论文部分内容阅读
分析了均匀各向同性介质的超声检测中，耦合因素和仪器因素造成的脉冲回波幅度的差异，提出了误差校正方法。以校正后的缺陷回波和底面反射波的峰－峰值为特征量，利用人工神经网络进行缺陷类型识别和大小评价。为模拟自然缺陷的二基本要素——光滑曲面和带棱边的平面，用有机玻璃制作了代表性的横穿孔和平底孔缺陷样品共１８个．对１８个缺陷样品的缺陷回波和底面反射波的峰－峰值测量了四次，并进行的校正．用人工神经网络对这组缺陷样品进行的处理结果表明：（１）设定的缺陷类型全部准确识别。（２）估计缺陷大小与标称孔径吻合较好。最后，对测量误差和缺陷大小估计误差进行了分析。
The difference of pulse echo amplitudes caused by the coupling factors and instrumental factors in the ultrasonic testing of homogeneous isotropic media is analyzed. An error correction method is proposed. Based on the peak-peak value of the corrected defect echo and the bottom reflected wave, the defect type identification and size evaluation were carried out by artificial neural network. In order to simulate the two basic elements of natural defects - smooth surface and edge plane, a total of 18 samples of transversal and flat-bottomed defects were made of organic glass. The peak-to-peak values of the flaw echoes and the bottom echoes of 18 defect samples were measured four times and corrected. The artificial neural network processing results of this group of defective samples show that: (1) the set of defect types are all accurately identified. (2) It is better to estimate the defect size and the nominal aperture. Finally, the measurement error and defect size estimation error were analyzed.