Nondestructive Detection of Osmotic Damage in GFRP Boat Hulls Using Active Infrared Thermography Methods

Boat building is one of the oldest industries in coastal regions worldwide. Since ancient times, boats have been essential for fishing, as well as transporting people and goods. In modern times, the industry has evolved to meet the growing demand for personal sport and recreational vessels. Since the 1970s, composite materials have become integral to the construction of small boats and luxury yachts, owing to their notable advantages, including reduced weight, enhanced resistance to fatigue and corrosion, and the capability to be molded into intricate shapes [1]. Glass-fiber-reinforced polymers (GFRP) and glass-reinforced polymers (GRP) are the most commonly used materials for boat construction. Throughout their lifespan, boats are exposed to the adverse effects of surrounding water, which causes the onset of osmosis [2]. Minor dynamic impacts can lead to cracks in the protective gelcoat layer of the boat, thereby accelerating the osmosis process. This process begins as soon as the boat is submerged in water, with individual water molecules being small enough to penetrate the composite material layers [Figure 1. Stage 1]. During migration, water molecules react with other chemicals within the laminate, primarily with water-soluble materials such as compounds that hold the mat, or with resin in pockets where the resin has not cured. This way, they form larger molecules such as acids or glycols that can no longer migrate through the GRP [Figure 1. Stage 2]. This process is known as hydrolysis. New water molecules migrate through the hull, moving from areas of lower solute concentration (surrounding water) to areas of higher solute concentration (liquid pockets within the boat’s hull) [Figure 1. Stage 3]. Therefore, the problem of osmosis is more prevalent in fresh water than in salt water. Over time, pressure within the laminate increases, forming blisters on the boat’s hull. These blisters are known as osmotic blisters, and the entire process is called osmosis. When the pressure within the osmotic blister becomes greater than the laminate can withstand, the laminate layers crack–delamination occurs. When kept dry, the osmotic blister dries out and is no longer visible to the eye. However, upon re-contact with water, the water is reabsorbed into the blister, and the process continues. Such hull conditions require timely detection and subsequent repair using known processes. The boats made of GFRP (or GRP) material cannot be environmentally disposed of, so extending their lifespan is of great importance. Also, determining the condition of the boat’s hull is a crucial issue in the boat market, as such boats have often spent some time in dry docks, making osmosis damage no longer visible.

To make hull maintenance possible, it is necessary to ensure a nondestructive method that provides insight into the current condition of the boat’s hull. The methods for nondestructive testing of anisotropic non-metallic materials are not sufficiently developed, and so far, hull maintenance is performed by measuring humidity using humidity gauges, Figure 2 Moisture meters are essential for determining moisture uptake in the laminate and variations across different sections of the hull. Comparisons with areas above the waterline are necessary to establish a baseline. High moisture content does not inherently indicate structural failure. Many older vessels exhibit elevated readings without signs of blistering or delamination. Accurate interpretation of moisture meter readings is crucial. Providing scale conversions and actual moisture content helps clarify findings and avoids misrepresentation. This method is reliable for a certain period, but after the hull is dried, this method does not provide a clear picture of the damage, preventing the evaluation of the boat hull or diagnosing the curing method. The condition of the boat’s hull can also be assessed by hitting it with a hammer and listening to the return echo. The hammer is invaluable for detecting manufacturing faults, delamination caused by blistering, or repair voids. It also helps assess the extent and distribution of delamination across the hull. This requires a highly experienced operator, and even with these requirements met, repeatability of testing is not guaranteed. Thus, the presented approaches to osmosis detection and evaluation based on active infrared thermography are new full-field visualization methods capable of detecting the severity and shape of osmotic damage [3].

The widespread use of glass (fiber)-reinforced polymers in nautical applications requires suitable nondestructive testing (NDT) methods to address structural safety and quality issues, which are of primary concern for customers [4]. The development and application of various nondestructive testing techniques originated in the aerospace industry and have since expanded to fields such as automotive, marine, and construction. Compared to the extensive literature on nondestructive testing of composites, research focused on composites used in marine applications is relatively limited. Often, publications only briefly mention that NDT methods can be applied to composites in marine environments, without detailed studies or specific applications being thoroughly explored [5]. Techniques like acoustic emission (AE), ultrasound testing (UT), infrared thermography (IRT), shearography, digital image correlation (DIC), and X-ray imaging (XRI) are integral to the composite industry. Among these, UT, IRT, and DIC stand out as versatile and cost-effective methods, widely applied in industrial settings and academic research. Each technique offers unique capabilities for detecting and evaluating defects and monitoring damage evolution in composite materials and structures [6]. This study aims to demonstrate the potential of infrared thermography as a nondestructive testing method for detecting osmosis in GRRP boat hulls through the analysis of three detailed case studies. Herein, used methods of postprocessing IR signals are as follows: lock-in thermography, pulse thermography, pulse phase thermography, and gradient pulse phase thermography. Each of the methods have advantages and disadvantages, which will be considered in this paper and supported by examples.

Lock-in thermography was first mentioned by authors Carlomagno and Berardi [7], while its first significant application was achieved in the 1990s by author Busse [6,8]. In this method, the object is stimulated by modulated thermal flow; the thermal wave propagates within the object through its thickness and reflects when it encounters zones where the thermal properties of the material change (local inhomogeneity or a wall). The heating modulation period is selected depending on the sample thickness or the expected depth of the defects. The selected modulation period usually varies from ten to several hundred seconds. This method is limited by the capabilities of the equipment, depending on whether the sinusoid periods can be selected continuously or discretely, and within what range of values [9]. A sinusoidal thermal wave is achieved by regulating the current voltage used to power external heat sources (typically halogen lamps) that heat the surface of interest. Sinusoidal thermal waves propagate within the sample and reflect off the sample walls or defects, thus forming an outgoing sinusoidal wave. The outgoing sinusoidal wave will have the same frequency as the incoming wave (due to the property known as sinusoidal fidelity), but it will have a different amplitude and phase shift [7]. The reflected wave can be recorded by an infrared camera as a temperature change on the surface of the object. If the frequency of the incoming thermal wave and the temperature distribution on the surface of the sample are known, the amplitude and phase can be calculated from four points, equally spaced in time over the modulation period, according to Equations (1) and (2). Thermograms recorded in this way allow direct evaluation of the shape and depth of defects without the need for additional signal processing [10]. where $A_{tw}$ is the amplitude of the sinusoidal wave, ${\varnothing }_{tw}$ is the phase shift in the sinusoid and $S_i$ is the temperature measured by the thermal camera in four points equally distributed over the sinusoid period on the time curve.

Phase pulse thermography (PPT) converts data from the time domain to the frequency domain using the fast Fourier transform (FFT). The resulting data are stored as a 3D matrix, as shown in Figure 3, where $x$ and $y$ represent the spatial coordinates, and $t$ is time. The temperature at a given point on the surface decreases at a rate that can be approximated by the square root of time $\sqrt{t}$, particularly during the initial stages. However, in areas with defects, the cooling rate differs from this typical pattern [11]: where j is the imaginary unit ($j^2=-1$), $n$ denotes the frequency increment ($n=\mathrm{0,1},\cdots ,N$), $\mathrm{∆}t$ is the sampling interval of the thermal images, and $Re$ and $Im$ are the real and imaginary parts of the transformation from which the amplitude $A$ and the phase $\varnothing$ of the signal can be calculated:

The PPT coherently organizes the available information according to increasing frequencies. This method allows for the detection of some new characteristics, unlike previously mentioned methods (e.g., temperature contrast).

After processing the recorded sequence with the FFT algorithm, each pixel in the frequency domain has its own phase profile. Depending on the recording duration and the sampling interval, the phase profiles in the sequence have a certain frequency resolution, $\mathrm{∆}f$ [12,13]: where $w\left(f\right)$ is total recording duration, $N$ is the number of thermograms captured in total, and $\mathrm{∆}t$ is the sampling time interval.

The 1D discrete Fourier transform (DFT) algorithm is applied on a pixel-by-pixel basis, resulting in a 3D complex matrix from which amplitude and phase delay data can be extracted [3]. The DFT is versatile, as it can be used with any waveform and offers the advantage of reducing noise in the signal. While highly effective, Equation (3) is computationally slow. Fortunately, the fast Fourier transform (FFT) algorithm, which can be implemented or found in common software packages (either in full or in a simplified form), provides a faster alternative. The application of the FFT on thermography data were first proposed by Maldague and Marinetti in 1996 [15]. The discrete Fourier transform (DFT) can be used with any waveform of external excitation; therefore, besides PPT, it can also be applied to lock-in thermography (LT) and vibrothermography (VT). The phase of the signal is particularly valuable in thermography, as it is less sensitive to external factors such as variations in emissivity, non-uniform heating of the sample, reflections from the surroundings, and the condition of the surface under inspection.

Figure 4a shows collected thermal images over a time period and temperature decay curves for a defective (red) and a non-defective (blue) pixel. Also, the temperature contrast curve between a defective and a non-defective (green) pixel is shown. As can be seen from Figure 4b, it is obvious that it is not possible to differentiate amplitude profiles for a defective pixel (red) and a non-defective (blue) pixel. The phase profiles for a defective pixel (red) and a non-defective pixel (blue) as well as the phase contrast profiles (green) are shown in Figure 4c. Figure 4c portrays the 3D phase matrix reconstructed from pulsed data using the fast Fourier transform algorithm.

As seen in Figure 4b,c, this kind of function (real) will produce an amplitude and phase response that are even and odd, respectively, with respect to $f=0 \mathrm{H}\mathrm{z}$ (i.e., $n=N/2$) after the application of the FFT. Therefore, from a sequence of $N$ thermograms, there are $N/2$ useful frequency components; the remaining half of the spectrum provides redundant information and can be discarded. Therefore, the entire data sequence can be processed using the FFT, allowing for the reconstruction of both amplitude and phase sequences.

To perform the FFT of thermal data, the continuous temperature signal $T\left(t\right)$, is sampled at $\Delta t$ time intervals and truncated with a rectangular window $w\left(t\right)$. Both $w\left(t\right)$ and $\Delta t$, or using its reciprocal, the sampling frequency $fs=1/\Delta t$, are strongly dependent on the thermal properties of the material being inspected. The appropriate selection of fs primarily depends on the thermal properties of the specimen, but also on a variety of factors that complicate the development of analytical tools [12]. The sampling theorem ($fs \ge 2fc$) should be respected for all defects present on the inspected specimen; the challenge arises in determining $fc$. As a result, fs is generally established empirically by taking some basic guidelines, e.g., high conductivity materials require a higher fs to avoid a loss of information. In addition to being a function of the specimen’s thermal properties, time–frequency duality plays an important role on $w\left(t\right)$ size determination. Frequency resolution $\Delta f$, is directly related to $w\left(t\right)$ by the following equation: $\Delta f=1/w\left(t\right)$ [15].

The gradient-based approach to PPT is an image processing method proposed in [6,7]. The idea came from a standard thermographer’s software tool procedure when analyzing thermograms, i.e., sliding the isotherm. An isotherm in infrared image processing is a line connecting the field of apparent temperatures within a thermal range of interest. The normal to the front line of the isotherm coincides with the temperature gradient. The temperature gradient function may be written as follows: where $T(x,y,t)$ is a temperature of each pixel forming the thermal image at an observed (constant) time t, i.e., coordinates x, y are the horizontal and vertical directions of a pixel position on the pixel map. Such an image processing method enables flattening the differences in thermal signal with a lower gradient, while keeping visible zones with a higher thermal gradient, i.e., it enables visualizing the transition zones between anomalies and an intact object. This method filters out parasitic influences caused by non-uniform heating.

The application of lock-in thermography on three boat hulls in significantly different conditions is described below. In the first case study, it is applied on a motorboat that has been in dry dock for some time. The surface of this boat has been sanded down, and the anti-fouling layer and gelcoat layer have been removed. The hull of the boat has been washed and dried. In the second case study, the testing was conducted on a motorboat that was just taken out of the sea. The osmotic damage is noticeable on the hull. On this hull, in addition to nondestructive testing using active infrared lock-in thermography, nondestructive testing with ultrasonic control was also performed. In the third case study, an image processing method based on the fast Fourier transformation (FFT) approach is presented with the goal of enabling the location of damaged zones on a real-life boat’s hull, which is covered with several layers of anti-vegetative coatings that make detection of osmotic blisters challenging.

Figure 5a shows a motorboat with a severely damaged hull. The measurement is performed on a dry GPR surface where the anti-fouling paint layers and the gelcoat layer have been removed. After washing and neutralizing the acidic phase, the boat hull was dried for several months. Figure 5a shows the equipment setup for the measurement, and Figure 5b shows the region of interest (ROI).

During the measurement, the hull surface is heated with two halogen reflectors, each with a power of 0.5 kW. The surface heating is performed periodically, as shown in Figure 6. The sinusoidal thermal wave of the halogen reflectors is achieved through control by an industrial relay ELMARK ZG1NC-2-20D, which is connected to the control circuit. Changing the period of the sinusoid is possible in discrete steps. This control allows for a minimum sinusoid period of 0.1 s. The maximum period that can be achieved is 200 h, although such a long period currently has no practical application, Figure 6. The thermographic sequence was recorded with a mid-wave infrared camera FLIR SC 5000 (Teledyne FLIR, Wilsonville, OR, USA) with a resolution of 320 × 256 pixels, a sensitivity of 0.02 K, and a maximum acquisition frequency of 150 Hz.

The duration of one heating and cooling cycle of the surface, excitation frequency, recording time, sampling frequency, and the number of thermal images are provided in Table 1.

Influences such as uneven heating, environmental reflections, and surface coatings degrade the thermal image (Figure 7), making damage difficult to discern. Figure 7 depicts the unprocessed thermal image where a small portion of the damaged area is visible (area B in the image).

The damaged surface of the ship’s hull was recorded with an infrared camera three times at different heating periods (Table 1), resulting in three sequences of thermal images. In the data processing using lock-in thermography, four thermal images were utilized from each sequence. Signal processing was conducted using the Matlab R2010b software package. The processing time was several minutes on a standard computer. For each sequence of thermal images, the amplitude and phase of the sinusoidal wave were calculated according to Equations (1) and (2).

Figure 8, Figure 9 and Figure 10 illustrate the obtained phase shifts after thermal excitation with different periods on the inspected ship hull. In the phase shift shown in Figure 8, the modulation period is the shortest (P = 24 s), corresponding to the highest modulation frequency. The entire hull surface was scratched after removing the gelcoat manually with a grinder, making it difficult to differentiate osmotic damage from surface damage.

In the phase image in Figure 9, the modulation period is P = 72 s, corresponding to the mid-range frequency. The thermal wave has enough energy to penetrate the material. The marked indications represent deeper osmotic damage, and the effects of surface grinding are less pronounced.

In the phase image in Figure 10, the modulation period is P = 120 s, corresponding to the lowest frequency. The thermal wave has high energy, penetrating deep into the material. In images captured in this regime, deep damage caused by osmosis is clearly visible. The depth of osmotic damage indicated by marks A and C in Figure 10 corresponds to the depth at the recording frequency, f = 0.0083 Hz. The phase shift in the osmotic damage marked by label B is strong, indicating that the damage has penetrated into deeper layers of the material. To precisely determine the depths involved, recordings should also be made at lower excitation frequencies. However, unfortunately, this was not feasible with the limitations of used modulation period equipement.

The detection and assessment of osmotic damage in GRP boats require a fast and reliable NDT method. Lock-in thermography is an NDT method capable of visualizing osmotic processes in composite structures where conventional approaches (such as moisture measurement) may lack accuracy. The presented approach is based on the Lock-in method where the heating period is selected based on thermal properties, thickness of the tested material, and the expected depth of damage. To detect damage at various depths, it is necessary to conduct several tests with different excitation periods. The method yields good results if the appropriate excitation periods are chosen for the range of depths where damage is expected to be found.

The method was applied to a motorboat just pulled out of the sea onto dry dock. The boat hull is protected by a gelcoat and protective paint, as shown in Figure 11a. Osmotic damage is filled with liquid, so it can be seen with the naked eye shortly after the boat is removed from the water, as depicted in Figure 11b.

Ultrasound testing is a widely utilized nondestructive testing (NDT) method for assessing the structural integrity of composite materials, including glass-reinforced polymer (GRP) used in boat hulls. This technique relies on the propagation of longitudinal ultrasonic waves through the material, with their velocity and attenuation providing critical insights into the material’s condition. Osmosis was confirmed both by lock-in thermography and A-scan ultrasonics, using a low-frequency probe. Since the hull is made of thin laminate, a special plexiglass attachment is needed for ultrasonic testing to concentrate the soundwave beam near the surface of the laminate. In ultrasonic testing of metal materials, higher frequency probes are usually used. Such probes penetrate the material with greater energy and provide a sharp output signal. A comparison of the output signals from a flat 4 MHz probe and a flat 1 MHz probe on calibration steel block K1 is shown in Figure 12. Generally, lower frequency probes are used only when readings cannot be achieved using higher frequency probes; for example, in ultrasonic testing of non-homogeneous materials (composites, gray cast iron, heat-affected zones, and Duplex steels).

The device used for ultrasonic testing is the GE Krautkramer USM GO (GE Inspection Technologies, Hürth, Germany), as shown in Figure 13a. The probe used for testing GRP is the K1S-C 1 MHz frequency probe GO (GE Inspection Technologies, Hürth, Germany), depicted in Figure 13b. The soundwave of this probe has a longer period, making it easier to bypass obstacles resulting from material inhomogeneities. In water, the velocity of longitudinal ultrasonic waves is approximately 1480 m/s, whereas in GRP, the velocity typically ranges between 2500 m/s and 3000 m/s, depending on the laminate’s specific composition and manufacturing process. This disparity is a key factor in interpreting ultrasonic signals, as the transition between different media significantly influences wave behavior. For instance, the presence of moisture within a GRP hull diminishes sound conductivity, as soundwaves travel faster through GRP than through liquid, affecting the signal shape and attenuation.

By calibrating ultrasonic equipment to match the material properties of the GRP, defects such as osmotic damage, delamination, or voids can be accurately detected. This makes ultrasound testing an indispensable tool in marine applications, where ensuring the structural reliability of boat hulls is critical for safety and performance. The sound velocity was calibrated on a vacuum-molded sample of fiberglass, 10 mm thick, as shown in Figure 13b.

Figure 14 is an A-scan of the glass-reinforced polymer hull of the boat. The signal peaks represent reflections from different layers of the laminate, such as anti-fouling paint, epoxy primer, gelcoat, mat layers, and the inner gelcoat layer. In Figure 14a, signal loss, i.e., the zone without reflections, represents the beam passing through osmotic damage, followed by signal reflection from the laminate layers again. The shape and attenuation of the signal after the A-scan depend mainly on whether the vessel has recently been pulled onto dry dock (the case shown) or has been out of the water for some time. Sound conductivity is better through a wet hull, as soundwaves travel faster through a liquid medium than through air. After a few weeks, the boat hull will completely dry out, and osmotic damage will no longer be visible. The signal peaks at a depth of around 10 mm are reflections from the rear walls of the hull, Figure 14b. The measured values (depths) are not entirely accurate because the speed of the ultrasonic wave was calibrated on a vacuum-laminated reference block, while the hull of this boat was made by hand laminating.

In Figure 15, phase shifts are shown for different excitation frequencies. Osmotic damage in zones A and B (osmotic damage A and B in Figure 11) are located in laminate layers near the surface.

Damage A is closer to the surface and shallower than damage B, so at the excitation frequency of f = 0.0083 Hz, damage A is not visible. The depth to which the hull is damaged in zone A is between the excitation frequencies of 0.0083 Hz and 0.0139 Hz. Since damage B is visible even at the excitation frequency of 0.0083 Hz, to locate the total depth to which the hull is damaged in zone B, even lower excitation frequencies need to be used. Lower excitation frequencies correspond to longer wavelengths of the thermal wave. To investigate deeper material layers, higher energy input is required (longer heating period, or lower frequency). Due to equipment limitations, the lowest frequency with which this testing can be performed is 0.0083 Hz.

Figure 16 shows the area where damage A and B are located after removing the anti-fouling paint and protective epoxy coating. In one part of the hull, a cluster of smaller osmotic damage was observed, as shown in Figure 17a. After grinding the anti-fouling paint and epoxy substrate, it was revealed that the osmotic damage was between the layers of gelcoat and epoxy coating, as depicted in Figure 17b. The damage did not affect the hull of the vessel.

The surface covered with minor osmotic damage was inspected using lock-in thermography. Phase shift images show damage in the surface layer of the laminate, as shown in Figure 18. Damage is not present at greater depths, as seen in the phase shift captured at the excitation frequency of 0.0083 Hz.

The detection of osmotic damage in polymer composites presents a significant challenge due to the weak signal response, both in ultrasonic testing and active infrared thermography. This chapter has highlighted the applicability of the lock-in thermography method for detecting osmotic damage on composite boat hulls. The phase delay data obtained through LT is particularly valuable in nondestructive testing, as it is less influenced by issues such as noise and external factors compared to raw thermal data. This makes LT a highly effective diagnostic tool, not only for qualitative inspections but also for the quantitative characterization of materials [11,12].

The image processing method based on the fast Fourier transformation (FFT) approach is presented with the goal of enabling the location of damaged zones on a real-life boat’s hull, which is covered with several layers of anti-vegetative coatings that complicate detection, Figure 19. The presented research is conducted with a 1 kW halogen lamp and a cooled FLIR SC 5000 MW InSb 320 × 256 pixels IR camera (Teledyne FLIR, Wilsonville, OR, USA). The optimal position of the camera and the lamp is evaluated. According to the conducted experiments, the best thermal contrast is achieved when the lamp is positioned normal to the surface within the distance of 1 to 2 m. These conclusions are valid for observed GFRP structures and a 1 kW halogen lamp. In here-presented examples, the camera was positioned 1 m from the surface (the distance depends upon the specimen size and the angle of the camera lens) and with a 15° offset from the normal direction. This camera offset is particularly important when using the cooled MW camera where the focal point array (FPA) reflection cooled down to −200 °C will influence the image. For LW bolometric cameras, the 15° is a strong recommendation. The hull was heated for 2 min, and after that the hull was left to cool off for 3.33 min. The cooling process is acquired by the cooled MW thermal camera (Figure 19a). The cooling dynamics and thermal contrast ΔT (i.e., the difference between two curves) depends upon the damage depth. The thermal contrast ΔT between two observed points diminishes during the time. The thermal contrast between the damaged and undamaged regions can be observed during the heating, but the reflections of the heating lamp are strongly influencing the image. The strongest thermal contrast is achieved at the beginning of the cooling process (Figure 19c). Due to the several layers of anti-vegetative coating, it is hard to distinguish delaminated zones from the paint layers.

The FFT algorithm is applied on the recorded thermal sequence. The displayed diagram of the cooling process consists of $N=200$ frames acquired at sampling frequency $fs=1 \mathrm{H}\mathrm{z}$, i.e., with time resolution $\Delta t=1 \mathrm{s}$. This corresponds to truncation window size $w\left(t\right)=200 \mathrm{s}$ which is more than enough to ensure that even the deepest defects are not missed. The frequency resolution is equal to $\Delta f=1/w\left(t\right)=0.005 \mathrm{H}\mathrm{z}$. According to the Nyquist–Shannon sampling theorem, the frequency is $fc=fs/2=0.5 \mathrm{H}\mathrm{z}$ which reassures that the phase contrast is zero at a frequency lower than $fc$, i.e., the time resolution is correctly chosen. Figure 20 portrays the phasegrams reconstructed from the pulsed data using the FFT algorithm. As is already theoretically analyzed and shown in Figure 4, it is enough to take into consideration only the phasegrams in the right-hand plane of Figure 20b. Based on the phasegrams in Figure 20b, it is obvious that blister detection zones strongly depend on the observed frequency. The shallower zones of blister osmosis are easier to detect on the higher frequencies and vice versa.

The boat’s hull phasegram with a maximum phase contrast between blister osmosis zone B and zone C and the sound boat’s hull (f = 0.01 Hz in Figure 20b) is shown enlarged in Figure 21. Thus, due to the lack of parasitic influences (non-uniform heating, emissivity variations, and reflections from the environment and atmosphere) defect blister osmotic zone B and zone C are clearly distinguished among the sound boat’s hull.

The higher frequency phasegrams (f = 0.075 Hz in Figure 21c) are revealing damage closer to the surface (zone A), while the lower ones (f = 0.01 Hz in Figure 21a) reveal damage deeper in the material (zone B and zone C). As a consequence of signal processing with the FFT, the noise increases at higher frequencies.

The thermogram acquired at the first frames of the cooling process (Figure 22a) is affected by the uneven heating due to the lamp position and convexity of the boat’s hull. Due to the several layers of anti-vegetative coating, it is hard to distinguish delaminated zones from the paint layers. The thermal gradient (Figure 22b) eliminates these influences, enabling a clearer image of the structure’s thermal response.

In Figure 23, the gradient Function (7) is applied to the phasegram image shown in Figure 21a. The location of osmotic blisters near the trailing edge is more embossed, i.e., pronounced, and the damaged zones can be even easier to observe than in Figure 21a.

The locations of the detected damaged zones have been verified by removing the material. The material affected by osmosis is characterized by delaminated “dry” looking zones visually appearing like the resin is not present in the composite (Figure 24).

There are examples in [6,7] where gradient-based image processing has been used as an image analysis tool for the better detection of structural reinforcements of wind turbine blades and glider planes. Moreover, glass-reinforced polymer composite structure anomalies such as delamination, and cracks or air pockets trapped in gelcoat can also be detected. When applied to submerged boat hull structures, due to the remains of several anti-vegetative paint remains, it is hard to distinguish the damaged zone from the influence of the paint remains, which was not the case for the wind turbines and glider pane structures, where the surface conditions are significantly better.

This study demonstrates the effectiveness of advanced nondestructive testing (NDT) methods, particularly in addressing challenges associated with detecting defects and osmotic damage in composite materials like GFRP and CFRP. Each of these methods has advantages and disadvantages, as summarized in Table 2. For each method mentioned, it is necessary to determine the areas of application and the capabilities of the method depending on the available equipment and the physical and geometric characteristics of the type of boat hull being tested. In future research, it would be beneficial to try to eliminate the mentioned shortcomings of the methods by combining known or new methods.

The application of the pulsed phase thermography (PPT) method, using FFT-based signal processing, successfully mitigates parasitic effects such as non-uniform heating, emissivity variations, and environmental reflections. The phasegram analysis confirmed that defect depth is frequency-dependent, with higher frequencies being more sensitive to shallower defects.

A novel gradient pulse phase thermography (GPPT) method was introduced to enhance localization and assessment of osmotic damage in boat hulls. Experimental validation on osmotic blistering in GFRP hulls highlighted the method’s superior diagnostic capabilities compared to standard phasegram analysis, providing clearer visualization and better evaluation of affected zones.

This research underscores the ecological and economic importance of extending the service life of GFRP structures through effective damage detection and maintenance strategies. By offering improved diagnostic accuracy, the proposed methods support more informed decisions about repair versus replacement, particularly for vessels where full osmosis treatment may not be economically viable. Periodic monitoring remains crucial, with future research directed at optimizing these techniques and integrating new approaches to overcome existing limitations.

In the end, the osmosis process will depend on the owner’s usage patterns. Prolonged immersion, seasonal use, and the type of water (fresh vs. salt) significantly impact the likelihood of blistering. Economic considerations must inform recommendations. In smaller vessels, the cost of full osmosis treatment may exceed the boat’s market value. If no structural degradation is evident, periodic monitoring (e.g., every five years) is a more practical approach.