accuracy of the magnetic characteristic indexes β and γ reaches 92.5%. The support vector machine (SVM) statistical learning algorithm is introduced to evaluate the proposed evaluation system, and its accuracy rate reaches 90%, which proves the credibility of the proposed evaluation system.

Key words: bridge engineering; steel bar; corrosion degree; magnetic characteristic index; evaluation system; SVM

摘要： 为了准确地检测和评定钢筋的腐蚀情况，基于自发漏磁检测技术开展了40组不同直径的HRB400钢筋锈蚀分级评估试验，分析了钢筋不同锈蚀度α和磁感应强度By、Bz之间的关系，重点探究了钢筋试件的锈蚀度α和磁特征指标β、γ之间的定量关系，建立钢筋锈蚀等级和磁特征指标β、γ之间的评价体系，并对该评价体系的可信度进行了验证。研究结果表明：当提离高度一定时，磁特征指标值β、γ和钢筋试件的锈蚀度α具有较强的相关性，且随着锈蚀度α的增加，磁特征指标值β、γ整体呈递增趋势，但都具有一定离散性。基于此，将钢筋锈蚀度α划分为4个等级，通过建立不同的评价体系，发现磁特征指标β、γ综合判别准确率达到了92.5%。引入支持向量机（SVM）统计学习算法，对该评价体系进行评估，其准确率达到了90%，证明了该评价体系具有可信度。

关键词: 桥梁工程；钢筋；锈蚀度；磁特征指标；评价体系；SVM