Thermal Non-destructive Testing and Evaluation for Inspection of Carbon Fiber Reinforced Polymers

1. Introduction

Defects that occur during the design, development, or manufacturing phases due to process imperfections or those introduced during actual use can compromise a device's durability and functionality, because it will be unfit for its intended purpose and leads to the failure. Consequently, the existence of defects can result in severe consequences, such as product loss, a shorter lifespan for the component, environmental harm, repair expenses, and increased costs. Detecting these defects not only prevents failures in critical components during their use but also holds the potential to restore and repair these components by addressing the defects. The widely used Non-destructive Testing and Evaluation (NDT&E) techniques are Magnetic Particle Testing (MPT), Electromagnetic Testing (ET), Visual Inspection Testing (VT), Acoustic Emission Testing (AET), Thermal/InfraRed Imaging (IRI), Ultrasonic Testing (UT), Radiographic Testing (RT), Dye Penetrant Testing (PT) and. Depending on their complexity, these methods are often categorized and classified as either classical or innovative solutions. Non-Destructive Testing and Evaluation (NDT&E) typically serve as necessity for various purposes such as qualifying raw materials before processing, inspecting sub-components, finished products during and after manufacturing and maintenance phases.

In recent years, InfraRed Imaging (IRI) for Thermal NDT&E (TNDT&E) has gained immense popularity for its use for inspecting and characterizing a wide range of industrial and biomedical materials [1,2,3,4,5,6]. This approach revolves around mapping the temperature distribution across the test sample to identify hidden irregularities, assess their properties, and map stress patterns [6,7,8,9,10]. Due to substantial capabilities of IRI, diverse applications found in various industries, including aeronautics, civil engineering, electrical engineering, mechanical engineering, automotive, and biomedical [11,12,13,14,15].

Among the various Non-Destructive Testing and Evaluation (NDT&E) techniques, Infrared Imaging (IRI) has gained recognition as a dependable inspection method. This is because it offers, non-invasive, whole-region, rapid, safe, and non-contact inspection capabilities. IRI can be applied in either a passive or an active mode for NDT&E purposes [14,15,16,17,18,19,20].

Passive infrared imaging involves capturing the inherent heat distribution of a test object without any external stimulation. This method is effective when the object naturally exhibits significant thermal variations. It finds applications in various scenarios such as detecting heat loss in structures, monitoring corrosion in pipe walls within thermal power plants, measuring power dissipation in electronic chips, identifying building seepage issues and many more. But the biggest disadvantage of this passive mode is its inability to provide clear thermal contrast for deeper subsurface anomalies and its limited capacity for quantitative analysis. These constraints restrict its applicability in certain situations.

Whereas Active thermal imaging [20,21,22,23,24,25,26] employs a deliberate and controlled heat stimulus to introduce thermal energy applied to the test specimen which have predefined characteristics like amplitude, phase and bandwidth. To reveal subsurface deviations that are located deeper and have relatively smaller lateral dimensions, various signal, image, and video processing methods are applied to the temporal thermal response data recorded from the test object.

Compared to the Passive approach, Active thermography is preferred because it offers substantial enhancements in thermal contrast, particularly when it comes to subsurface anomalies. Additionally, it provides quantification capabilities and improved ability to analyze deeper layers within the test object, making it a valuable technique for inspections that require thorough depth analysis.

2. Active Thermal Infrared Imaging Techniques

In the realm of active infrared imaging methods for TNDT&E applications, three of the most employed techniques are Pulse Thermography (PT), Lock-in Thermography (LT), and Pulsed Phase Thermography (PPT) [27,28,29,30,31,32,33].

PT finds out as the quickest and simplest method compared to other thermal non-destructive testing and evaluation (NDT&E) techniques [1,3,5]. In PT, a brief and intense heat stimulus with high peak power is applied to the test specimen, and its temporal thermal distribution is typically observed during the cooling phase of the sample. The presence of defects leads to differences in thermal properties between the defective and non-defective areas, creating thermal contrast on the test object. However, there are limitations to the applicability of PT. It requires high peak power heat sources to generate the necessary high-frequency thermal waves for detecting subsurface defects with sufficient resolution. Additionally, challenges related to inconsistent emissivity and unequal heating on the object can significantly impact the temperature distribution over the test sample, further restricting its use.

Lock-in Thermography (LT) takes a continuous, single-frequency heat source with relatively low to medium peak power when applied to the test object [2]. This heat source generates and sends thermal waves into the test object. Subsequently, phase analysis is conducted on the thermal response data obtained (as illustrated in Figure 1(a)) to uncover subsurface information about the test object. This phase analysis can be carried out using approaches such as Fourier transform or phase shifting. This makes LT a time-consuming approach to resolve the defects located at various depths of different lateral dimensions due to mono-frequency sinusoidal thermal distribution (Figure 1(b)). The autocorrelation of the mean removed thermal distribution of the Figure 1(a) is shown in Figure 1(c).

Pulsed Phase Thermography (PPT) shares similarities with Pulse Thermography (PT) in terms of the experimental setup. However, its analysis differs as it involves the application of Fourier Transform (FT) applied to the temporal thermal data recorded for each pixel during the cooling phase [4,5]. PPT combines the benefits of the wide bandwidth found in PT with the advantages of phase-based analysis seen in LT.

Frequency Modulated Thermal Wave Imaging (FMTWI) [16,18,21], probes thermal waves into the test specimen within a desired band of frequencies in a limited time determined by thermo-physical properties of the sample and its dimensions, as depicted in Figure 3(a). The frequency and its auto-correlation responses are illustrated in Figure 3(b) and Figure 3(c), respectively. FMTWI stands out as an economical and feasible experimental technique that addresses the restrictions related to the high peak power heat sources needed for the pulse-based approaches (PT and PPT), as well as the need for repetitive experimentation in LT to detect defects at various depths with sufficient resolution. The ability to sweep the desired band of thermal wave frequencies ensures the complete depth scanning of the specimen under examination in one run, even when using low peak power heat sources. This makes FMTWI a practical and efficient solution for inspecting objects with varying depths and spatial dimensions.

Figure 1. Illustrates (a) mean zero temporal thermal distribution for an incident sinusoidal heat flux onto the test specimen (b) frequency response (of Figure 1(a)) and (c) auto-correlation response of the Figure 1(a).

Figure 1. Illustrates (a) mean zero temporal thermal distribution for an incident sinusoidal heat flux onto the test specimen (b) frequency response (of Figure 1(a)) and (c) auto-correlation response of the Figure 1(a).

Figure 2. Illustrates (a) thermal response over the object for an incident pulse shaped heat flux on to the test specimen (b) Frequency response (of Figure 2(a)) and (c) the Auto-correlation response of the Figure 2(a).

Figure 2. Illustrates (a) thermal response over the object for an incident pulse shaped heat flux on to the test specimen (b) Frequency response (of Figure 2(a)) and (c) the Auto-correlation response of the Figure 2(a).

Figure 3. Illustrates (a). an aperiodic linear frequency modulated thermal response for an incident frequency modulated heat flux with frequencies varying from 0.01 Hz-0.1 Hz for duration of 100 s (b) frequency response of Figure 3(a) and (c) the auto-correlation response of the Figure 3(a).

Figure 3. Illustrates (a). an aperiodic linear frequency modulated thermal response for an incident frequency modulated heat flux with frequencies varying from 0.01 Hz-0.1 Hz for duration of 100 s (b) frequency response of Figure 3(a) and (c) the auto-correlation response of the Figure 3(a).

Regarding post-processing methods, the commonly employed traditional frequency domain phase-based analysis distributes the applied energy onto the examined sample into distinct frequency components [23,24]. However, this approach has limitations in terms of resolution and sensitivity for detecting subsurface defects using phase information from a specific frequency component, as depicted in Figure 4. A more effective alternative is the matched-filter based time-domain analysis, which includes correlation coefficient and phase analysis [24,25] as illustrated in Figure 5, outperforms the conventional method by providing improved results for detecting subsurface defects.

In matched filter-based time-domain data processing methods, the energy supplied within a specified frequency band during a predefined time is condensed into a narrow time duration, rather than being spread out among individual frequency components, as is the case with frequency domain analysis. These matched filter-based time-domain approaches, which include pulse-compression based correlation coefficient and time domain phase analysis, have shown promising results for defect detection when compared to the classical post processing approaches. The time domain phase and pulse compression based correlation co-efficient methods have been investigated to find their potential in detecting the defects.

The matched filter-based approach discussed in this context involves correlation-based post-processing data analysis. It offers several advantages, such as being resistant to multiplicative noises that can occur during experimentation from factors like inconsistent illumination and variations in emissivity over the sample surface. Also, it is robust against additive noise introduced during the detection process as well as thermal noise.

3. Results and Discussions

Experiments using LT, PPT, and FMTWI have been conducted on a Carbon Fibre Reinforced Polymer (CFRP) test specimen with blind holes as shown in Figure 6(a), to investigate the defect detection capabilities of commonly used active infrared imaging techniques and their associated post-processing approaches. Further, the frequency (phase) and time domain (correlation coefficient and phase) based post-processing schemes have been adopted. PT has been performed by using two halogen lamps each 2 KW peak power with a duration of 12 s. Moreover, post-processing strategies based on the frequency and time domain (phase and correlation coefficient) have been implemented. PT has been performed by using two halogen lamps each 2 KW peak power with a duration of 12 s. The temporal temperature distribution is performed for 100 s using a capturing frequency of 20 Hz. In contrast, LT and FMTWI were performed for 100 s using a 0.05 Hz modulation frequency and a 0.01 Hz to 0.1 Hz sweep frequency, respectively, while maintaining the same capturing frequency (20 Hz) as PPT. Two 1 KW peak power halogen bulbs have been used to achieve this. The acquired zero mean experimental thermal responses have further been subjected to frequency domain (stack of phase image sequence) and time-domain based data processing (stack of correlation coefficient and phase image sequence). Reconstructed stacks of thermograms using the suggested frequency and time domain data processing techniques are displayed in Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12, Figure 13, Figure 14 and Figure 15. Experiments for the LT (Figure 7, Figure 8 and Figure 9), PPT (Figure 10, Figure 11 and Figure 12), and FMTWI (Figure 13, Figure 14 and Figure 15) were used to gather the results.

The analysis of thermal stacks reconstructed from the lock-in thermographic (LT) results reveals several noteworthy observations such as the time domain phase images (Figure 8) show the better spatial thermal contrast than that of the widely used frequency domain phase approach Figure 7). Further it is clear from the thermal images reconstructed from the correlation coefficient (Figure 9) approach on LT data at different cross-sectional areas failed to provide enough contrast to visualize the blind holes. This is due to the sinusoidal thermal excitation of LT, as shown in Figure 9(c). Furthermore, it can be shown from the thermal pictures of stacks that were reconstructed for various cross-sectional views from the obtained PPT thermographic data that the time domain phase images (Figure 11) have better spatial thermal contrast than the frequency domain phase method (Figure 10), which is often employed. Additionally, it is evident from the thermal stacks that were rebuilt using the correlation coefficient method (Figure 12) and PPT data at various cross-sectional regions that insufficient contrast was provided to enable the visualization of the blind holes. This is due to the pulsed thermal excitation of PPT, which is not a pulse compression favourable excitation scheme. However, compared to the traditional frequency domain phase techniques (Figure 13), the processed data of FMTWI with correlation-based data processing shows superior depth as well as spatial resolvability (Figure 15). Nevertheless, the obtained stack from the time domain phase images still provides better spatial contrast to detect the hidden sub-surface defects as shown in Figure 14.

4. Conclusions

This discussion has focused on the application of widely used thermal imaging modalities (LT, PPT, and FMTWI) and their associated post-processing methods for non-destructive testing and evaluation of Carbon Fiber Reinforced Polymer (CFRP) specimens containing blind holes as defects. Also, the superior sub-surface anomaly visualization capabilities offered by phase images retrieved from the time-domain analysis over conventional obtained phase images from the frequency-domain based analysis has been highlighted in the results. It also emphasizes that the best defect identification is achieved with a correlation coefficient-based post-processing approach, which is suitable for excitation schemes that benefit from pulse compression (such frequency-modulated incoming heat flux over the sample). This method provides improved depth resolvability together with enhanced spatial contrast, making it a useful tool for detecting flaws in CFRP specimens.