摘要
针对超声相控阵无损检测(NDT)中近表面缺陷识别困难的问题,提出一种基于分形理论的近表面缺陷智能识别方法。运用基于线性插值的盒计数维数算法,计算140组超声A扫(A-Scan)信号的盒计数维数,并运用统计的方法详细分析其分布情况。实验结果表明超声A-scan信号具有分形特性,分形理论可应用于A-Scan信号分析;而且有无缺陷信号的盒维数分布区间差异明显,盒维数可作为A-Scan信号的特征识别近表面缺陷。在超声相控阵自动化检测中,运用分形理论能提高近表面缺陷的检出率,减少人为因素引起的漏检。
The near-surface defects are hard to identify in ultrasonic phased array Non-Destructive Testing（ NDT）, thus a new intelligent identification method based on fractal theory was proposed to solve this problem. A box-counting dimension algorithm based on linear interpolation was described to calculate the box-counting dimension of 140 groups of ultrasonic AScan time domain signals. Then the distribution of box-counting dimension was analyzed using the statistical method. The experimental results show that ultrasonic A-Scan signal is obviously fractal and it is effective to analyze the A-Scan signal with the fractal approach. This method has the potential to identify near-surface defects since the values of the box counting dimension of defective signals are different from those of defective signals. As a result, the detection rate of near-surface defects can be improved and the omission rate caused by man-made factors can be reduced in ultrasonic phased array automatic testing.
出处
《计算机应用》
CSCD
北大核心
2014年第11期3365-3368,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(61201039)
关键词
超声相控阵
近表面缺陷
分形理论
盒维数
缺陷识别
ultrasonic phased array
near-surface defect
fractal theory
box-counting dimension
defect identification
