NDT.net • Sep 2004 • Vol. 9 No.09
2nd MENDT Proceedings

A New Eddy Current Imaging System for Enhancement of Nondestructive Evaluation

Ibrahim Elshafiey* and Lalita Udpa**
*Electrical Eng. Dept., Cairo University, Fayoum Branch, Fayoum, Egypt
**Electrical Eng. Dept, Michigan State University, Michigan, USA

Abstract

Eddy current techniques are widely used in a variety of nondestructive evaluation (NDE) applications. Eddy current inspection is fast and easy in implementation and it is not associated with radiation hazards. This paper introduces a research conducted to enhance the performance of nondestructive inspection using eddy current imaging. The research resulted in a graphic user interface computer based system, which automates the process of eddy current data acquisition, as well as signal processing, display and classification. Visualization of the eddy current signal is enhanced using several display forms: electronic strip chart, complex impedance plane, A-scan, and C-scan images. Signal interpretation is also enhanced using utilities such as user control of the image colormap and adjustment of the vertical and horizontal A-scan track positions. Automatic classification of the eddy current signal is available based on a neural network approach.

The system was introduced to inspect aircraft wheels. Feedback comments obtained from airline company’s personnel pointed various advantages of using this system. The user can easily locate features of the wheel on the electronic strip chart, while simultaneously following the trajectory signal of these features on the impedance plane window. C-scan images allow visualization of feature size and location on the wheel, in contrast to paper strip chart, which required experience in recognizing the number of flaws and flaw size and location. Automatic retrieval of old data of the wheel and using a display of difference images between the current and previous data, using image registration techniques can enhance the process of monitoring flaw initiation and growth. The use of automatic signal classification using neural network approach, to obtain on-line decisions about signal type, and then mark flaw locations, expedite the inspection process and increase its reliability. The capability of remote control and monitoring of the inspection process is appreciated. Automation of nondestructive inspection would establish a common language for NDE personnel, where eddy current images obtained at one part of the world, could get inspected and analyzed, immediately, thousands of miles away.

Keywords: Eddy Current Imaging, Nondestructive Evaluation, Image Registration, Aging Aircraft

I. INTRODUCTION

Eddy current inspection is widely used in a variety of nondestructive evaluation applications for inspecting electrically conducting materials. Examples of these applications include inspecting oil pipelines, natural gas pipelines, pump impellers, nuclear power plants, and aging aircraft. The principle of eddy current inspection is to bring a coil carrying an alternating current close to the specimen under test, as shown in Figure 1, and observe its impedance. The impedance variation is often represented as a complex trajectory in the complex plane [1,2]. Various forms of eddy current coil probes are designed in accordance with the requirements of each application. The role of computer in eddy current nondestructive evaluation (NDE) has grown significantly with the introduction new systems to automate the process of data acquisition.and signal interpretation. Computer based systems can significantly increase the reliability of NDE by avoiding errors related to human factors, such as operator fatigue, inexperience and inconsistency. This paper introduces a new software analysis package: Wheel Inspection and Signal Analysis (WINSAS), Figure 2, developed to enhance the performance of eddy current inspection of aircraft wheels. WINSAS is designed to drive a data acquisition system for storing horizontal and vertical channels of eddy current signal, and then to perform processing and classification of the signal.

Fig. 1: Alternating Current Coil over a Conducting Specimen Fig. 2: Main Window of WINSAS

II. INSPECTION OF AIRCRAFT WHEELS

Aircraft wheels are subject to excessive stresses during the landing and take off cycles. It is therefore required to periodically inspect the aircraft wheels according to regulation standards. Eddy current inspection of aircraft wheels is the technique of choice used in this type of inspection, since it is effective, fast and easy to implement.

Fig. 3: One Model of ANDEC Eddy Current Aircraft Wheel Testing System.

A commercial eddy current system manufactured by ANDEC is currently in use in several airline companies including: Northwest, USAir, Canadian, Lufthansa, and Delta airlines. The system has been produced in different models ranging from fully analog to a more recent digital model used by Lufthansa. An image of the system is shown in Figure 3. The operation of the ANDEC system involves mounting the wheel on a rotating table and moving the eddy current probe downward while the wheel is kept rotating. Information about different wheel types is stored in the PC. This information is used to control the scanning motors and scan plan. A typical operation is carried out as follows:

    1. Start rotating motion of the table, which carries the wheel.
    2. Position the eddy current probe (or two probes for high and low frequency testing in some systems) until it touches the top of the wheel.
    3. Move probe downward, to obtain a helical scan of the wheel outer surface.
Two probes, namely high and low frequency probes, can be used simultaneously to allow inspection of both the outer and inner surfaces of the wheel. The high frequency probe is operated in the range of 60-80 KHz, while the low frequency is operated in the range of few hundred Hertz. In the case of aluminum, the skin depth, which determines material depth that affects the eddy current measurements, is around 2.6 millimeters at 1 KHz, and 0.26 millimeters at 100 KHz, frequencies.

The eddyscope phase is usually adjusted to make the horizontal channel sensitive to the lift off between the probe and the wheel surface. This is done by moving the probe while adjusting eddyscope phase until the impedance plane trajectory becomes almost horizontal. The lift off signal is thus a source of noise in eddy current measurements, and the vertical channel which is affected less by lift off has a higher signal to noise ratio than the horizontal channel. The eddy current inspection results including both impedance channels (horizontal and vertical), are recorded on a strip chart. However, only the vertical data channel is analyzed and interpreted by the inspector. In contrast, WINSAS offers data acquisition, display and processing of both horizontal and vertical channels allowing decision making to be based on all available data.

III. WINSAS FEATURES

The automated signal analysis system runs on any PC equipped with an Input/Output card, and controls the functions of data acquisition, display, analysis, and storage. The I/O card is connected to the horizontal and vertical BNC connectors of the impedance channels coming out.of the eddyscope, which performs signal filtering and amplification. The I/O card samples both these channels at time intervals determined by an encoder installed on the shaft. Use of the encoder synchronizes the data acquisition process and thus makes the data display and processing invariant to variations in the shaft speed. A typical encoder produces 5000 pulses per revolution. The I/O card is programmed to digitize and store one of the analog inputs to its channels whenever it receives a pulse from the encoder. This means that for a single frequency operation, the number of data samples for each channel will be 2500 samples/ revolution. This number will be divided by 2 in the case of dual frequency mode of operation. The synchronization of the sampling process allows the display of the data as C-scan and A-scan image.

The main window has three main buttons: Scan, Plot, and Image. Each button activates a corresponding window as described next.

III. A. Scan Window:

Fig. 4: WINSAS Scan Window
Scan window (Figure 4) is used to control the I/O card using start, stop and continue buttons. The user first chooses the wheel type using the shown dialog box. The wheel can also be automatically identified by the computer using a bar code reading system. The function of the different buttons are as follows:
Start: Initiates both scanning of the wheel and data acquisition.
Stop: Stops scanning and data acquisition.
Continue: Resumes scanning and data acquisition.
Cancel: Aborts this window.
OK: Ends the window and stores the acquired data.

III. B. Signal display (Plot and Image Windows)

Four basic forms of display are available in WINSAS to present the acquired data:
    1. Electronic strip chart
    2. Impedance plane trajectory
    3. C-Scan Images
    4. A-Scan plots of a selected track across the wheel
The first two display types are given as subwindows of the plot window. The C-scan and A-scan displays are given in the image windows. Descriptions of both of these windows are given next. Plot Window: Figure 5 shows a sample plot window taken from a Boeing 747 nose reference wheel. Description of the features of this wheel is presented next along with its C-scan image. The plot window offers various advantages over the paper strip chart. The user can scroll through and zoom into the data from both channels. The user can thus track the formation of the impedance plane trajectory. Scale factor slider can be used to detect small features in the scan. The strip chart window show the value of both channels corresponding to each index by moving a horizontal line across the chart with the mouse. Figure 5. A Sample Plot Window in WINSAS.

In case of on-line scanning, Plot Window is initiated automatically, and the program updates the display with the incoming data. The user can then save the data in a file, and also obtain a C-Scan image format of the data, by simply clicking on the C-Scan menu. The next section explains options available in the C-Scan image format.

Image Window: To display a C-scan image, the data points are scaled, quantized, decimated and then assigned to image pixels. Three images are produced using the horizontal and vertical channels and also the magnitude values derived from both channels, as illustrated in Figure 6. The Utilities menu in the window is used to control the colormap of the display. The utilities also offer a choice of displaying a single vertical or horizontal A-scan in any of the three images, by moving a horizontal cursor bar. An example is shown in Figures 7 and 8. The wheel has 3 big holes, which were drilled to condemn the wheel after a natural crack was detected. The natural crack signal appears below the hole signals. Other features shown on the wheel are 15 EDM notches made in the bead seat area--where the tire presses on the wheel. In the example shown, the user chooses to obtain a horizontal A-scan and drags the white horizontal line in the magnitude window to the desired position, and the corresponding A-scan values are then displayed in the A-scan subwindow.

Figure 9 shows an Air Bus 310 wheel containing signals form 3 holes, a natural crack, an EDM notch and the bead seat area.

Fig.6 : Image Window in WINSAS Fig.7 : A Sample Image Window of a Boeing 747 wheel. Fig.8 : A-Scan of the impedance magnitude. Track marked by white line is passing through a natural crack and EDM notch. The A-Scan window in bottom shows the signal magnitude for this track Fig.9 : An example of an Air Bus 310 Wheel

III.C. Automatic Signal Classification in WINSAS

In the field of nondestructive evaluation, critical interest in industry is directed towards development of automation signal classification systems. The objective of these systems is to provide accurate and consistent interpretation. The expertise of senior inspectors gained over a long period is invaluable and computer based systems provide the capability for storing such information for future use. WINSAS utility for signal classification is based on an artificial neural network approach [3,4], where a multilayered perceptron neural network is used for signal classification. The structure of the neural network consists of neurons connected together via interconnection weights, which determine the functionality of neural networks. The neural network utilities operate in two phases: training phase, and operation phase. During the training phase, the interconnection weights are adjusted iteratively in such a way that neural network learns to classify correctly all the training input data. In the operation phase, the computer stored weight values are used to classify input test signals.

Fig.10 : Obtaining Neural Network Features from C-Scan Images

Figure 1. Alternating Current Coil over a Conducting Specimen. Figure 2 . Main Window of WINSAS. Figure 3. One Model of ANDEC Eddy Current Aircraft Wheel Testing System. Figure 4. WINSAS Scan Window. Figure 5. A Sample Plot Window in WINSAS. The input vectors are derived by preprocessing of a subimage around the considered feature as indicated in Figure 10. Two-dimensional Fourier transform of the subimage is used to extract the input vector. Using a set of reference wheels with known feature types (natural crack, EDM notch, beadseat, man-made hole, etc.), the network is trained to obtain the weight values, during the installation of WINSAS, using the reference wheels. This is similar to the training of eddy current inspectors. The network can thus capture the experience of the supervising operator automatically. The automatic signal classification capability in WINSAS can be used to trigger a pen or ink-spray markers for marking the flaw locations on the wheel. The inspector can then direct the search towards selected locations.

IV. Additional Features

A variety of functions have been designed as an extension to the basic system upon request from airlines. The main tools are
    1. Automatic Marking Tool (AMT): This tool uses signal classification results of WINSAS to control a plotter and mark the wheel with color-coded marks at crack positions. In this process the wheel, is first scanned with the eddy current probes, and eddy current signal is.acquired. The wheel is then scanned with color pens to mark crack positions with color codes, which indicate the size of the crack and whether it is a surface or subsurface crack.
    2. Electronic Log Tool (ELT): This tool allows automatic retrieval of the old records of a wheel using bar code serial number identification of the wheel. Historical data can be used to subtract unwanted features from the C-scan images, and display clear images of the wheel cracks. Previously obtained crack images would then offer easy detection of new cracks and a better monitoring of crack growth.
    3. Advanced Image Processing Tool (AIMT): Some of the wheel features such as holes or bead seat signal might obscure crack signals. This tool uses neural network analysis and image processing algorithms for enhancing the C-scan image display and reduce noise level and other unwanted feature signals from the display [5]. Image registration capability [6] is also incorporated into this tool, to display difference images between current and older inspection data.

V. Capabilities of WINSAS

Some of the features offered by WINSAS can be summarized as
  • The C-Scan display helps visualize the wheel, by enabling the user to relate flaw signal to its location on the wheel, and also predict flaw size.
  • The colormap and the A-scan features enhance the process of flaw detection. The spatial correlation of the signal helps easy interpretation of the eddy current signal, and estimation of the number of different features on the wheel. On the other hand, using paper strip chart requires training and skill in signal interpretation.
  • The electronic strip chart offers many improvements over the paper strip chart:
    • The electronic data acquisition process offers a larger bandwidth in contrast for a mechanical strip chart and consequently results in better signal to noise ratio.
    • The electronic strip chart allows scrolling back and forth of the data.
    • WINSAS also displays the impedance plane trajectory signal. By dragging the mouse, the user can watch the formation of the impedance plane trajectory of selected parts of data.
  • WINSAS offers a remote control of the inspection process. Inspectors can monitor wheel inspection from their offices. NDE personnel at airline companies can easily exchange data files stored by WINSAS, with each other, and with university researchers. This would help create a larger knowledge database for training new inspectors, and also to help the ongoing research on this subject.
  • The computer would easily retrieve the old inspection file for each wheel. This allows monitoring flaw initiation and growth in each wheel in the fleet.
  • Automatic signal classification can increase the reliability and speed of the inspection process, while keeping the requirements of inspector’s experience moderate.

VI. CONCLUSIONS

This paper presents the WINSAS system designed to work with ANDEC eddy current aircraft wheel inspection system, currently in use in several airline companies. The objective of WINSAS is to enhance the inspection process and improve the eddy current signal interpretation. The system operates on a signal acquired from eddyscope in a synchronized pattern depending on pulses obtained from shaft encoder. Signal processing and display in WINSAS offer various advantages over the currently used system. The user can easily locate a feature on the wheel on the electronic strip chart and also follow the formation of the complex impedance trajectory of this feature. C-scan images allow visualization of feature size and location on the wheel, in contrast to paper strip chart, which requires experience in recognizing the number of flaws and flaw size and location. Automatic retrieval of old data of the wheel and computation of difference images between the current and previous data, using image registration techniques can enhance the process of monitoring flaw initiation and growth. The use of automatic signal classification using neural network approach, to obtain on-line decisions about wheel condition, can expedite the process of wheel inspection and increase its reliability. The capability of remote control and monitoring of the inspection process, using a computer terminal on an inspector’s desk is a novel feature for enhancing the performance of eddy current inspection of aircraft wheels. Such capabilities can be developed for a variety of other aging aircraft inspector problems. For example, signal analysis systems have also been developed for the analysis of eddy current data obtained from rotating probe inspection of bolt-holes, impeller bores and dove tail slot in engine disks.

ACKNOWLEDGMENT

This material is based upon work performed at the FAA Center for Aviation Systems Reliability operated by Iowa State University and supported by the Federal Aviation Administration under Grant No. 95-G-032.

REFERENCES:

  1. S. R. Satish, “Parametric Signal Processing for Eddy Current NDT”, Ph.D. dissertation, Colorado State University, Ft. Collins, Colorado, 1983.
  2. H.L. Libby, “Introduction to Electromagnetic Nondestructive Testing Methods,” Wiley Interscience, New York, 1971.
  3. C.M. Bishop, Neural Networks for Pattern Recognition, Oxford University Press, 1995.
  4. L.V. Fausett, Fundamentals of Neural Networks, Prentice Hall, 1994
  5. Al Bovik, Handbook of Image and Video Processing, Academic Press Limited, 2000.
  6. J.V. Hajnal, , D.L.G. Hill, D. J. Hawkes (editors), Medical Image Registration, CRC Press, 2001...

© NDT.net |Top|