Principle, Equipment and Applications of Line-Scanning Infrared Thermographic NDT

The Line Scan Thermography (LST) technique for thermal nondestructive testing (TNDT) uses continuous heating from an incandescent (quartz) linear tube heater. This provides a continuous linear inspection with a lower noise level than the classical single area Xenon flash TNDT technique. The basic theory of the LST technique is described, along with a description of a special, self-propelled LST device. It is shown that TNDT processes using a moving heat source can be evaluated by solving simpler, static TNDT problems using flash heating. An example of the experimental detection of hidden corrosion in steel is presented. The LST method is recommended for the continuous inspection of large objects, especially flat, cylindrical and/or conical shapes.


Introduction
Infrared (IR) thermographic nondestructive testing (TNDT) is based on the evaluation of differential non-stationary temperature signals T appearing in defect areas.Therefore, when comparing test procedures, a natural figure of merit could be the relative increase of the temperature in defect ("d") areas in regard to non-defect ("nd") material; this parameter is often called the running temperature contrast C (T d − T nd )/T nd T /T nd .Back to 1975, Karpelson et al. mathematically showed that maximal values of C appear if an instantaneous heat source with a pulse energy of W scans the bulk of a solid body from one volumic point to another [1].Schematically, this hypothetic situation is shown in Fig. 1a.Assume that two small adiabatic volumes V are chosen in a solid body.One volume belongs to the sound material and another one is located within a defect.Assume that an instantaneous heat source consecutively flashes in each elementary volume.The value of C can be calculated as the specific heat, and ρ is the material density.Here we use the well known formula: W cρV T .For example, an air-filled defect in carbon fiber reinforced plastic (CFRP), provides C 1024 (CFRP: cρ 1.23 × 10 6 J K −1 m −3 ; air: cρ 0.0012 × 10 6 J K −1 m −3 ).The above-described procedure can hardly be implemented in practice, however, there are a couple of test procedures, which can be regarded as approximations of this one.For example, a two-sided test procedure (Fig. 1b) allows investigation of material properties throughout an entire sample, even if such procedure is impractical in many real cases.
In a one-sided TNDT procedure, the points of thermal excitation and temperature monitoring are located on a front surface of a test object (Fig. 1c).In the former USSR, Bekeshko et al. proposed using a plasmotron and infrared (IR) radiometer for point-by-point testing of metallic tubes [2], and the same scheme was applied to the inspection of soldered joints [3].In 1968, Green elaborated an apparatus and methodology for inspecting nuclear fuel elements [4].A special attention was paid to suppression of emissivity noise by using a concept of non-stationary temperature signals caused by subsurface defects, and the proposed solution called a dual scan ratio technique has proven to be useful until present.In the same period of time, some basic research on applying scanning TNDT was fulfilled by Vogel [5].A simple solution for testing low-conductivity rubber-like materials was suggested by Gavinsky et al. in the former USSR who used near-surface thermoelectric sensors instead of IR devices, while heating was performed by means of a vortex tube [6].
A single-point scanning procedure, which is sometimes called a "crawling spot" technique [7], ensures an excellent repeatability of point-by-point measurements, but its scan rate is typically low, except the "flying spot" version of the procedure where remote heating and temperature recording are performed by means of swinging mirrors thus allowing fast inspection of thin materials (coatings) [8].Inspection productivity can significantly be enhanced by implementing a line scan procedure (Fig. 1d).In this case, thermal excitation is performed within a line of whose length can reach 1 m, and temperature monitoring requires using a line-scan IR radiometer.For example, a photo-recording IR device (FID-1) developed in the 1970s at Tomsk Polytechnic University, Russia, used a strip-like 500 W nichrome heat source with the length of 60 mm and a home-made IR line scanner, while amplitude temperature profiles were recorded on a photofilm, see Fig. 2 [9].This device was intended for detecting dry joints in brazed metallic parts with the scanning speed of 5 mm s −1 .A prototype of this device was the line scanning TNDT unit TD-21A developed by Storozhenko [10].The unit included a 250 mm-long line heater (power up to 70 kW m −2 ) and a line-scanning IR radiometer (spatial resolution 1 × 1 mm) with test results being recorded on an electro-chemical paper.The productivity of line scanners can be higher in Fig. 2 Photo-recording IR device (FID-1) [9] regard to point IR devices approximately by N times, where N is the number of spatial resolution elements in a scanned line.In practice, a gain in test productivity is lower because spot heating can be more powerful thus allowing a higher speed of scanning.For example, the above-mentioned TD-21A unit allowed productivity of up to 7 m 2 h −1 in the inspection of large-size wound shells made of glass fiber reinforced plastic (GFRP).Another advantage of both single-point and line heating is diminished volumic heat diffusion that may enhance both defect spatial and depth resolution, but this issue requires further investigation.
In the two procedures above, detectability of defects at particular depths depends on the time delay between the moments of heating and temperature measurement τ d .Respectively, if V is the speed of scanning, the distance d between areas of heating and temperature monitoring is d V τ d .In 1984, Shiryaev et al. suggested to scan temperature distributions in the direction of heat spot movement in order to improve the detection of defects located at different depths and characterized by different delay times [11].
For many years, a popular TNDT procedure has involved thermal stimulation of large areas on sample surface accompanied by the IR thermographic monitoring of the same areas at the heating and/or cooling stage (Fig. 1e).In this case, an optimal observation time τ m defines an IR image where particular defects are detected best of all.Numerous implementations of this technique were introduced years ago by Vogel [5], Kutzscher and Zimmermann [12], and Kubiak [13].Until now this procedure seems to be typical in both pulsed and thermal wave implementations of TNDT.
Finally, a combination of the two last procedures was introduced and patented in 1999 [11].The idea has been that heating is performed by a moving line heater while temperature is recorded by an IR camera (Fig. 1f).Such approach allows performing a single test where many time delays τ d can be analyzed at once by producing synthetic IR images composed of particular columns /lines corresponding to par-ticular values of τ d .Following this concept, Woolard and Cramer from NASA, USA, developed the so-called "thermal photocopier" where heating was performed by a 1 kW quartz lamp with an elliptical reflector and the temperature was monitored by an IR camera with the image format of 256 × 256 [14].This technique was called Line Scan Thermography (LST).The same authors performed the comparative study of LST versus flash thermography [15].In 1982, Vavilov and Taylor analyzed a 3D numerical model of TNDT by using a line moving heat source [16].Khodayar et al. proposed an analytical model of LST [17], and Morgan and Rajic suggested a technique of Dynamic Pulse Phase Thermography (DPPT) for enhanced data processing [18].In the 1990s, Lehtiniemi and Hartikainen developed a hand-held TNDT device implementing the principle of induction line heating for inspecting carbon fiber reinforced plastic (CFRP) [19].Later, the induction heating technique was applied by Oswald-Tranta and Sorger for inspecting metals [20].At a microscopic level, scanning laser heating was used to detect poorly bonded microcircuits [21].
In this study, the features of the LST technique are summarized and an innovative TNDT device is introduced along with experimental results obtained by detecting corrosion in a steel sample.For illustration, consider LST inspection of cylindrical shells made of carbon/carbon composite (thermal conductivity 5.9 W m −1 K −1 , thermal diffusivity 5.56 × 10 -6 m 2 s −1 ).The Tables 1 and 2 contain results of solving the corresponding test problems in the Cartesian coordinates x, y and z in the case of uniform and line heating of 5, 10 and 20 mmthick samples, which contain air-filled 10 × 10 × 0.2 mm 3 defects at the depths of 25, 50 and 75% of the sample thickness L. In all cases, the heating power was Q 10 kW m −2 and the heating duration was 10 s.The heat strip with the width of 20 mm moved with the speed of 2 mm s −1 .The two figures of merit were used to compare optical parameters of each procedure: the temperature signals T m caused by the defects and the times of their observation τ d (line scanning test) and τ m (area heating test).In the case of uniform heating, the time τ m was counted from the end of heating while in the LST procedure this time (τ d ) was calculated in regard to the rear edge of the heat strip.The calculations were performed by using the specialized ThermoSource software from Tomsk Polytechnic University, Russia.
The data in Tables 1 and 2 shows that both test procedures provide close values of temperature signals, namely, from 5 to 0.02 °C, for the defects located at depths from 1.25 to 14.8 mm.As for optimal observation times, in the case of uniform heating, they appear at the end of heating.In the LST procedure, the τ d values are from − 3 to 9 s if to count from the rear edge of the heat strip.The negative values of τ d mean that optimal observation points are located within the heated line area.The general conclusion is that some important LST parameters can be evaluated by analyzing a simpler TNDT model based on uniform heating.
The values of T m and d obtained by following the methodology above represent the input data for designing optoelectronic parameters of experimental apparatus.Note that T m is linearly proportional to absorbed energy Q•τ h where τ h is the heat pulse duration, therefore, temperature signals over defect can be enhanced by increasing the absorbed energy.In practice, defect detectability mainly depends on surface noise, of which amplitude can be expressed in terms of noise running contrast [22,23], but a deeper consideration of noise is beyond the scope of this study.
Consider LST of a cylindrical object (Fig. 3).The input test parameters are as follows: (1) image format M x × M y , (2) test object size (diameter), (3) IR imager field of view α x × α y , (4) distance H between the IR imager and the object, (5) distance d between the heat strip rear edge and the central All defects size is 10 × 10 × 0.2 mm 3 *d is the distance between the rear edge of the heat strip and the recording point, τ d d/V ; negative values of d and τ d mean that optimal observation points are located either within the heat strip or, in some cases, even in front of it

Self-Propelled TNDT Device and Experimental Results
The LST principle has been implemented in an innovative self-propelled test device intended for TNDT of flat largesize parts, see the photo in Fig. 4. The device includes a tubular 1 kW halogen lamp and an IR thermographic Optris 450 module, as well as a stepper motor and a wireless control system.A mobile platform has been realized on the base of rollers and V-rib belts; this ensures smoothness of movement and a reasonable contact area with a test object thus preventing the sliding of the device on slightly inclined test surfaces.
According to the LST principle, the operator sets up the following unit parameters for detecting defects in a particular material at particular depths: movement speed, heating power and frame acquisition rate.By taking into account that the field of view is 0.35 × 0.25 m and the test speed varies from 5 to 50 mm s −1 , optimal detection times are in the range from 4 to 11 s for the 50 mm s −1 speed and from 30 to 100 s for the 5 mm s −1 speed.It is also true that, in many cases, the choosing of non-optimal delay times still allows detecting defects within a certain range of depths.In details, the technical characteristics of the flaw detector were reported in [24].Depending on the pre-set test parameters, the unit ensures inspection productivity of up to 20 m 2 h −1 .The efficiency of LST and potentials of the test equipment described above were evaluated by inspecting a reference steel sample.Such samples are widely used in thermographic detection of corrosion.A 310 × 235 × 10 mm steel plate contained 9 flat bottom-holes with the diameter of 10, 20 and 40 mm simulating corrosion with the material loss of 10, 30 and 60% (Fig. 5).The sample surface was painted black with a matte acrylic dye.The methodological aspects of corrosion detection were discussed both theoretically and experimentally elsewhere [25][26][27].
The sample was tested by using both the LST and "classical" (area-by-area) TNDT procedures.The speed of scanning was 40 mm s −1 thus providing the total test time of about 8 s. Figure 6a shows the synthetic (pseudo static) IR thermogram for the delay time of 4 s where up to six defects can be detected.The set of synthetic IR thermograms was processed by applying the technique of Principal Component Analysis (PCA) to result in the identification of eight defects including 10 mm-diameter corrosion areas with the material loss of 60 and 30% (Fig. 6b).Note that the PCA analysis was applied to a set of LST images composed for time delays from 5 to 15 s.To some extent, such set of images is equivalent to image sequences, which are obtained by using pulsed TNDT.It is also worth mentioning that the technique of PCA is one of the most popular in TNDT along with the Fourier Transform (FT).However, the two techniques differ by their physical base.FT provides a clear physical sense for processing test results in the phase domain where magnitude-related phenomena can seriously be suppressed.In the case of PCA, the statistics of IR image temperature evolution is analyzed to stress phenomena of spatial and temporal behavior of temperature in defect and defect-free areas.In fact, it is not easy to predict which technique will be more appropriate in a particular test case.In this study, PCA has been been used to illustrate improvement of TNDT efficiency when applying LST.For comparison, Fig. 6c, d shows the results obtained under uniform heating for 5 s by means of two halogen lamps with the total power of 2 kW.Up to four areas of material loss can be identified in the raw image (Fig. 6c).The PCA processing also allowed to detect eight defects (Fig. 6d), although visually the corresponding image seemed to be noisier than that in Fig. 6b.This can be explained by a higher noise resistance of the LST procedure due to efficient suppression of the reflected radiation in the self-propelled TNDT device.
The classical test procedure was fairly time-consuming (about 60 s), and the advantage of the LST technique by test productivity may be crucial in the inspection of large objects.It is worth noting that, in the case of LST, the best test result appeared in the 1st PCA component (Fig. 6b), while the classical inspection technique revealed the defects best of all in the 2nd PCA component (Fig. 6d).This can be explained by the different texture of LST and classical images, namely, by greater heating non-uniformity in the case of whole field heating.

Conclusion
The LST TNDT technique provides a continuous linear inspection and a lower noise level compared to the classical single area flash TNDT technique.The LST technique synthetic IR thermograms that correspond to varying time delays, and this allows the determination of defect depth.The thermophysical problem with a moving heat source can be approximated by solving a simpler problem using uniform static heating.This is explained by the same principle of temperature signal build-up in defect areas when using either TNDT technique.
The LST technique requires the synchronization of heating and data recording processes resulting to produce images that are free of uneven heating patterns.These images can be efficiently processed by using known TNDT algorithms, such as PCA.The LST method is recommended for the continuous inspection of large objects, especially flat, cylindrical and conical shapes.Large flat parts can be tested by a selfpropelled TNDT device.The practical implementation of this is also described in this paper.

Table 1
TNDT parameters (10 KW m −2 uniform heating for 10 s, onesided procedure)* All defects size is 10 × 10 × 0.2 mm 3 **τ m is the delay time after the end of heating; zero values mean that the maximum T m signal appears at the end of heating *