· Table of Contents
· Methods & Instrumentation
Development of Flexible Ultrasonic Waveform Analysis for NDEMatthew Ibrahim
Defence Science and Technology Organisation
506 Lorimer Street, Fishermans Bend VIC 3207 AUSTRALIA.
Key words: LabVIEW, virtual instrumentation, waveform analysis, ultrasonic software.
Ultrasonic waveforms contain considerably more information than is utilised by conventional flaw detectors, which are generally unsuited to the identification of complex or obscure features in the time- and frequency domains. Analysis of ultrasonic signals, as implemented on commercial instruments, is typically limited to measurement of the amplitude or time-of-flight of the signal within a user-defined time gate, with signal pre-processing generally restricted to the application of a distance-amplitude correction (DAC). For many applications, such a measurement may be sufficient to satisfy the requirements of an inspection and to guide decisions regarding repair and structural integrity of a component. However, for certain types of defects, this level of analysis may be inadequate, and more sophisticated procedures and algorithms may be required to process the acquired data.
Such algorithms could include amplitude comparison of two or more sections of a gated waveform, signal level compared to DC level, relative amplitude of two or more peaks in the frequency spectrum, and various other possibilities. Advanced analysis improves nondestructive inspection capability, by increasing the probability of finding and quantifying the defect being detected, and by aiding in signal interpretation, reducing the dependence on a high degree of operator understanding and experience.
DSTO has developed a flexible PC-based system for acquisition and analysis of ultrasonic waveforms in order to enhance research capability and to provide an adaptable tool that may be modified to suit a given application, with scope for future implementation in the field. The DSTO system is implemented within the National Instruments LabVIEW (Laboratory Virtual Instrument Engineering Workbench) environment.
The term virtual instrumentation refers to the replacement of fixed test and measurement equipment by PC cards and controlling software, or its enhanced and automatic control. Several instruments in the form of PCI- or ISA-bus cards can be housed in and operated from a single PC, saving the user from transporting several fixed pieces of laboratory test equipment unnecessarily, and making an entire test system more compact. Software libraries are available to perform waveform time and frequency measurements, plus various other applications including process monitoring and control. System components may be selected according to task requirements, and one does not have to rely on the manufacturer's choice for cards that operate in the system. The key advantage of such a system is its flexibility for modification, as any component may be updated separately, it runs with the fastest processing speed available, and the controlling software may be developed to suit evolving needs.
The virtual aspect of the system is evident in the controlling software. Often the graphical user interface (front panel) is designed to imitate the control panel of the hardware instrument it replaces, while the computer controls, mouse, keyboard, are used to adjust the on-screen dials and buttons. This may then be modified to any suitable configuration for the user and simplified, or complicated, depending on the needs of the application, and the capability of the card, or instrument. As a graphically based programming language, LabVIEW makes the creation of a front panel a matter of a few minutes, while each component seen by the user is controlled as a separate node in the code diagram, depending on its data type.
A feature of LabVIEW that makes it attractive for a flexible acquisition and analysis system is its modular nature, in which each software virtual instrument (VI) can be embedded as a component node, named a sub-VI, within a larger program. Thus, a series of sophisticated processing steps can quickly and flexibly be linked together. By creating VIs for particular operations, one makes each sub-VI a portable program of its own. The processing stages of a waveform might include signal rectification, extraction of a segment within a gate or sub-gate, processing of an FFT on the selected segment, and the extraction of features from the FFT to give a numerical result. Each process is performed in a separate sub-VI, and these are linked together, according to the direction of data-flow. Each sub-VI may have its own panel to modify its particular process settings, which is preferable to separate the graphical user interface from the data handling processes, or settings may be configured initially in a larger panel, and passed through each sub-VI with the data. A description of the DSTO waveform processing system follows, together with examples of its application to nondestructive testing measurements.
The general form of the data acquisition and processing system consists of an ultrasonic pulser and receiver, a high digitisation-rate analogue-to-digital converter and the data acquisition software (figure 1).
|Fig 1: Waveform analysis experimental setup.|
The basic functions of the system emulate those of a conventional ultrasonic flaw detector. The digitised signal may be initially high-pass filtered, and also full-wave, or else positive- or negative half-wave rectified. Up to four waveform "gates" may be set in the main window to facilitate analysis in the time domain (figure 2). A gate is indicated by a horizontal line in the waveform display, which defines a sub-window in time, where the selected segment is enlarged (figure 3). Gates are triggered either from the initial pulse, or from the beginning of an interface gate, which itself is triggered by a peak in the waveform, as on a normal pulse-receiver unit or ultrasonic flaw detector with these capabilities. Within each gate window, two further sub-gates may be set, enabling precise selection of any part of the waveform.
|Fig 2: The main display window of the waveform analysis software. Three gates are set over sections of the continuously-acquired waveform.|
|Fig 3: Three gate windows, displaying a section of waveform, according to their settings in the main window, with sub-gates set over part of the signal.|
The frequency spectra of the gated waveforms, and several other frequency-domain calculations including power spectrum, convolution and auto-correlation, may be performed and displayed in an FFT window for each corresponding time gate (figure 4). Up to two frequency-domain sub-gates may be set within each of the FFT windows to permit analysis of features of the frequency spectra. A grid, and two cross-hair cursors may also be applied to the gate, in order to locate a point on the waveform precisely, the coordinates of the cross-hair being displayed as numerical fields in the window.
Waveforms may be stored and subsequently recalled for comparison with other waveforms. Any arithmetic function may be applied to compare acquired and stored waveforms, displaying the result in the third window of the multi-waveform display (figure 5). An analysis panel computes a quantitative measure of a specified feature in any gate (figure 6).
Features may be compared between gates by any arithmetic operation. Available features include the amplitude of the waveform, as positive peak, negative peak or peak-to-peak, DC (background) signal level, number of waveform cycles within the gate, and the time of a positive or negative peak level occurring inside the gate. Any of the thirty gates and sub-gates may be compared with one another, using any of the listed features as a quantitative measurement. These scalar results are themselves able to be re-scaled, and compared with one another, then output from the software to a file, or for use in another software process.
|Fig 4: The three amplitude and phase spectra in the frequency domain for the gates set over the waveform in figure 2.|
|Fig 5: The multi-waveform display, showing the result (bottom) of a stored reference waveform (middle) being subtracted from the real-time waveform (top).|
|Fig 6: Analysis panel. The data from any gate or sub-gate, measurable by several methods, is compared with that from another gate, with any necessary scaling.|
LabVIEW is also flexible in its communication with other Windows-based software, programmed in various languages, and is adaptable to standard and other data types for reading and writing data between applications. For example, the analysis component of the DSTO waveform processing system was modified to communicate with a beta-release version of the ANDSCANTM NDE scanning system. The beta-release version in this case incorporates acquisition, analysis and storage of a full ultrasonic waveform as a function of probe position for a C-scan.
Many of the basic features of the DSTO waveform acquisition and analysis system are incorporated in the ANDSCAN software, with the added capability of storing waveforms as a function of position to permit subsequent post-processing of previously acquired C-scans. This is of great advantage for technique development during an aircraft tear-down program, in which components are inspected for defects, and subsequently torn down to assess the accuracy of the inspection method. Stored waveforms for each scanned point of the component give a valuable, permanent record for post-processing to develop and evaluate new processing algorithms so that defects not detected prior to teardown may be found.
By interfacing the DSTO analysis system to the ANDSCAN system via a software link, the flexibility of the LabVIEW-based system is combined with the advantages of the ANDSCAN system for storage and subsequent post-processing of acquired waveforms as a function of position. While running both applications simultaneously on a PC, waveforms are exported from ANDSCAN to the LabVIEW system, and scalar values, extracted from analysis of the waveform in LabVIEW, are returned to ANDSCAN and plotted as a function of position in a C-scan presentation. Novel analysis algorithms are promptly implemented by the rapid coding and library of analysis sub-VIs available in LabVIEW. Interfacing the LabVIEW code to another application also allows for the insertion of new capabilities, without requiring access to the source code, which is particularly valuable in a research environment.
Thin metal structures, such as aircraft skins, whose thickness is of the order of a few wavelengths of an incident ultrasonic excitation can sustain plate (or Lamb) waves that cause the entire thickness of the plate to oscillate as the wave propagates within the plane of the plate. Previous research has demonstrated that ultrasonic Lamb waves may be used to detect hidden corrosion in thin Al-alloy aircraft skins. Because of the waveform processing requirements, this is an ideal application for the LabVIEW-based system.
Several Lamb wave modes may exist, depending on the excitation frequency, plate thickness and plate material. The modes travel at different group and phase velocities, and are distinguishable from one another as illustrated in figure 7, which shows Lamb wave modes propagating in a thin Al alloy plate with differing degrees of corrosion and amounts of corrosion product on the back face. A pair of 5 MHz 30º angle-beam probes, spaced 100 mm apart, were used to generate and detect the waves. The first mode, s0, arrives after 35ms and is centred about 2 MHz. The next mode, s1, commences at approximately 38ms, superposed with s0 and has a centre frequency of 8.5 MHz. A third mode a1 with centre frequency 6 MHz arrives at 42ms. Clearly, the amplitude of the a1 mode is severely attenuated by the presence of the corrosion product. This attenuation is significant even before any appreciable loss of metal thickness has occurred.
To reduce the scatter in amplitude measurements due to variations in coupling, the a1 amplitude may be normalised against the amplitude of the s0 mode, which remains relatively unaffected for corrosion product layers less than 10% of the skin thickness, and has been found to be useful as a reference level for the a1 mode. A ratio of the peak-to-peak amplitude of a1 and s0 is proposed to reduce measurement error due to coupling variation. Using this ratio, a signal reduction threshold of 50% has been used to detect corrosion product layers of 5% of skin thickness, corresponding to material thickness losses of less than 1%. Peaks in the frequency spectrum, which can be measured concurrently, will indicate the presence or absence of any mode, as shown in figure 4. Measurements such as these go beyond the scope of commercial ultrasonic flaw detector units.
Fig 7: Three ultrasonic waveforms, displaying the different Lamb wave modes that propagate in a thin aluminium plate with various amounts of corrosion product on the back face. The corrosion affects the Lamb wave modes to differing extents.|
(a) Blank, uncorroded specimen,
(b) Corrosion product formed after one day of treatment,
(c) Corrosion product formed after five days of treatment.
Aircraft control surfaces vary significantly in thickness from one point to another. Consequently, when scanning a component in through-transmission, the ultrasonic pulses transmitted through the component shift in time by an amount which depends on the velocity of wave propagation through each layer of the structure. However, because different ultrasonic modes propagate at different frequencies, the different modes present in a through-transmission waveform shift by different amounts in the time domain as a function of specimen thickness. This is illustrated in Figure 8, in which mode 1 is time shifted by an amount D t(m1) when the probe is moved from position 1 to position 2, whereas mode 2 is shifted by a greater amount D t(m2) for the same change in probe position. This differing time shift means that simple gating procedures, in which a number of analysis gates are set at fixed delay times with respect to a reference or interface gate, are inadequate for analysis of such waveforms. Attempting to use such simple gating results in the pulses for different modes moving out of their respective analysis gates as the specimen thickness changes.
|Fig 8: Through-transmission scanning of a variable-thickness aircraft honeycomb component.||Fig 9: The shift in time of two wave modes, m1 and m2, in an ultrasonic waveform for two different component thicknesses.|
For the case of control surfaces containing honeycomb core, the time-shift of the pulse due to a change in honeycomb thickness can be calculated by consideration of the ultrasound velocity for each ultrasonic mode transmitted through the honeycomb core and waveform gate settings are able to be adjusted automatically based on a measurement. This capability is currently being implemented within the DSTO LabVIEW-based system to allow post-processing of C-scan images over variable-thickness honeycomb components. The implementation of the waveform gate adjustment will allow several wave modes to be monitored simultaneously during a scan of an entire component.
Structures of complex geometry such as aircraft control surfaces require a high level of analysis to find hidden, internal defects. The extent to which ultrasonic NDE is currently utilised can be enhanced by analysing in greater detail the information contained in reflected or transmitted waveforms. At DSTO, this analysis has been realised by the development of flexible software that is able to expand and compare waveform components, and present them in various sophisticated ways. This process has been facilitated by the use of a flexible data acquisition package that allows portable programs to be created, while greatly simplifying the construction of a set of virtual instruments, which themselves can be modified to suit particular applications.
Input and assistance from Susan Bowles, AMRL, is greatly appreciated. The cooperation received from DERA-Farnborough UK, and especially from Mr Robert Smith, in assisting with interfacing the DSTO system to the ANDSCANTM scanning system is gratefully acknowledged. Calculation of the relative time-shift of different lamb wave modes was performed by David Hatt, AMRL.
|© AINDT , created by NDT.net|||Home| |Top||