УДК 620.179.16
ULTRASONIC INSPECTION AND PATTERN RECOGNITION OF WELD DEFECT BASED ON MANUAL ULTRASONIC SCANNING
METHOD
Wengang Hu*, Tie Gang State Key Laboratory of Advanced Welding and Joining, Harbin Institute of Technology, Harbin, China, 150001 * E-mail: huwengang0105@sina.cn
Abstract. As is known to all, the data of conventional manual ultrasonic testing are hardly stored in real-time, and the diagnosis of defect is performed completely by the experienced operator, therefore, the automatic assessment of defect in qualitative and quantitative is unreliable. In this paper, a manual ultrasonic scanning system based on USB-camera was developed. The position of probe in the scanning path could be extracted through the image obtained by the camera. At the same time, echoes reflected from defect were stored, which could offer more information for defect identification. Experiments were carried out by this system. Several welds, containing defects of hole, slag and crack, were inspected, and the images of weld defects were described intuitively by the method of 3D-projection imaging technology. According to the large number of stored echo signals of each defect, signal features were extracted in time domain, frequency domain, time-frequency domain and morphological features were also obtained through the image processing of weld defects. Then these features were optimized by classification criteria based on Euclidean distance. Finally, a back propagation (BP) neural network was adopted and trained by the optimized features to classify the three kinds of flaws. The classification result is satisfying and it will be helpful for weld assessment. Compared to the simple signal features, the fusional features of signal features and morphological features could offer more information of weld defects, thus the recognition rate of weld defect was improved by using these fusional features.
Keywords: ultrasonic testing, weld defect, feature extraction, euclidean distance, BP neural network.
УЛЬТРАЗВУКОВОЙ КОНТРОЛЬ И РАСПОЗНАВАНИЕ ОБРАЗОВ ДЕФЕКТОВ СВАРКИ, ОСНОВАННЫЙ НА РУЧНОМ УЛЬТРАЗВУКОВОМ МЕТОДЕ СКАНИРОВАНИЯ
Венганг Ху, Тай Ганг Ведущая лаборатория сварки и соединений, Технологический институт в Харбине, Китай
Как известно, данные обычного ручного у. з. контроля трудно обрабатывать в режиме реального времени. Диагностика дефектов при этом может быть выполнена только опытным оператором, поэтому актуальна задача автоматической качественной и количественной оценки дефектов. В данной статье рассмотрена система ручного сканирования, которая осуществляет мониторинг положения преобразователя на изделии с помощью видеокамеры. Положение преобразователя в реальном времени определяется по изображению от видеокамеры. Одновременно с этим в компьютере сохраняются эхосиг-налы, полученные при прозвучивании изделия. С помощью системы было проведено несколько экспериментов по прозвучиванию сварных соединений с несплошностями в виде отверстий, шлаковых включений и трещин. Изображения дефектов сварки были описаны интуитивно посредством технологии ЭО-проекций изображений. Каждый дефект был представлен большим количеством эхосигналов, сохраненных в компьютере. Особенности сигналов были извлечены во временном интервале, а также в частотной области. Морфологические особенности несплошностей были получены посредством обработки изображения дефектов сварки. Далее эти особенности были оптимизированы критериями классификации, основанными на эвклидовом расстоянии. Для распознавания типа дефекта использовалась нейтронная сеть, которая способна распознать три вида несплошностей. Получены удовлетворительные результаты классификации и это будет полезно для оценки качества сварки. По сравнению с простыми и известными критериями браковки использование морфологических особенностей эхосигналов дает больше информации о дефектах сварки, что позволяет усовершенствовать критерии качества сварных соединений.
Ключевые слова: у.з. контроль, дефект сварки, определение состояния, нейронная сеть.
1. INTRODUCTION
The inspection of welds is a very important task for assuring safety and reliability in several industrial fields [1], where weld failure can be catastrophic [2]. Nowadays non-destructive testing (NDT) plays a crucial role in ensuring cost effective operation, safety of use and reliability of a wide range of industrial components especially on aerospace, power generation, automotive, railway and petrochemical applications [3]. NDT is based on techniques that rely on the application of physical principles to determine the characteristics of materials and to detect and assess flaws or harmful defects without change of the usefulness or serviceability of said materials [4].
Recently ultrasonic testing technique has become one of the primary methods and is widely applied to detect the inner flaws of material, where one of the main tasks with respect to quantitative non-destructive evaluation is the determination of location, shape, size and orientation of defects [5]. Moreover, the identification of flaw is also an important and difficult task. As is known to all, the flaw information of material is all contained in the ultrasonic echo waveforms, but only a little of data are used presently. Therefore, it is crucial for flaw identification to be full use of the ultrasonic echoes.
In previous investigations, just the single echo signal of each defect was used and processed to identify the flaws. Case et al. [6] analyzed the single waveform in time and frequency domain to obtain 69 features for distinguishing the different waveforms from counterbore, root and crack. Drai et al. [7] extracted features of single echo in time domain, spectral domain and discrete wavelet representation to characterize defects in nature: planar or volumetric. Cau et al. [8] constituted an artificial neural network model to classify position, width, and depth of the defects by the features extracted from the reflected signal. A mass of ultrasonic echo signals from each defect were stored and processed to identify the defect in this article. In addition, morphological features obtained through the image processing of weld defects were also used to identify the defect, so the recognition rate of weld defect was improved by the fusion algorithm. Firstly, several welds, containing defects of hole, slag and crack were inspected by this system. Then the position information and a large number of ultrasonic echo signals from each defect sample were stored in realtime, and the images of weld defects were described intuitively by the method of 3D-projection imaging technology. Then the echo signals were processed in the time domain, frequency domain and time-frequency domain and the signal features were extracted in these domains. The morphological features were also obtained through the image processing of weld defect. Then these features were optimized by classification criteria based on Euclidean distance. Finally, a fusion algorithm of back propagation (BP) neural network was adopted and trained by the optimized features to classify the three kinds of flaws.
2. EXPERIMENTAL METHODOLOGY 2.1. Manual Ultrasonic System
When flaws were detected and the maximum amplitude of ultrasonic echo signals was found, the process of detection must be stopped in conventional manual ultrasonic testing. Then the position of probe was measured manually and the ultrasonic echo signal was shown in oscilloscope. Finally, the location and type of flaws can be obtained by the experience of manipulator. It made the defect inspection very inconvenient. Thus a manual ultrasonic testing system based on video positioning of USB-camera was studied and developed. It
can automatically save the position information of probe and store a series of echo signals from defect sample in real time. This system has a virtue of low cost, simple structure and convenient portability. Its primary hardware includes USB-camera, PC, ultrasonic probe and ultrasonic collect board (Fig. 1).
Fig. 1. Scanning system based on camera.
This system consists of two work modules (Fig. 2). On the one hand, the ultrasonic echo signals were acquired by ultrasonic collect board in real-time. Then signal processing was performed to obtain the buried depth of defect. On
4+-
L/1^4 \TV
Butt Weld
Ultrasonic Probe л К
NT ■■
Control Box
Position Sensor Image Transfer
sr V
Ultrasonic Collect Board
USB Port
Signal
Process Process
PC
3E
ДА \rv
Data Save
ЛЛ
SÍV
Defect View
Image Process
Fig. 2. Schematic diagram of system.
the other hand, the images of ultrasonic probe were acquired by USB-camera controlled by computer. Then image processing was performed to obtain the position of ultrasonic probe.
Finally, 3D-position information of flaw was obtained and described by the method of 3D-imaging technology (Fig. 3). As shown in this figure, the red area is a scan zone which was calibrated before the ultrasonic testing. After the ultrasonic testing, the location, size, distribution and orientation of weld
Fig. 3. Schematic diagram of 3D-projection of image.
defects can be described conveniently, quickly and intuitively by the method of 3D-imaging technology. It could also offer more information for defect identification.
2.2. Feature Extraction of Flaw
Several welds, containing defects of hole, slag and crack were inspected by this system. When a defect was found, the ultrasonic probe was moved to obtain the echoes of flaw with different position and direction. Then the position of ultrasonic probe and a large number of ultrasonic echo signals from this defect were stored in real-time. The maximum amplitude of these echo signals was extracted and 30 % of this value was selected as the threshold value. To avoid the influence of systemic noise and electrical noise, the ultrasonic echo signals whose amplitude is greater than threshold value were used. Obviously, compared with the single echo signal, these echo signals inc
Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.