1 | INTRODUCTION
The global wind energy production has been growing steadily for years as demand for clean renewable energy increase to meet global climate goals. Every year, tens of thousands of turbines are installed world-wide, which makes reliability and low-maintenance operation critical for the scalability and economic viability of wind energy production. One of the most maintenance-heavy parts are the turbine blades. In particular, the leading edge of the blade, which cuts through the air at speeds of around 300 km/h and therefore experiences the strongest forces and harshest conditions of the entire structure. Over time, collisions with rain droplets and other particles may lead to erosion, degrading the aerodynamic efficiency and thereby reducing the efficiency of the turbine. As the length and tip-speed of turbine blades have increased over the years to improve energy capacity per turbine, the challenge of controlling leading edge erosion has become increasingly difficult. One of the frontiers in this endeavor is in designing and testing durable protective coatings that can absorb and dissipate the kinetic energy from these impacts. The current industry standard for testing leading edge protection and for estimation of the coating lifetime is a liquid impingement test known as the whirling arm rain erosion test (RET) [1]. In the whirling arm test, the erosion process is simulated by spinning a blade sample inside an artificial rain field resulting in repeated droplet impacts on the leading edge. The test uses high rotational speed to accelerate the process, causing high stress loading, initiation and propagation of cracks, and eventually complete erosion of the coating. Previous studies have shown that the presence of subsurface defects reduces the mechanical performance of the coating, leading to crack formation and erosion [2,3]. Therefore, it is important to understand the correlation between manufacturing defects and the degradation of the coating performance. For characterization of RET specimens and for quality control in manufacturing of blades, a fast, non-destructive, portable, and cost-efficient scanning technology for coatings is important.
Current methods for subsurface coating inspection include high-frequency ultrasound (HFUS), THz imaging, thermographic imaging, and X-ray computed tomography (XCT). HFUS and THz has the advantage that they can penetrate all the way through the coating and into the underlying filler and composite layers. However, due to their operation wavelength the overall spatial resolution is typically limited to the order of hundred or few hundreds microns [4,5]. In addition, HFUS requires physical contact and a coupling medium, such as ultrasonic gel, which is impractical when scanning curved surfaces. Thermographic methods infer the presence of subsurface defects from the surface thermal radiation, and as such, it is difficult to detect small or vertically stacked defects, and the results are highly dependent on the thermal properties of the sample [6,7]. XCT on the other hand can offer sub-micron imaging resolution and in principle unlimited penetration. However, the technology is limited by long processing time and the need for a rotating system to perform 3D scanning [8–10]. As such, the technique is mostly used as a laboratory technique on small cut-out samples. Furthermore, the use of ionizing radiation not only poses a risk to human health, but also a risk of damaging the sample by breaking chemical bonds and thus creating artificial defects. Another promising technology for non-destructive coating inspection is optical coherence tomography (OCT).
OCT is an interferometric imaging technique based on backscattered laser light from internal microstructures and material interfaces. It was developed in 1991 for biomedical imaging and is widely used today as a diagnostic tool within ophthalmology and dermatology [11–13]. In recent years, the technology has also found applications within non-destructive testing (NDT) due to the possibility of high resolution, non-contact 3D imaging. However, the penetration depth in many materials are severely limited by absorption and scattering, which depend on the type of material and the wavelength of the laser. Consequently, there are only a few examples in the literature where OCT has been applied for the inspection of coatings, primarily for art and cultural heritage preservation, as well as automotive and pharmaceutical coatings. Within automotive coatings, OCT systems operating around the 0.83 to 0.93 µm central wavelength range has been used to map the thickness of individual coating layers with an axial (depth) resolution of 4–6 µm. However, due to the short center wavelength only the top clear coat layer was transparent, limiting the maximum penetration depth to about 100 µm [14–17]. Zhang et al. (2016) characterized automotive paints with metal flakes using OCT at 0.832 µm and were in some cases able to distinguish the clear coat, base coat, and primer layers, each around 20–60 µm in thickness with a depth resolution of 5 µm. In their system, a 2 × 2 mm scan of 1536 × 600 × 2048 pixels was acquired in ~45 s and took ~60 s to process [16]. Moving towards longer wavelengths in the mid-infrared, Cheung et al. (2015) used a broadband supercontinuum (SC) laser to compare OCT at 0.93 and 1.96 µm central wavelength for inspection of artistic oil paintings. They found that despite the lower axial resolution of 13 µm, the longer wavelength was able to penetrate the ~340 µm layer of yellow ochre pigmented paint and provide more structural information about the chalk base layer below [18]. SC lasers in particular has had a major contribution to the development of OCT, since they can provide a high spectral brightness over a broad bandwidth from ultraviolet to mid-infrared, even exceeding that of synchrotron radiation sources [19]. Using a SC laser, Zorin et al. presented improved penetration in oil paints using 4 µm central wavelength, although with a poor axial resolution of 50 µm and a slow line rate of 2.5 Hz [20]. Fast scanning and high-resolution in the mid-infrared was first demonstrated by Israelsen et al., most recently achieving a 3 kHz line rate and 5.8 μm axial resolution at 4.1 µm central wavelength [21,22]. The first study to investigate industrial coatings in the mid-infrared was by Petersen et al. that demonstrated subsurface imaging in marine coatings, including monitoring of wet film thickness during curing of a 210 μm thickness blue-pigmented anti-fouling coating based on cuprous oxide particles, and detection of substrate corrosion through 369 μm thickness white-pigmented high-gloss alkyd enamel [23]. In those marine coatings, the surface roughness and large functional particles presented the main limitations in terms of penetration depth.
So far, there has been no work published with OCT in relation to wind turbine and leading edge coatings. Liu et al. used OCT with a central wavelength of 1.55 μm to monitor delamination growth in an uncoated fiber-glass epoxy composite used for the spar webs in wind turbine blades [24]. They were able to image the delamination through 2 mm of the composite material with an axial resolution of 17 μm and a scan speed of 4 mm/s. In this work, OCT is used to non-destructively inspect coated glass-fiber composite samples to investigate the penetration depth and identify subsurface coating defects. The technique is compared with traditional XCT and optical microscopy to quantify the imaging depth and illustrate the difference in image contrast.
2 | MATERIALS AND METHODS
2.1 | Coated samples
Four coated glass-fiber composite samples were considered for testing. Three of the samples (A, B, C) were rectangular with the dimensions of 40mm x 15mm and varying thickness from 4-11 mm. The fourth sample (D) was curved, as it was cut from a model blade used for RET. The coatings on all samples were made from polyurethane (PU) and applied in varying thickness. For samples A and B, the PU coating was poured onto a horizontal composite panel to obtain a thick layer, while for C and D the coating was applied by airbrush to obtain a smooth and thin layer. Images of the samples are shown in Figure 1, and the sample characteristics are summarized in Table 1.