Inductively excited thermography has been shown to detect cracks in metallic components with good sensitivity. It is discussed as an alternative to magnetic particle testing. An open question to achieve acceptance in the industry is its testing reliability. A study with in total 200 forged steel parts was performed in order to compare the testing reliability of automated inductively thermographic testing and magnetic particle inspection. A robot supported thermographic inspection station was used. An inductor with orientation-independent crack detection was built up and tested. The thermographic phase images obtained were analysed by an automatic defect detection procedure based on machine learning techniques. Results of magnetic particle inspection served as a reference. Depending on the type of test object, an agreement of 68% to 82% was achieved, if only large indications of thermography were considered. The weak thermographic indications turned out to be due to shallow cracks (<150 mu m depth). Improvement of the testing speed can be achieved by inspection inside large coils.

The one-dimensional propagation of electromagnetic waves and the propagation of the resulting thermal waves in conducting material are analysed in a coherent way. The heat release due to resistive losses has a static and an oscillating part. Both are considered as heat source terms for the thermal diffusion equation. The time dependence of the temperature is described by analytical solutions. Electrically and thermally conducting materials are classified by the ratio of thermal penetration depth to the skin depth. Experiments performed on ferritic steel, stainless steel and carbon-fibre-reinforced polymer show the time dependence of the thermal signal after heating begins, as described by the theory. At low induction frequencies, an oscillating part of the surface temperature at the double of the induction frequency is detected in accordance with the theory. The results point out new opportunities for induction thermography.

A new approach for detection of cracks in metals with low emissivity was investigated. Liquid gas is sprayed in a linear movement across the test object. The gas propagates into cracks open to the surface by capillary effects and by the pressure of spraying.This causes cooling and subsequent warming up. It was found that the time profiles of the temperature close to the crack have characteristic properties which are different from the time profiles of a pure surface distribution of the emissivity. Cracks of different depth were investigated in aluminium and steel. An advantage of the technique compared to liquid penetrant testing is the residual-free evaporation of the gas that allows testing in a single step

A study with in total 200 forged steel parts was performed in order to compare the testing reliability of automated thermographic testing and magnetic particle inspection. A robot supported thermographic inspection station was built up. The thermographic phase images obtained were analysed by an automatic defect detection procedure based on machine learning techniques. Results of magnetic particle inspection served as a reference. A good testing reliability was found.

Single planar fuel cell elements consisting of metallic interconnectors that are bonded and sealed by a thin glass solder layer form the core of a solid oxide fuel cell. For reliable operation, the bonding layer has to adhere well and must be without voids or foreign material inclusions, which might cause gas leakage, electrical shorts or mechanical weakening and structural failure. Nondestructive testing (NDT) by the high-frequency ultrasound in immersion technique and by air-coupled ultrasound was optimized to find such defects. Another technique was flash light excited thermography for detection of voids. The NDT techniques used are complementary to each other, as they are based on different physical principles. Voids and small steel platelets of different sizes were prepared in the glass solder layer before the high-temperature bonding process and then monitored by the NDT techniques through the interconnector plates. Two selected NDT techniques were then validated in a probability of detection (POD) study. The study resulted in detection limits for the two main types of defects. As a step towards production testing, a demonstrator was built combining testing by air-coupled ultrasound and that by flash thermography. During the testing steps, the cell elements were handled by a collaborative robot.

Active thermography is considered in this paper as a non-destructive testing technology to ensure the quality of solid oxide fuel cells. The acquired infrared images are automated processed by artificial intelligence methods in order to detect defects within the glass solder layer of solid oxide fuel cells. For this purpose, three supervised machine learning methods are investigated: (1) Support Vector Machine, (2) Adaptive Boosting, (3) U-Net. Among those methods, the U-Net method outperforms the other considered methods with higher accuracy and F-measure.

Two modified algorithms for pulsed phase thermography were defined. The principle is to mask out parts of the thermal signal that only add noise but contain no significant information. A first algorithm uses a frequency dependent Gaussian window function. A second algorithm leads to a rectangular window function. The algorithms were tested on synthetic signals modeling a circular hidden defect in a plate and experimentally by flash excited thermography on a steel and a polymer sample. For both materials, a significant improvement of the contrast-to-noise ratio from the defects was obtained for higher analysis frequencies.

Two interconnector plates made out of ferritic steel are joined together by a thin layer of glass-ceramics and form an airtight assembly as a core part of a solid-oxide fuel cell (SOFC). The sealant has to withstand temperatures above 800°C and has to be gas-tight and mechanically stable. The solder layer must be free of larger voids and metallic inclusions. In particular, electrical shorts between the steel plates have to be detected and localized. Flash-light excited thermography in one-side access and in transmission was employed to detect artificial and natural local voids and inclusions in the glass-ceramics. Two flash-lights with each 6.4 kJ of energy were used. The recorded thermographic image sequences were pre-processed by pulsed phase thermography (PPT). Short- and long-time reproducibility tests were performed. Sets of samples with prepared test defects (air voids and metal platelets) of different size were measured after optimization of the excitation and detection parameters. One of the steel plates consists of two tightly joint separate sheets. This may cause false alarms in thermographic testing due to small air gaps between the sheets. The experimental findings were supported by numerical FEM simulations using COMSOL Multiphysics. A POD (probability of detection) analysis was performed showing that voids with a diameter of 2.3 mm can be detected reliably. Electrical shorts between the steel sheets could be localized by a lock-in technique using a modulated electrical current. Resistive losses at the internal contact points generate heat which becomes visible as a hot-spot in the thermal image.

Two modified algorithms for pulsed phase thermography were defined.The principle is to mask out parts of the thermal signal that only add noise but no significant information.The algorithms were tested on synthetic signals from a circular hidden defect and experimentally by flash excited thermography on a steel and a polymer sample.A significant improvement of the contrast-to-noise ratio of defects was obtained for higher analysis frequencies.

A fully convolutional neural network was set up for the detection of crack-type defects and for the defect shape prediction of thermography datasets. The method uses a supervised neural network for sematic segmentation (U-Net). For these tasks, training datasets of forged parts were acquired through induction thermography. The approach provides a significant improvement over conventional methods of thermal signal and image processing used in active thermography. Furthermore, the results may lead to new procedures for a quantitative evaluation of flaws and defects in non-destructive testing using infrared thermography

A fully convolutional neural network was set up for the detection of crack-type defects and for the defect shape prediction of thermography datasets.The method uses a supervised neural network for sematic segmentation (U-Net).For these tasks, training datasets of forged parts were acquired through induction thermography.The approach provides a significant improvement over conventional methods of thermal signal and image processing used in active thermography.Furthermore, the results may lead to new procedures for a quantitative evaluation of flaws and defects in non-destructive testing using infrared thermography.

A neural network (NN) for semantic segmentation (U-Net) was used for the detection of crack-type defects from thermography sequences. For this task, data sequences of forged steel parts were acquired through induction thermography and the corresponding phase images calculated. The results for defect detection were quantitatively evaluated using Intersection over Union (IoU) metric. Further, a combination of 2D convolutional layer as well as LSTM (Long-Short-Term-Memory) is shown, which includes three-dimensional aspects in the form of time dependent and spatial changes and allows a defect shape reconstruction of back wall drillings. Therefore, pulsed thermography sequences were simulated with COMSOL Multiphysics. Finally, the reconstruction results were compared with the ground-truth defect profile using Mean Squared Error (MSE). The approaches provide improvements over conventional methods in non-destructive testing using infrared thermography.

In this paper, an active thermal imaging technique using a laser as an excitation source is used to detect aluminum specimens containing artificial defects. The presence of defects in the local excitation of the laser can produce better edge effects, using different laser excitation methods to improve edge effects. In the experiment, the laser scanning method is designed according to the characteristics of artificial defects, and the experimental data is analyzed. PPT and Sobel filtering are used to further data processing to obtain clearer defect edges and detect the detailed features of the defects. The study found that the detection effect is the best when the laser power and pulse width are large.

In this paper, carbon fiber reinforced polymer (CFRP) is tested using lock-in thermography, which is excited by an induction coil. A impact-damaged CFRP plate was available and measured as a function of the angle between a linear excitation coil and the main fiber directions. The amplitude and phase images of high signal-to-noise ratio (SNR) was obtained by Fast Fourier Transform (FFT). The relationship between angle sensitivity of included angle and crack shape is analyzed. The results show that the circular symmetry is basically maintained when the defect is linear. When the structure of defects is similar to that of bifurcated trees, the angular distribution is quite different, which are consistent with the results of X-ray imaging.

The effect of non-metallic polymer coatings on the induction thermographic signal from metal surfaces with cracks was studied.Analytical calculations for the backward thermal wave propagation were performed both in the frequency and in the time domain.Numerical simulations were used to show the thermal patterns of cracks under coatings.Experiments were performed on coated samples with artificial cracks.The crack contrast was analysed under coatings of various thickness.The contrast in amplitude and phase is decreasing with coating thickness.The effect of infrared transparency of a model coating was shown in experiment.

Experiments were performed by induction thermography on carbon fibre reinforced plates. Uni-axial, bi-axial materials as well as woven fabric were studied. The uni-axial material was investigated as a function of the fibre orientation angle and showed heating far away from the inductor and heating patterns with local minima. The occurrence of minima was explained by induction current cancellation effects. The experimentally measured thermal contrast patterns were confirmed by the results of numerical simulations. Fiber breakage could be detected better by induction thermography than by optically excited thermography.

In this paper, eddy current pulsed thermography was used to evaluate ballistic impact damages in basalt-carbon hybrid fiber-reinforced polymer composite laminates for the first time, to our knowledge. In particular, different hybrid structures including intercalated stacking and sandwich-like sequences were used. Pulsed phase thermography, wavelet transform, principle component thermography, and partial least-squares thermography were used to process the thermographic data. Ultrasound C-scan testing and X-ray computed tomography were also performed for comparative purposes. Finite element analysis was used for validation. Finally, an analytical and comparative study was conducted based on signal-to-noise ratio analysis.

In this paper, eddy current pulsed thermography in transmission mode was used to detect the damages caused by low-velocity impacts in carbon fiber-reinforced polymer and basalt-carbon hybrid fiber-reinforced polymer laminates. In particular, different hybrid structures including intercalated stacking and sandwich-like structures were used. The impact energy of 12.5 was used for the evaluation of the impact damage level. Ultrasonic phased-array C-scan was performed for comparative purposes. In addition, the advantages and disadvantages of the two structures were identified and discussed.