This is a modern pulsed thermography system with portable flash inspection head. Source: Thermal Wave Imaging
Until recently, thermography has been widely regarded as a qualitative adjunct to traditional nondestructive test (NDT) techniques. However, advances in analytical methods, improvements in infrared (IR) camera technology and the development of NDT-specific hardware have enabled portable thermographic NDT systems capable of highly accurate quantitative measurement, with sensitivities far beyond that of previous generations. Although it is possible to detect very severe near-surface defects using a simple excitation setup, such as a heat gun or heat lamps and an IR camera, deeper or subtler features are not detectable in the raw IR camera image. In such cases, the entire sequence of images captured during or after thermal excitation must be analyzed.
Thermography is widely used as a stand-alone inspection tool for both in-service and manufacturing applications in the aerospace and power generation industries. The technique is based on heating the sample surface with a brief pulse of light from a flashlamp and monitoring the surface temperature of the sample with an IR camera, which is interfaced to a PC. Source: Thermal Wave Imaging
The Thermographic Signal Reconstruction (TSR) method was developed specifically for this type of sequence analysis. In TSR, each pixel is converted to an equation that represents its deviation from expected defect-free behavior. The result is a noise-reduced image in which true subsurface defects are highlighted. Because only the equations' coefficients are stored, significant data compression also occurs in TSR, facilitating rapid inspection of large structures. Applications include thickness measurement of thermal barrier coatings or turbine blade walls, corrosion detection and measurement of material loss, and measurement of porosity and backing film inclusions in composites.
This system offers fast, noncontact, wide-area inspection of flat or curved structures, and can measure depth and area of subsurface defects. Source: Thermal Wave Imaging
Heat transfer concepts
The idea that the surface temperature response of a sample to an applied heat pulse is affected by its internal structure follows basic heat transfer concepts. The introduction of the infrared camera in the 1970s allowed researchers to experiment with image-based IR to detect subsurface flaws. In these early systems, samples were excited with hot air, light and various other mechanisms, while the surface temperature was monitored with an IR camera. Large internal voids or discontinuities blocked the heat flow from the surface to the bulk of the sample, producing areas of elevated temperature on the surface, corresponding to the rough outline of the flaw. The range of materials inspected this way was limited, because of the slow frame rates of early opto-mechanical scanners. Systems also were hindered by limited sensitivity and dynamic range, and the fact that the need for LN2 cooling made the cameras cumbersome and impractical for many real-world inspections. As a result, only very severe, near-surface flaws could be detected. Although encouraging, results from these early experiments were entirely qualitative, and required extensive (subjective) interpretation by an expert operator.
Despite these initial shortcomings, the enormous potential of IR NDT was widely recognized, in terms of speed and wide-area coverage, compared to conventional techniques such as ultrasound, which required point-by-point scanning. The situation improved gradually, in part because of advances in the underlying camera and computer technology.
The introduction of commercially available focal plane array (FPA) IR cameras allowed larger dynamic range and sensitivity than scanning systems, which allowed either dynamic range or sensitivity to be optimized, but not both. Furthermore, the FPA cameras provided digital data compatible with new data communication formats such as RS-422 and 1394. These improvements provided researchers real-time access to uncompromised data and faster computer processor speeds allowed them to perform processing on the raw signals.
Today, thermography is widely used as a stand-alone inspection tool for both in-service and manufacturing applications in the aerospace and power generation industries. Pulsed thermography has emerged as the most widely used form of the technique; it allows rapid inspection and can be configured as a portable system. This technique heats the sample surface with a brief pulse of light from a flashlamp and monitors the surface temperature of the sample with an IR camera, which is interfaced to a PC. The data are processed within a few seconds of capture providing an image of subsurface defects that may exist in the sample.
The Thermographic Signal Reconstruction method allows detection and measurement of subsurface features with aspect ratios that are undetectable using conventional processing approaches. It is based on the analysis of the time histories of individual pixels, and as such, is capable of detecting defect that span the entire field of view. Source: Thermal Wave Imaging
In pulsed thermography, the criteria for defect detectability are most conveniently expressed in terms of Aspect Ratio, the ratio of the defect diameter to its depth beneath the surface. For a flash-heated sample, the raw signal from the IR camera will show indications of large, near-surface defects, such as defects with large aspect ratios-in most materials, defects with an aspect ratio > ~10 will be detectable in the raw camera data. However, inspection requirements typically have much more stringent requirements for minimum detectable defect size and aspect ratio. As a result, additional processing of the camera data is required.
The traditional approach to viewing the pulsed thermography experiment results is to view a "movie" of the data sequence, played back in slow motion on the PC display. This approach will reveal the large aspect ratio flaws discussed above to the viewer. However, there are significant limitations, including very low sensitivity, to this approach. In an image-based scheme such as this one, defect detection is based on the assumption that there is feature contrast in the image. It assumes that there are both good and bad portions of the sample, so the inspector's eye becomes trained to detect this contrast, a task that the human eye is well suited to.
In an actual inspection, a sample may be uniformly good or uniformly bad, such that there will be no feature contrast in the image field of view. In such cases, it would be entirely possible to miss a serious flaw covering a very large area. Furthermore, using a visual approach to analysis, it is relatively easy to confuse surface features, emissivity variations or heating nonuniformity with true subsurface defects.
It is possible to overcome these limitations, get quantitative measurement of defect properties and even perform automatic defect recognition if instead of viewing a time sequence of images, every pixel in the image is treated as a time series that is completely independent of all other pixels, eventually looking at an image, but only after the individual time histories have been processed.
Thermographic images of a composite panel with Rohacell inserts between each ply layers. The inserts in each row are 0.25 , 0.5 and 1 inch in diameter. At left, a raw IR image of the panel 1.6 seconds after flash heating is shown. At right, a depth map is created from reconstructed data. Source: Thermal Wave Imaging
Thermographic signal reconstruction
The TSR method was developed specifically to take advantage of heat diffusion physics in order to remove noise and extraneous non-thermal signal components from the individual pixel time series, and to accentuate signals that deviate from typical cooling behavior. The TSR process generates an equation based on a least-squares fit of a low-order polynomial to the logarithmic time history of each pixel. The result is a noise-reduced replica of the original time history. The overall result of the TSR process is a significant increase in sensitivity to low-aspect ratio features, as well as an order of magnitude reduction in RAM and storage space requirements.
In the TSR process, several hundred frames of raw data representing the time history of each pixel are reduced to a set of equations. The fact that the TSR information is presented mathematically in a closed form allows advanced manipulation-such as differentiation-calculation of inflection points or FFTs to be performed quickly, and no adverse noise effects. The time derivatives of the logarithmic time history are particularly useful in discriminating between defective and intact points. With the TSR method, pixels that deviate from linearity in their logarithmic time evolution are enhanced by differentiation. They are easily identified by their zero crossings and inflection points, compared to conventional parsing of the raw data.
The effects of the TSR process on inspection results are shown in the graphic, "Effects of the TSR Process on Inspection Results," on a composite sample fabricated from 350 F cure carbon/epoxy, 3K-8 harness- satin weave prepreg. Carbon fibers were 33 MSI, AS4. The laminate consisted of a non-symmetrical Ω0,90Ω 5-ply stack, which was cured at 80 psi for 90 minutes. The resulting laminate was 0.129 inch. Defects representative of voids and disbonds were fabricated from 2-millimeter thick Rohacell that was crushed to a thickness of 0.013 inch prior to placement in the laminate. The simulated defects' sizes were 1 inch, 1/2 inch and 1/4 inch in diameter. These were placed in series under each consecutive ply of the laminate. One of the longitudinal sides of the laminate was stepped at each ply. The panel was inspected using a commercial pulsed thermography system using a 320 by 256 pixel InSb focal plane array camera operating at 60 hertz. The raw IR images show the near surface inserts and thinner steps, but features are blurred and deeper features are not detectable. Furthermore, there is little correlation between pixel intensity and defect depth. However, in the depth map that was created by applying the TSR process to the same data sequence, all inserts and steps are detected, and blurring has been significantly reduced.
Conversion of a 300-frame, 5-second, data set to a TSR data set requires approximately 10 seconds on a 1 GHz PC. Once the data has been converted, creation of a TSR image, or a reconstructed time derivative image, requires only a few milliseconds. The original data set of the image sequence is typically quite large. For example, a 300-frame, 5-second sequence from a 320 by 256 pixel camera operating at a 60 hertz frame rate occupies approximately 49 MB of storage space or RAM.
However, the TSR data structure created from this same data set occupies less than 5 MB. As a result of this data reduction provided by TSR, it is possible to operate on several data structures simultaneously-the arrays of equation coefficients from each image sequence are combined into a single large array and processed in a single operation.
This capability is useful when large structures requiring multiple shots for complete coverage are to be inspected. The individual shots are automatically combined into a single shot, and processed and analyzed simultaneously using a dedicated software program. In practice, the inspector acquires the data, views the finalized combined shot and then zooms in to view areas of particular interest.
Modern pulsed thermography systems offer unprecedented speed, sensitivity and ease of use compared to early contrast based systems. The TSR method allows detection and measurement-depth and size-of subsurface features with aspect ratios that are undetectable using conventional processing approaches. It is based on the analysis of the time histories of individual pixels, and as such, is capable of detecting defects that span the entire field of view. Because the TSR data is a representation of the raw data in equation form, and only the coefficients of the equation are stored and manipulated, there is a significant reduction in the PC RAM requirements to compute the TSR data. As a result, it is possible to process many files of TSR sequences simultaneously in the space that a single raw data structure would normally consume. As a result, sequences of large structures, which comprise several thermography sequences, can be processed and viewed as a single entity.