Page 138 - 201901

第 38 卷 第 1 期 Vol. 38, No. 1
2019 年 1 月 Journal of Applied Acoustics January, 2019

⋄ 研究报告 ⋄

多层结构中脱粘缺陷的超声检测方法

郑善朴 1† 陆铭慧 1 王俊东 2 罗 艺 1 董俊冬 1

(1 无损检测与光电传感技术及应用国家工程实验室 (南昌航空大学) 南昌 330063)
(2 天津市交通科学研究院 天津 300399)
摘要 固体火箭发动机多层装药结构中的脱粘类型、位置和尺寸决定了其对整体安全性能构成的威胁程度。
该文研究了多层结构中脱粘缺陷的超声检测方法，通过对不同脱粘缺陷超声脉冲回波的特征分析与统计，实
现缺陷的定性、定位和定量。首先，采集含有多类脱粘缺陷的粘接结构的超声脉冲回波信号，分析信号中主能
量波包所代表的声程，提取五种声程的波峰时刻和幅值作为特征值，组建已知脱粘类型训练样本并输入至 BP
神经网络，实现特征值域到类别域的非线性映射，即脱粘类型分类；其次，采用阈值法确定缺陷的界面位置；最
后，提出分段线性插值 -相关性定量法将待检测缺陷的定量结果缩小到 ±2 mm 以内。该文利用 COMSOL 有
限元仿真和实验操作验证了多层粘接结构中脱粘缺陷的定性、定位和定量方法的可行性和可靠性。
关键词 脱粘缺陷，BP 神经网络，超声检测，阈值定位，线性插值
中图法分类号: TB332, TB52.7 文献标识码: A 文章编号: 1000-310X(2019)01-0134-08
DOI: 10.11684/j.issn.1000-310X.2019.01.018

Ultrasonic testing method for debonding defects in multilayer structures

ZHENG Shanpu 1 LU Minghui 1 WANG Jundong 2 LUO Yi 1 DONG Jundong 1

(1 National Engineering Laboratory for Nondestructive Testing and Photoelectric Sensing Technology and Applications,
Nanchang Hangkong University, Nanchang 330063, China)
(2 Tianjin Institute of Transportation Science, Tianjin 300399, China)

Abstract The type, location and size of the debonding defects in a multi-layer charge structure of a solid
rocket motor determines its threat degree to overall safety performance. In this paper, the ultrasonic detection
method of debonding defects in multilayer structure is studied. According to the analysis and statistics of
ultrasonic pulse echoes of different debonding defects, the judge of type, position and area of defects is realized.
Firstly, the ultrasonic pulse echo signals of the bonding structure containing multiple types of debonding defects
are acquired, the sound paths represented by the main energy wave packet in the signal are analyzed, and the
peak moments and amplitudes of the five sound paths are extracted as the eigenvalues to construct the known
debonding defects’ training samples and input the samples into the BP neural network to realize the nonlinear
mapping from the eigenvalue domain to the category domain, equivalent to the classification of debonding
defects. Secondly, the threshold method is used to determine the interface position of the defect. Finally, the
segmented linear interpolation-correlation is proposed to reduce the quantitative result of defects to be detected
to within ±2 mm. In this paper, the feasibility and reliability of the judge method of type, position and area
for debonding defects in multilayer bonded structures are verified by COMSOL finite element simulation and
experimental operations.
Key words Debonding defect, BP neural network, Ultrasonic testing, Threshold positioning, Linear inter-
polation
2018-04-28 收稿; 2018-10-23 定稿
作者简介: 郑善朴 (1994- ), 女, 山东泰安人, 硕士研究生, 研究方向: 超声无损检测。
† 通讯作者 E-mail: xin_toyu@163.com