·Table of Contents
·Computer Processing and Simulation
Analysis & Result of Data Obtained from a new Laser Based NDT System (A new concept in NDT)
John M. Webster
John Webster Associates
Jackie M. Mew, Robert Topp
Computer Science Division
University of Portsmouth, Portsmouth UK.
The "RAID" (Remote Acoustic Impact Doppler) NDT system is unique and employs a
proprietary design acoustic transducer which produces an air coupled shock wave and a
customised scanning laser Doppler velocimiter. The result of this design configuration is that a single brief, but extremely high velocity shock wave of broadly unidirectional characteristics is launched into the air and used to impact and excite the object undergoing testing . The resulting relaxation frequencies occurring on the object surface are interrogated with the Doppler velocimeter.
The technique depends upon the hypothesis that any change in substructure will locally affect the surface frequency response spectrum. Thus, surface relaxation frequencies for any given material are dependent upon the underlying substructure of the object.
Remote interrogation of the relaxation frequencies is accomplished with a highly customised
scanning laser Doppler velocimeter. The acquired time domain signal is processed to a Fast
Fourier Transform (FFT). A velocity based image is computed and presented on a monitor
overlaid on an image of the object. Advanced techniques for computerised automatic analysis of
the images have been developed. Our results show both deep subsurface defects in solid as well
as honeycomb materials, delaminations in metal assemblies and the presence of hidden corrosion
in aluminium structures. It is also apparent that the system has many applications in field other than aircraft.
Initially the research program was initiated by DARPA (US Defence Advanced Research
Projects Agency) primarily for the detection of defects in composite aircraft structures. The
technique has however, proven itself in application to metal where delaminations, hidden
corrosion and other defects can be both located and characterised, and also in thick ceramic
structures such as insulation tiles. The RAID system is currently undergoing commercialisation
by a large aircraft repair and servicing company.
Defects, in composite structures, either in manufacturing or through field usage, usually manifest themselves at a subsurface level. Impact damage will frequently cause delamination of the structure either between laminates or skin/core interfaces. Delamination or bond weaknesses also occur in metal laminates. In a not dissimilar manner corrosion often manifests itself in such regions as lap joints in aircraft bodies where the defect lay between the bonded or fastened layers of metal .
To give a simplified understanding of the system the well known 'Tap Test" is a limited analogy.For the detection of defects in, for example composite materials, the operator taps the surface and listens for the audible response. In this case the subsurface structure is effecting the excited resonance response of the surface; hence the sound change according to the subsurface structure. In a similar manner, delamination in metals or the presence of hidden corrosion will effect out of plane vibration surface response; albeit perhaps not audibly. The tap test is a very subjective method of NDT, although there are scientific principles that can be exploited with modern technology. The RAID system has been designed to achieve this end.
The system is totally remote and requires no physical contact with the object undergoing testing;it has been demonstrated to operate at a stand-off distance of three meters. The system employs a proprietary design acoustic transducer that produces an air coupled shock or pressure wave,similar to that produced from a small explosion. This is achieved by discharging a high voltage capacitor within a period of less than 5 microseconds. The discharge is contained within a small ceramic chamber with an annular design anode providing an exit for the resultant hot gasses. The result of this configuration is that a single brief, but extremely high velocity shock wave of broadly unidirectional characteristics is launched into the air and used to impact the object undergoing testing[2,3] . The objective of the brief impact is to excite natural relaxation frequencies in the object and avoid any "blanketing" effect that would be present if continuous wave white noise was applied.
As discussed above, surface relaxation frequencies for any given material are dependent upon the underlying substructure of the object. Remote interrogation of these relaxation frequencies on the surface of the object undergoing NDT, is accomplished with a highly customised computer controlled scanning laser Doppler velocimeter employing special analysis programs. The technique depends upon the hypothesis that any change in substructure will locally affect the surface frequency response spectrum. The tapping analogy is very limited in satisfying an explanation because an important aspect of this technique is that it does not limit itself to the excitation of a single spot, but impacts a large area of the structure, thus primarily exciting out of plane vibrations.
The scanning laser Doppler vibrometer receives an analogue velocity time domain signal from
each of up to 500 x 500 data points. This is passed into the processing computer where, in our
initial work, it was processed to a Fast Fourier Transform (FFT), which is effectively a frequency
spectrum over the selected bandwidth recorded. Later work that we carried out subsequent to
prepared software programs, post processed the time domain data. The result of this latter
technique is that alternative processing procedures can be carried out. This data is then stored in the computer as individual frequency bands.
Up to a total of 6500 individual FFT lines can be recorded over a vibration spectral range of up to 200 kHz. More usually 1024 lines are employed from 1 to 30 kHz. However, this represents a vast amount of information where each of the individual FFT lines might be considered as a out of plane velocity snap shot photograph of the object vibrating at that individual frequency. This data stack, which might be thought of as a cube where the X & Y co-ordinates are the plain of the object and the Z axis is the frequency. The data stack is later analysed to select only the relevant information carrying frequency bands and a velocity based image is computed then presented on a monitor overlaid on an image of the object which was simultaneously grabbed by a CCD camera located in the vibrometer.
One of the most important aspects of this research program has been the development of a
system for the analysis of this very large quantity of data. Typically a defect in a given material might exhibit oscillations over several relatively narrow bands of the recorded spectrum. The problem was to devise a computer algorithm[4,5] that would separate out the relatively few frequencies containing the defect information whilst rejecting the remaining mass of the recorded frequency spectrum and thus reducing noise. The process has been likened to searching for the proverbial needle in the haystack. Along with the analysis program, we have recently developed a technique of post processing the recorded data and to "zoom" into selected regions of the time domain trace; this does speed up the whole data acquisition process and again reduces any noise problems.
An FFT applied to this situation can be considered as a frequency spectrum with a predetermined number of discreet frequency lines of information over the chosen frequency range. As discussed above, an alternative way is to describe it as a data stack with potential image forming data in the horizontal X&Y axis and frequency information in the vertical or Z axis, as illustrated in Figures 1 & 2. For simplicity, the diagram only shows three primary areas of a complex situation in which image information was contained in small groups of frequencies ranging from 10 kHz to 25 kHz. In very brief and simplified terms, the analysis algorithm operates on the principle that
the bandwidth of a defect will be greater than that of noise, thus it searches for vertical links
within the data cube and extracts the relevant frequencies. It then applies a weighting factor to
compensate for the lower velocities that occur at higher frequencies.
Fig 1: Selection of defect frequencies
Fig 2: Schematic representation of the random noise, bending modes, and signal in the
frequency spectrum obtained in a typical scan.
2. Results: Composite & Metal Bonds
The RAID technique has been successfully applied to the location and characterisation of defects
in a wide variety of composites, ceramic and metal bonded materials including the situation
where kiss contact delaminations are present. This presentation only permits room for a few
examples. Figure 3 shows a defect in a carbon composite nomex honeycomb cored material in
which a parametric repair has been carried out. The defects in this test specimen were simulated
by the presence of Teflon shims, which give the effect of a kiss contact situation. In some cases
the defect was on the rear face between the skin and the nomex core and in other either below the
surface skin or on the scarf of the repair. Note the defect information has been overlaid on the
recorded image obtained from the CCD camera located within the vibrometer
Fig 3: Seven defects in a carbon composite parametric repair panel.
Fig 4: Delamination revealed in the 50 mm thick black heat insulating tile. The defect area is
clearly delineated in green
Figure 4 shows a debond in a 50 mm thick heat insulation soft ceramic tile of the type used to
protect during re-entry of space vehicles. The RAID system has also reliably detected subsurface
cracking in other modern foam materials now used in airframe construction.
Applied to metal weak bonds, figure 5 shows a weak bond in a rib to skin joint in an aluminium
helicopter rotor blade. The computer has been used to enhance the image of the defect region.
Fig 5: Metal helicopter blade: Application of RAID to detect weak bond in metal
Figure 6 shows an aluminium aileron with a aluminium honeycomb core that has been damages
by an ingress of water freezing and crushing the core region.
Fig 6: Aluminium aileron section showing crushed core
3. Application of RAID to the detection & quantification of hidden corrosion
Corrosion in aluminium structures is largely calendar driven and is an increasing problem in our
ageing air transport fleet where a structure weakened by metal loss in a pressurised cabin
situation, could potentially threaten the safety of the aircraft. Currently there are aircraft in
regular service that are over thirty years old [6,7,8] An example of this is the US fleet of KC135
Tanker aircraft, which are actually the old Boeing 707's with a future service life projected to be
another 35 years. Reliable detection and quantification of metal losses under field conditions
does present a difficult situation. Currently, no reliable and practical method exists, although
there are considerable efforts now directed towards a solution. This technology presents
promising early results in detection and quantification of corrosion in lap joint situations.
In very simplified terms, the system operates very much as it does in the field of composites. The
effect of the presence of corrosion will affect the local relaxation frequencies. A rather over
simplified analogy is a drum surface: The frequency produced by the drum will be the sum total
of the boundary conditions, the skin thickness and the tension of the skin. In the case of hidden
corrosion the metal skin thickness is a function of the metal losses due to corrosion, the boundary
condition will be the extent of the corrosion whilst the tension is a function of the corrosion by
product. Aluminium oxide has a lower density and a larger volume than the original metal, thus
causing a tension on the surface.
Initially corrosion samples were obtained from the US Airforce Logistics command. The results
were promising and corrosion was detected, however, no attempt was made to quantify at this
stage of our work. As a result of these tests, we made up a lap joint in the laboratory and
simulated metal by the removal of metal: 5%, 10%, and 15% respectively. The result is
illustrated in Figure 6. This result clearly shows all three regions of simulated corrosion spots
overlaid on a CCD record of the lap joint sample.
From the data contained in the rich data set on the computer record of the scan, we extracted the
FFT spectrum of each data point of the primary antinodes in each of the fault regions. What is to
be noted here is that the peak response for each of the different defects is in an entirely different
region of the vibration spectrum and relates directly to the metal thinning in that region. Figure 7
is a plot of the metal loss condition against the resulting frequency. It conforms to the equation
for membrane resonant frequency, wn, given by :
Fig 6: Lab joint panel manufactured with 5%, 10% and 15% metal thinning. Note: all three
regions are clearly revealed.
B is a variable dependent upon the shape of flaw and edge conditions, E is the Young's modulus,
t is the thickness of plate, r is the mass density, a is the dimension of plate, and v is the
Fig 7: Graph of metal thinning against resonant frequency for the results in figure 6
Figure 8 is a record of an aircraft shin lap joint containing hidden corrosion between the shin
layers. Interestingly the antinodal vibrations of the various regions comprising the corroded
region can be broken down into their component parts each at it particular frequency .
Fig 8: corrosion in a lap joint
Fig 9: A breakdown of the corrosion illustrated in Figure 8: Region vibrating at 8750 kHz.
Other regions, not shown here, vary up to 23 kHz.
This work was sponsored by DARPA Phase III # DAAH01-98-R152 Contract "Development of
an Optical/Acoustic Method for the Remote non-destructive Inspection of Large Area Composite
Materials for Structural Defects". "Approved for Public Release Distribution Unlimited"
- J.M. Webster et al. Optical technique for the detection and imaging of subsurface defect in
composite structures using Remote Acoustic Doppler. J. Sci. Imaging. No1. Vol. 46. 1998.
- J. M. Webster. Acoustic Wave Generating Apparatus. US Patent No. 5,616,865 1997.
- J. M. Webster & J. M. Mew Method & Apparatus for Non-Destructive Inspection of
Composite Materials and Semi-Monocoque Structures US Patent No. 5,505,090 1996
- J. M. Webster, et al., "The detection of subsurface defects in composite structures & metal
structures by the application of remote acoustic Doppler technique" - First Pan-American
Conference on Non-destructive Testing, Sept. 14-18, 1998, Canada.
- J. M. Mew et. al. ASNT Spring Research Conference 1998
- R. A. Smith, Non-destructive evaluation for corrosion in ageing aircraft. Part 1, Insight, Vol.
No. 37, No.10, 1995
- J. P. Komorowski et. al., "Quantification of corrosion in aircraft structures with double pass
retroreflection", Canadian Aeronautics & Space Journal, Vol. 42, No. 2, 1996.
- D. Perry, Sandia National Lab, "Emerging NDE technology for ageing aircraft", ASNT
Spring Conference and Research Journal, 19989
- C. M. Harris, "Shock and vibration handbook", Fourth edition, 1996.
- Note: the RAID system is covered by a series of patents registered in the names of
J.M.Webster & J.M. Mew.