3. Extended Non-destructive Testing for Surface Quality Assessment

AWT allows the inline monitoring of the surface state, specifically through an inspection of its wettability.

Throughout the last decade, various aspects of this technology have been enhanced, whereby the most recent advances were achieved in the ComBoNDT project. Hereby, both the hardware and software were adapted in order to achieve more relevant and more reliable results in terms of measurement, data evaluation, and post-processing.

Here, we present our findings achieved by applying AWT to assess distinct contamination scenarios for CFRP adherends with different shapes.

The AWT results obtained for the scarfed CFRP pilot samples are presented in the following. They cannot be directly compared with the results for the coupon samples or those of the realistic parts. Indeed, the variations between part type, geometry, and surface have a significant influence on the AWT measurement and evaluation results.

Therefore, we perform comparisons only among similar parts of the same type. The analytical process was set up following the objective of detecting surface contamination on the respective part.

In order to establish and enhance the performance of AWT when inspecting the surface states of CFRP parts relevant for specific aeronautical user cases, we iteratively determined and advanced the abilities of the system on flat coupon samples, pilot samples, and realistic parts and then conducted an assessment of the potential inline application of the technology. As explained, this technology relies on a comparative assessment of the surface state. Hereby, the very powerful systematic AWT inspection procedure, based on a convolutional neural network (CNN), must be taught to differentiate contaminations. Once the system had been trained to correctly detect the droplets and to differentiate the contaminations, its use in inline applications was straightforward.

The inline application allows fast and non-destructive monitoring and classification of the surface states and clearly exceeds the performance of the NDT approaches used so far (e.g., the water break test). Most contaminations investigated during the ComBoNDT research project in distinct scenarios with various degrees of contamination were successfully detected, and even the relevant contamination levels investigated here could be differentiated.

In conclusion, we developed sensitive and productive AWT procedures that not only facilitated the differentiation between the surface states of clean and intentionally contaminated parts but also permitted discrimination between distinct levels of contamination for several contamination scenarios. As the significance of AWT investigations and the thus obtained findings are based on contrasts in the wetting behavior of the inspected surfaces, our investigations plausibly indicate that AWT is a very surface-sensitive technique that enables a significant differentiation between clean surfaces and surfaces with sub-monolayer and monolayer contamination. However, AWT does not allow discrimination between surface states composed of a few or several molecular layers of contaminants. Finally, the integration of the technology in inline applications without major constraints was achieved.

OSEE is a surface analytical technique that relies on recording the photocurrent emitted from a sample surface region illuminated with UV light, typically under environmental conditions that prevail in cleaning or adhesive bonding processes. There are two modes of operation. During “microscopic” OSEE mapping, the sample is scanned using an electron collector, typically by applying lateral movements with a speed ranging from 1 mm to 1 cm per second. Meanwhile, during “spectroscopic” local measurements, the photocurrent is measured at fixed positions upon varying and recording the hold-up time. We performed our OSEE experiments under ambient conditions using an SQM300 surface quality monitor (purchased from Photo Emission Tech., Inc. (PET), USA).

Regarding the principle of an OSEE measurement in more detail, during the local inspection, regions of the sample surface are exposed to UV light from a mercury vapor lamp with prominent emission maxima at 4.9 and 6.7 eV. As the work function of the respective substrate surface amounts to approximately 5 eV, the emission maximum at the higher energy of 6.7 eV essentially contributes to the photoelectrons emitted by the sample surface, and the emitted electrons exhibit kinetic energies of less than approximately 2 eV. A sub-micrometer information depth of this method is observed when investigating the surface of a solid sample covered with a film exhibiting a thickness of some nanometers. This allows for the sensitive detection of thin films on substrates that do not significantly trap electric charges upon photoelectron emission. The interaction of the emitted photoelectrons with the ambient atmosphere is dominated by an electric field effective at the collector of the sensor to an extent that permits sensor-surface distances in the millimeter range during OSEE measurements. Carefully controlling the distance between the sensor and the surface is a prerequisite for effectively applying the setup presented in Fig. 3.26, see Fig. 3.13.

To mount the substrates for analysis, we equipped a commercial OSEE device with an electrically conductive and earthed moving table upon which we positioned the analyte sample using movements in two perpendicular horizontal directions (x and y) under the sensor. The vertical distance between the sample surface and the sensor (z) was set using a micrometer screw attached to the holder of the sensor. The enhanced OSEE setup developed during the ComBoNDT research project permitted vertical sensor movements with the aid of an electric motor, and the most advanced OSEE procedure was achieved with robot-aided three-dimensional sensor positioning. In OSEE implementations depending on a moving table for positioning the CFRP specimens, a surface scan was performed, with the table programmed to move according to a certain step size and number of steps in both horizontal directions, defined by the user through the machine-associated software. As the scan advanced, a photocurrent was obtained for each part of the scanned surface, and an apparently dimensionless value (which actually indicates a current converted into a voltage [7], given in centivolts), henceforth denoted as the OSEE signal, was indicated on a display and digitally recorded. Finally, a digital worksheet with the emission values for the entire analyzed sample, i.e., an OSEE map, was obtained as a result of the test. For further evaluation of these maps especially for non-localized contaminations, the mean value of all the data points in the map together with its standard deviation was calculated.

In the following, we report the OSEE enhancements and findings obtained in the ComBoNDT research project, in which the consortium partners at Fraunhofer IFAM performed the in-process monitoring of CFRP adherends with different shapes that are relevant for distinct technologically relevant user cases and which had undergone an intentional application of various contamination scenarios.

In this section, we summarize the findings highlighting the performance of the OSEE-based ENDT technique for the surface quality assessment of CFRP specimens in distinct user cases.

In a first and trendsetting approach within the ComBoNDT research project, we technologically advanced OSEE and successfully applied this ENDT technique to map CFRP test coupons prepared by intentionally depositing distinct contaminants within several contamination scenarios, among them scenarios comprising combinations of contaminations that might have adverse (and, thus, possibly compensatory) effects on the OSEE signal. Specifically, the reference (RE) states of the production (P-RE) and repair (R-RE) user cases were clearly differentiated from each other, and the suitability of using OSEE to assess the results and reproducibility of a CFRP surface grinding process was indicated. When investigating samples from production scenarios, we applied adjusted OSEE settings that differed from the ones applied for samples from the contamination scenarios of the considered repair use case. In a nutshell, when comparing the sample surface states obtained within different contamination scenarios of the production user case with the ones prepared following the P-RE scenario, OSEE enabled all sample sets from the P-RA, P-MO, and P-FP scenarios as well as the sets with combined contaminations (P-RA-1+FP3 and P-RA-2+FP3) to be differentiated from P-RE. Moreover, within the contamination scenarios of the repair user case, OSEE enabled all sample sets from the R-FP scenarios, most of the samples from the R-DI scenarios (namely those with higher levels of contamination than R-DI-1) as well as sets with combined contaminations (R-TD-1+DI-1 and the scenario R-TD-1+DI-2) to be differentiated from R-RE. However, the surface states of the samples from the R-TD scenarios and samples of the R-DI-1 scenario were not differentiated from those of the R-RE samples. Clearly, if the adhesive joints manufactured based on these adherend surfaces showed inferior mechanical properties compared to joints prepared following the R-RE scenario, then these findings would stand in opposition to recommending OSEE as the ENDT method of choice in this user case. Finally, when referring to the lateral homogeneity of the CFRP test coupons, OSEE indicated several peculiarities: The inhomogeneous grinding of one coupon sample from the repair scenario, the spreading of fingerprinted fluid within the R-FP scenarios, and the position of fingerprints within samples exposed to combined contaminations in the P-RA-1+FP3 and P-RA-2+FP3 scenarios.

Aiming at assessing the surface quality of the pilot level CFRP specimens, we further improved the OSEE technique to permit its use in an industrial environment by incorporating a laser line scanner in the measuring setup to precisely control the distance between the sensor and the studied sample, which might have a more complex shape than a flat sheet. Concerning the investigations conducted on the pilot samples from the production user cases, the OSEE method encountered some severe limiting challenges concerning its use as a measuring system for these kinds of samples. In the case of the pilot samples of the production user case, the closed epoxy film of the peel ply surface prevented a measurable OSEE signal from being obtained. ENDT procedures using OSEE as a measurement technique evaluate relative changes to a given reference signal, e.g., the one characteristic of a clean CFRP surface. Therefore, the procedure was not suitable for assessing the surface quality of the pilot samples of the production user case as the surfaces of the reference substrate did not provide a sufficient signal strength. In our opinion, the encountered challenges, in the form they are presented here, do not appear to be unsolvable if further development is invested in a more customized embedding of the powerful OSEE technique in a setup that enables the elevated positive charging of electrically non-conductive analyte surfaces to be counteracted.

Moreover, we achieved an integration of the OSEE technology in inline applications, e.g., a robot-aided approach, without fundamental constraints. We present an approach to assembling the OSEE sensing technique in a setup that provides further information on the geometrical constellation of the measurement situation. We suggest that a real-time height adjustment and an improved method of signal detection facilitating an automated adjustment of the signal range could provide new ways to use this technique, even on materials that present the challenges observed during the ComBoNDT project. Within the project, it was not possible to implement these improvements. However, as techniques to change the sensitivity of analog–digital converters become more and more available, these challenges could be tackled in a future project.

We would like to underline that OSEE is a highly sensitive inspection method that allows the detection of even very low levels of contamination [1]. Often, a monolayer of a contaminant on an otherwise well-emitting material can be safely detected, as was demonstrated here, e.g., in the case of release agent deposits that we diagnosed using x-ray photoelectron spectroscopy. Nevertheless, establishing OSEE-based ENDT procedures may encounter difficulties with some of the technical surfaces found in industrial applications that, for example, rely on manually performed abrasive surface pretreatment. Such technical surfaces, which are often laterally inhomogeneous in their surface composition or surface conditions, show a considerable variance in their OSEE signal. This may make it difficult to locally differentiate between contaminations and substrate effects. In other words, assessing merely the feature of optically stimulated electron emissivity does not enable the identification of the substantial source of signal variations and provides a signal contrast that is not sufficiently specific to assess analytical challenges, such as the material-related identification of (mixed) contaminants or the essential elements thereof. We estimate that these challenges can be compensated for to a certain level using modern data analysis methods for such applications that benefit from the high sensitivity of the OSEE technique.

Electronic noses (e-noses) are chemical multi-sensor devices, based primarily on microsensors, which are capable of conducting low-cost analyses of complex gas mixtures through chemical fingerprinting. Coupling an array of chemical sensors to pattern recognition algorithms is an idea dating back to the 1980s and began significant development during the 1990s. Due to low overall costs in terms of purchase, operation, and maintenance as well as interesting research results, they looked—and still look—very promising and appealing for the development of industrial-grade ENDTs [9]. A significant number of commercial platforms have been available since 2000 for the comprehensive detection of components in mixtures as well as identification and/or quantification in medical diagnostics, environmental monitoring, and the food industry [10]. Despite some success stories, increasing the application of this technology to a wider range of industrial scenarios and user cases is currently obstructed by technological limitations [11]. Apart from the well-known specificity and sensibility issues that affect single sensors, the variability of fabrication outcomes severely hampers the use of a shared calibration function for e-noses. Thus, ad hoc calibration procedures need to be implemented for each individual e-nose device. The development of calibration transfer strategies could help overcome these limitations but, currently, there are no definitive results [12, 13]. Drift effects (mostly due to aging and poisoning) in sensor devices as well as environmental parameters have affected the mass adoption of e-noses in the high-value production industry because they jeopardize operative requirements. Presently, most research contributions in advancing e-noses still target the headspace analysis of liquid/solid samples in controlled environments, which may amount to underestimating the challenges of on-field applicative scenarios. From the point of view of today’s user, we would like to highlight that on the one hand, specific requirements such as high reliability, fast response, and the possibility to be operated by a chemically non-expert workforce can be, in principle, met by up-to-date e-nose platforms. On the other hand, some instrumental operative requirements, including the need for operation in uncontrolled or even harsh environments [14], may still prove very challenging due to the abovementioned lags and issues.

Conversely, and from a prospective point of view, we observe that along with the requests for novel integrated health monitoring systems, the need for new NDT technologies that should be capable of coping with the new challenges brought by the lightweight aircraft industry is steadily growing [15, 16]. The adoption of lightweight composite materials like CFRP for primary structural components is a major trendsetting milestone and may contribute to a significant reduction of per-mile-passenger transportation costs. It is estimated that this could allow for a dramatic increase in cost efficiency for ground operations (up to 50%), a reduction in fuel usage (up to 20%), and consequently, a CO₂ emission rate reduction at the fleet level of up to 15% [2]. CFRP parts are assembled and joined by adhesive bonding, a critical and special process that requires the adherend panel surfaces to have a high grade of cleanliness. If not, the mechanical properties of the assembly itself could be severely compromised, leading to a high risk of potential structural failure [17]. Surface contamination can be caused by various processes that occur during the assembly or operative life of CFRP panels. Adherends with improperly removed deposits of contaminants, such as hydraulic oil or de-icing fluid, may come in contact with a CFRP surface during aircraft operation, while the release agents used in CFRP molding processes during composite production can severely affect the mechanical properties of a resulting adhesively bonded joint. Such deposits may produce chemical damage to the surface or create a physical screen to the adhesion, while the thermal degradation of the CFRP adherend surface can affect the adhesion properties of the panel, thus compromising the mechanical strength of the bond. Several authors have measured the mechanical parameters of adhesive bonds based on contaminated CFRP panels (e.g., [18]). In particular, Tserpes et al. showed a reduction in the fracture toughness (G_IC) in excess of 25% for Skydrol^®500-B contamination and of more than 60% for significant release agent contamination [19]. Specifically, a 7% silicon-containing substance (calculated as at.% based on XPS investigations) remaining on the surface after demolding can lead to a total lack of adhesion [20].

Bearing in mind these production–technological challenges and the instrumental current state of the art, we anticipate that customized e-noses may represent a suitable technology that can contribute to addressing the lack of a verified procedure for assessing bond quality, which has ultimately been slowing down the adoption of CFRP for primary structures. ENEA and Airbus, both partners in the H2020 ComBoNDT research project [2], have been and continue to be focused on developing e-nose ENDT solutions for surface cleanliness checks to be applied in composite pre-bond quality assurance. In the ComBoNDT project, this endeavor required the design of ad hoc sampling and measurement subsystems, the careful selection of an ad hoc sensor array as well as the design of proper machine learning algorithms to cope with the requirements of this safety-critical application field. The subsequently reported work accounts for the in-project development process, which aimed to reach and demonstrate a high TRL for the use of the e-nose as a detector of surface contaminations before bonding.

In this section, we first detail the experimental methodology for assessing e-nose data as well as the pattern recognition technique. Finally, we present the obtained results for the CFRP specimens from the distinct user cases, according to their shape and the respectively considered contamination scenarios.

Regarding the Airbus e-nose setup, during the different test campaigns, the device proved its capability in terms of contamination detection on the provided samples and chosen application cases. During the coupon sample investigations, the system showed its ability to distinguish the different contaminations in both the production and repair user cases. At the pilot sample level, the desorber measuring head managed to take air-tight odor samples from concave and convex real-world geometries while dealing with new and previously unknown scarfed surfaces. During the three-day test event on technologically realistic parts at IFAM, the detection system substantiated its ability to also function in surroundings other than a gas testing facility and once again showed great sensitivity in detecting volatile chemical compounds.

Quantitative determination of moisture content using the e-nose system in combination with the desorber device is feasible for the desired material. The achieved error of prediction of less than 0.0763 wt% is very encouraging. Enlarging the dataset with additional measurements will further reduce the error of prediction to values of RMSECV (about 0.035 wt%). One big advantage of the e-nose in the combined method of operation is that information on the chemical surface condition (clean or contaminated) can be obtained at the same time with a very good value in terms of the moisture content.

If more precise measurements of the moisture content are required, the measuring setup of the e-nose system can be changed to a pure moisture measurement setup. Under this setup, the resolution of moisture detection will be increased, and the error of prediction can further be reduced.

At the coupon level, the first e-nose version by ENEA was not capable of dealing with the challenges of the production user case, although it was able to reach a 78% correct classification rate for FP detection in the repair user case. After the coupon level tests campaign, the ENEA e-nose was significantly improved by developing a new sampling head and filtering subsystems. The second version proved its capability to detect contaminated samples in both the production and repair user cases at the pilot and realistic parts challenge levels. In particular, the second version was able to achieve more than 80% accuracy in detecting RA-FP mixed contamination samples at the pilot level and obtained a perfect score during the realistic parts testing event in Bremen.

Based on the analysis of these results, e-nose technology appears very close to the maturity stage for the detection of surface contamination prior to the bonding phase. Of course, the e-nose methodology can only be effective whenever residual volatile compounds are present on the surface under analysis. The results of the ComBoNDT campaign are significant for the detection task and are encouraging with respect to the possibility to distinguish different contaminants, even when the concentration level was close to what we would expect to find in real-world scenarios. Quantification capabilities at this level, however, still appear to be difficult to achieve. Currently, the primary limitations of the techniques are due to measurement times, which can slow down the screening of large surfaces.

In summary, during the ComBoNDT project, two different approaches to contamination detection using e-nose were followed. On the one hand, a custom-made-of-the-shelf system (Airsense, Airbus E-Nose) was employed to reach a high TRL in a very short time. The system had to be retrofitted with additional sensors, and in order to achieve stable results on a shop floor environment without the influence of interfering odors from solvents and volatile compounds, a desorber device also had to be developed. On the other hand, the e-nose system from ENEA was developed almost from scratch and went through several optimization steps. Both approaches proved their eligibility and showed outstanding detection capabilities, as described above.

Both approaches offer advantages. The ENEA approach can be considered as an open-source system. The measurement setups (e.g., duration, flow, temperature) as well as sensor signal feature extraction and all other parameters are easy to set and change, whereas the custom-made device can only be controlled or influenced in a way that the manufacturer allows by providing an insight into the device’s firmware. The Airbus approach, employing a working gas sensor system, saved a lot of time, which was used to work on a smart sample-taking device. With the integration of the e-nose and desorber device in one system, interfering influences from a shop floor environment can be ignored. The implementation of signal processing and multivariate data analysis could easily be achieved, tested, and adapted using the “open-source” ENEA system. Measurement control, data recording, feature extraction, and multivariate data analysis were successfully performed, displaying the results on a GUI designed and programmed by ENEA at the end of the project.

As already mentioned above, both approaches have their advantages, and the greatest advancements in the technology were achieved in the areas where research and experience could improve the system via further hardware and software developments. The only conclusion must be to combine these enhancements—smart sample taking and total system control—in an integrated common device to further advance sensor system performance.

LIBS is a spectroscopic method for elemental analysis that is routinely used to determine the elemental compositions of solids, liquids, and gases. For surface technology applications, a high-power laser pulse is focused onto the sample, whereby a small amount of material (typically hundreds of ng to a few µg) evaporates and forms a micro-plasma above the surface [30]. In this way, surface species are excited, and due to the following relaxation process, element-specific radiation is emitted. The emitted light is subsequently separated by its wavelength and is detected using a (often high-resolution) spectrometer; then, the measured intensities of the optical emission are evaluated using dedicated software for qualitative and quantitative analysis. Quantification to assess the surface composition is possible through, e.g., peak ratios in combination with a suitable calibration of the method [6]. For the detection of surface contaminants with LIBS, the contaminant needs to contain (atoms of) the elements that are detectable by means of LIBS and that are not (or only in a very defined amount) part of the clean surface. Detection limits can be in the range of ppm but depend on the material, contaminant, and experimental setup, and they need to be determined separately for each combination of materials and species to be detected [1].

In comparison to conventional surface analysis methods, LIBS requires relatively short measurement times in the order of a few seconds. Measurements can be performed under atmospheric conditions without the need for sample preparation. In addition, LIBS can be adapted to inline applications [31].

Regarding the findings reported here, the LIBS measurements were performed with the LIPAN 4000 system from LLA Instruments GmbH, Berlin, Germany. Laser pulses from a Q-switched Nd:YAG laser with a 1064 nm laser wavelength, a pulse width of 6 ns, and a repetition rate of 20 Hz were used for excitation. In addition, tests were also performed with another Nd:YAG laser system that emits laser light with a wavelength of 266 nm at a rate of 10 Hz with a 7 ns pulse width. Figure 3.53 gives an overview of the LIBS setup.

Spectra were obtained and analyzed using an Echelle spectrometer (LLA Instruments GmbH, Berlin, Germany), which allows simultaneous detection of wavelengths from 200 to 780 nm with a spectral resolution of a few pm. The spectrometer was combined with an ICCD camera (1024 * 1024 pixels). Measurements were typically done at laser energies ranging from 95 to 180 mJ, depending on the laser light wavelength. The control of the spectrum recording and the evaluation of the spectra were performed using the ESAWIN software developed by LLA Instruments GmbH. The software has a large database that allows the automated identification of many relevant atomic emission lines in the optical spectral region.

In the following, we report the LIBS advancements and findings obtained in the ComBoNDT research project, in which the consortium partners at Fraunhofer IFAM performed the in-process monitoring of CFRP adherends of different shapes that are relevant for distinct technologically relevant user cases and had undergone the intentional application of various contamination scenarios.

In the following, we summarize the findings revealing the performance of laser-induced breakdown spectroscopy (LIBS) as an ENDT technique for the surface quality assessment of CFRP composite specimens in distinct user cases. These comprised first a production and a repair user case, both based on coupon level samples, then pilot level CFRP samples with a more complex shape applied in a production and a repair user case, and finally user cases relying on technologically realistic CFRP parts.

In a first and trendsetting approach within the ComBoNDT research project, we applied LIBS to map CFRP test coupons prepared by intentionally depositing distinct contaminants within several contamination scenarios. Some of these scenarios comprised contaminants containing chemical components that feature elements that are not found on clean CFRP substrates and thus, may be used as tracer species for LIBS-based surface assessment procedures. We advanced these spectroscopic procedures to achieve technologically relevant detection limits and to accomplish fast data acquisition and evaluation that permit surface investigations by applying a high area density of LIBS spots. In this way, even locally applied contaminants relevant for aeronautical environments were detected.

In FTIR spectroscopy, molecules close to the surface are excited by infrared light. This excitement results in a partial absorption of the infrared radiation by the molecules, as depicted in Fig. 3.71. The absorbed infrared radiation is missing in the received infrared spectra, providing information about the chemical composition of the surface. The measured FTIR spectra have complex structures that can be interpreted by a partial least squares (PLS) algorithm that correlates the material properties given in the calibration with significant features in the FTIR spectrum recorded on a substrate. In the second step, the calibration can be used to predict the material properties from an FTIR spectrum made on a substrate with unknown properties. The advantage of this evaluation algorithm is that the FTIR spectra can be interpreted quantitatively.

The measurements were conducted with an Exoscan 4100 portable handheld FTIR spectroscope. The test parameter was set to 64 scans at a resolution of 8 cm⁻¹. All measurements were performed in diffuse reflection.

The following subchapter summarizes the results obtained for the three sample levels. Since no sensitivity toward release agent or fingerprint contamination was found during this experiment, a successful evaluation of these samples was not possible; therefore, the results are not described in this chapter.

During the different measurement series, FTIR spectroscopy demonstrated high sensitivity in detecting and quantifying thermally damaged parts, the amount of moisture uptake inside an assembly, and de-icing fluid contaminations on CFRP surfaces.

Moisture uptake could be correlated successfully with an error of 0.09 wt%. This sensitivity is sufficient to set a reliable process window to prevent bondline failure due to unknown moisture uptake.

Coupon level investigations proved the capability of determining thermally damaged samples with an error of 8 °K. This is a very good result for sanded CFRP surfaces. However, this accuracy was not achieved for the pilot samples. At present, it is assumed that the heat friction during the scarfing process leads to high temperatures and causes thermal damage on the sample surface with higher local variations.

On the coupon sample level, the surface contamination with de-icing fluid was well predicted with an error of only 1.4 at.% K. These results were subsequently validated with the contaminated pilot samples, with appropriate separation of lower (DI-I) and higher (DI-II) levels of de-icing concentration. For measurements on the demonstrator part, the de-icing fluid was found on the contaminated area as well as on the reference side (in smaller concentrations). It is assumed that during the drying step in the oven, the de-icing fluid was spread across the whole surface of the sample.

With this newly gained knowledge and the identified tasks, the process of FTIR sampling can be significantly improved. Hence, FTIR spectroscopy has made another step toward finally being employed on the shop floor.

A scanning laser Doppler vibrometer is a non-contact measuring device that makes use of the Doppler effect to register the vibration velocity. A beam of laser light is focused on a point on the measured surface, see Fig. 3.74. The light is reflected and due to the Doppler effect, its frequency is shifted proportionally to the velocity of the measured point. A built-in SLDV interferometer is used to estimate this shift, thus measuring the point velocity. A set of motors and mirrors in the scanning head enables laser beam steering and, together with developed software, scanning along a defined grid of points. Possibilities for SLDV application in research related to guided wave measurements have been extensively studied over recent years. In [33], a combination of experimental analysis using laser vibrometry and numerical analysis for the thin aluminum plate is presented. Ruzzene [34] proposed a technique for full wavefield analysis in the wavenumber/frequency domain for damage detection. This is a filtering technique that improves damage localization results. The author performed numerical and experimental analyses of simple aluminum plates with crack and disbonded tongue and groove joints. In [35], scanning laser vibrometry and imaging techniques were utilized for the detection of hidden delaminations located in a multi-layer composite. The authors analyzed the wave interactions with delamination and utilized several image processing techniques, such as Laplacian filtering. It should be noted that the mentioned work presented a completely non-contact system for elastic wave generation and sensing. Elastic wave generation was performed using a continuous wave (CW) laser source in connection with a photodiode that excited a piezoelectric transducer. Elastic wave sensing was performed using a conventional scanning laser vibrometer. Non-contact elastic wave generation based on a thermoelastic effect was utilized in [36]. The authors utilized broadband excitation based on a low-power Q-switched laser. The research presented here used a piezoelectric transducer (Sonox P502) with a diameter of 10 mm, which was attached to the surface to excite guided waves in the samples. An excitation signal in the form of a 5 µs Hanning was generated by an arbitrary waveform generator and amplified to ±180 V by a signal amplifier. The excitation was synchronized with the SLDV system.

In the research reported in this subchapter, the laser vibrometry measurements were first conducted at points defined along the samples’ diagonal. A signal processing method was applied that allowed the wavenumber to be extracted as a function of frequency. Based on this relationship, the wave group velocity was estimated. The most promising results were observed for the moisture contamination, although the samples were not fully saturated with moisture. The used approach was local, so only a limited part of the sample was measured (diagonal measurement). However, as the waves propagated in the whole sample, the change in the wave field caused by the surface modification may have been too weak to be noticed after the wave had traveled a considerable distance (from the transducer to the edges and then to the measurement point). Considering that the presence of degradation or contamination can have a very slight influence on the wave, a new approach was proposed based on full-field measurements. The time signals were registered in a dense grid of measurement points defined at the sample surface. It was decided to choose an index that could give information about the average state of the surface because it was assumed that the real distribution of the contaminant/degradation is unknown. It was observed that the samples contaminated with moisture clearly differed from the reference samples. However, a sensitivity to the moisture level was not observed in the results. The pilot samples for the production user case were measured from both sides due to their curvature, and the results obtained for the reference samples on the convex side were comparable; however, this was not observed for the concave side. Regarding the contaminated samples, it cannot be clearly stated that they differ from the referential samples. In the case of the pilot samples from the repair user case, the reference samples differ from each other. There is no clear separation between the results for the modified samples and the values for the reference samples. The same approach applied to the realistic part also showed no sensitivity to the contamination of the surface. This could be associated with the type of contamination used.

LIF is a technique based on the analysis of the spontaneous emission of atoms or molecules excited with a laser, whereby the analyzed material can be in the gas, liquid, or solid-state phase. Typical instrumentation for LIF analysis consists of a laser to illuminate the investigated material and a detection system to record the fluorescence. Depending on the chemical composition of the analyzed material, the laser can be tuned to match the wavelength to the absorption lines or bands of specified atoms or molecules, producing an electronically excited state that can radiate. The fluorescence emission can be detected using bandpass filters and sensitive detectors, i.e., photomultipliers in the case of specified spectral band analysis or monochromators equipped with sensitive CCD detectors for full spectra recording. Since the fluorescence signal is much weaker than the excitation laser radiation, a laser cutoff filter is recommended in the detection system. The LIF analysis can be provided in two regimes: (i) the time-integrated mode, in which the spectra are integrated over a long time—in this mode both CW and pulsed lasers can be used for sample excitation; (ii) the time-resolved mode, in which the fluorescence decay is taken into account—this mode requires a short pulse duration of the excitation laser. The fluorescence spectrum is usually represented as a combination of vibrational, rotational, and fine structures, depending on the chemical composition and spectral range. In the case of composite materials characterized by complex chemical composition, the fluorescence bands usually overlap, but subtle changes in the structure of the material can still be seen as changes in the intensity or profile of the fluorescence spectrum.

The main advantages of the LIF technique are its sensitivity, spatial and temporal resolution as well as the non-invasive nature of the analysis. Measurements can be provided for the individual points on the material, or large areas of the surface can be scanned with a laser to visualize the local structure or composition changes [37]. The analysis does not require sampling, and the measurement can be performed on the tested object. The detection of kerosene and hydraulic fluid on CFRP surfaces has been previously reported [38]. Good results were obtained for 266 nm excitation. In another work [39], three wavelengths were studied (266, 355, and 532 nm). The detection of hydraulic fluid, release agent, and moisture contaminations as well as thermal treatment has also been presented. It was shown that at 532 nm, the thermal treatment can be easily distinguished from the remaining cases. Moreover, it was shown that samples treated at 190, 200, and 210 °C are characterized by increasing LIF intensity and can be distinguished from each other and from reference samples.

For the results presented in the next sections of this chapter, the sample excitation was provided by a CW DPSS Nd:YAG laser operating at 532 nm (Spectra Physics). The laser power was set to 0.2 W and the laser intensity was \(1 \,{\text{W}}/{\text{cm}}^{2}\) (5 mm laser spot diameter). The spectra of the laser-induced fluorescence (LIF) were recorded in the time-integration mode. The emission spectra were dispersed by a SR-303i 0.3 m spectrograph equipped with gratings of 600 and 150 grooves/mm and coupled to a time-gated ICCD camera DH 740 (Andor Tech). Spectra were acquired in the range of 300–800 nm with resolutions of 0.3 nm or 1.2 nm. An illustration of the setup is presented in Fig. 3.93.