摘要
裂纹缺陷定量反演是管道远场涡流检测中的一个难点问题。首先针对远场涡流检测信号的特点,提出了一种基于最小均方误差准则的波形逼近技术,用以高精度提取检测信号的波形特征。接着通过建立非线性多项式反演模型、基于一维搜索寻优的BP神经网络反演模型和基于粒子群搜索寻优的支持向量机反演模型来实现由检测信号波形特征到裂纹缺陷尺寸的定量反演。仿真结果表明,支持向量机反演模型计算相对误差低于5%,具有较强的抗干扰能力,适合作为裂纹缺陷定量反演的有力工具。最后,通过实验也验证了这一结果。
Quantitative inverse of the pipe crack flaw is a difficult problem to be solved in pipe remote field eddy current （RFEC） inspection. According to the distinctive features of the RFEC detecting signals, a wave-approximating approach based on the minimum mean square error criterion is proposed in this paper, which is used to extract the wave characteristics of the detection signals with high accuracy. Through establishing the nonlinear polynomial inverse model, the BP neural network inverse model based on one-dimensional optimization and the support vector machine （SVM） inverse model based on particle swarm optimization （ PSO）, the quantitative inverse from the wave features of the detecting signals to the crack flaw size is realized. Simulation results show that the SVM inverse model can achieve a relative error of 5% , has strong anti-disturbance capacity, and is suitable to be an effective tool for the quantitative inverse of the pipeline crack defects. Finally, the experiments also verify the above results.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2013年第8期1681-1689,共9页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61201131)
"十二五"国家科技重大专项(2011ZX05020-006-005)
中央高校基本科研业务费(ZYGX2012J092)资助项目
关键词
裂纹缺陷
定量反演
波形逼近
粒子群优化
支持向量机
crack flaw
quantitative inverse
wave-approximating
particle swarm optimization （ PSO ）
support vector machine （ SVM ）
