In the realm of advanced manufacturing, particularly for high-temperature applications such as gas turbine components, the integrity of materials is paramount. Precision investment casting is a critical process for producing complex, high-performance parts from superalloys, like the nickel-based K438 alloy, which exhibits excellent heat and corrosion resistance. However, the very nature of precision investment casting, with its intricate molds and solidification dynamics, inherently introduces the risk of defects such as cracks, shrinkage porosity, inclusions, and misruns. These defects can severely compromise the mechanical performance and service life of components operating under extreme conditions. Therefore, developing reliable, efficient, and high-resolution non-destructive testing (NDT) methods is essential for quality assurance. Traditional film-based radiography, while established, has limitations in terms of processing time, chemical waste, and dynamic range. Computed Radiography (CR), a digital radiography technique, offers a compelling alternative with advantages in speed, image quality, environmental friendliness, and digital post-processing capabilities. This study delves into the application of CR for inspecting precision investment casting superalloy sheets, with a focus on optimizing detection parameters for crack defects and enhancing image interpretability through advanced processing techniques.
The fundamental principle of CR involves the use of a photostimulable storage phosphor imaging plate (IP) instead of radiographic film. When exposed to X-rays, the IP stores energy in trapped electron states. This latent image is then read out by a scanning laser beam in a CR reader, which stimulates the release of stored energy as visible light, captured by a photomultiplier tube and digitized. The resulting digital image can be manipulated, analyzed, and stored electronically. For the inspection of precision investment casting components, which often have varying thicknesses and complex geometries, CR provides a wider dynamic range, allowing for the visualization of features in both thick and thin sections of a single exposure. The core objective of this research is to establish a robust CR inspection methodology specifically tailored for equiaxed grain superalloy materials produced via precision investment casting. This involves creating reliable exposure curves, determining the detection sensitivity for artificial crack defects, and applying image enhancement algorithms to improve defect conspicuity.
The test specimens were fabricated using standard precision investment casting techniques to replicate the microstructure and potential defect scenarios of actual components. The base material was a K438 nickel-based superalloy, cast into flat plate geometries. To simulate service conditions, some plates were coated with a Thermal Barrier Coating (TBC) system, consisting of a bond coat (approximately 150 µm thick, composition: Ni-22Cr-9Al-37Co-0.5Y) and a ceramic top coat (approximately 350 µm thick, composition: 7-8 wt.% Y2O3 stabilized ZrO2). Artificial surface-breaking cracks were introduced into the substrate surface prior to coating using electro-discharge machining (EDM). The dimensions of these cracks were meticulously controlled, as summarized in Table 1. Additionally, a multi-step wedge block, also made from the same precision investment casting K438 superalloy, was produced to facilitate the development of exposure curves. Its dimensions are detailed in Table 2.
| Crack Identifier | Depth (mm) |
|---|---|
| #1 | 0.2 |
| #2 | 0.4 |
| #3 | 0.6 |
| #4 | 0.9 |
| #5 | 1.2 |
| Step Number | Material Thickness (mm) |
|---|---|
| 1 | 2 |
| 2 | 4 |
| 3 | 6 |
| 4 | 8 |
| 5 | 10 |
| 6 | 12 |
| 7 | 14 |
| Backing Plate | 14 |
The inspection system comprised a directional X-ray generator (model analogous to XXG2505), a high-resolution CR scanner (model analogous to Crown), and a calibrated medical-grade display (model analogous to MDRC-2120) for image evaluation. The experimental setup followed a standardized radiographic geometry. To ensure the maximum applicability of the technique for detecting surface defects, the artificial cracks and the image quality indicators (IQIs), or wire penetrameters, were always placed on the source side of the specimen. The source-to-detector distance (SDD), also referred to as focus-to-film distance (FFD) in this context, was a key parameter. The geometric unsharpness (Ug) is given by:
$$ U_g = \frac{f \cdot t}{d} $$
where \( f \) is the effective focal spot size, \( t \) is the distance from the source side of the object to the detector, and \( d \) is the distance from the source to the object. Keeping \( U_g \) minimal is crucial for high-resolution imaging. In our experiments, we maintained a constant FFD of 700 mm, which, combined with a small focal spot, ensured geometric unsharpness was below levels that would significantly degrade sensitivity for the crack sizes of interest.
The cornerstone of reproducible radiography is a well-defined exposure curve. In CR, the exposure parameter linking kilovoltage (kV), current-time product (mA·min), and material thickness is the resulting gray value in the digital image. Based on prior empirical work, an optimal baseline gray value for achieving good contrast and signal-to-noise ratio in our system was determined to be approximately 17,500 (on a 16-bit depth scale). We systematically exposed the multi-step block at various kV and mA·min settings, measuring the average gray value in a region of interest (ROI) on each step. By interpolating the parameters required to achieve the target gray value of 17,500 for each thickness, we constructed the exposure curve for the precision investment casting superalloy. This relationship can be modeled empirically. A common form for such a curve in the kV range used is:
$$ E = k \cdot V^{-n} \cdot e^{\mu T} $$
where \( E \) is the exposure (mA·min), \( V \) is the tube voltage (kV), \( T \) is the total material thickness (mm), \( \mu \) is the effective linear attenuation coefficient (mm-1), and \( k \) and \( n \) are constants dependent on the equipment and detector. By taking logarithms, a more linear relationship for charting can be derived:
$$ \ln(E) = \ln(k) – n \cdot \ln(V) + \mu T $$
For a fixed kV, the relationship between log exposure and thickness is approximately linear. Our experimental data for 170 kV, for instance, fits such a model well. The derived exposure chart is the primary guide for selecting parameters for inspecting unknown components of similar material from precision investment casting.
With the exposure curve established, we proceeded to inspect the TBC-coated plate specimen with the artificial cracks. The total equivalent thickness was approximately 15.4 mm (substrate + coating). Consulting the exposure curve for the target gray level of 17,500 at this thickness indicated an optimal voltage of 170 kV. Adhering to a standard minimum exposure of 15 mA·min for adequate photon statistics, the parameters were set to 170 kV, 15 mA·min, and 700 mm FFD. The resulting CR image exhibited an average gray value of 17,536, confirming the accuracy of the exposure curve. The image quality was evaluated using standard wire IQIs. The visibility of the wires corresponded to a sensitivity level where the single wire designated W12 and the double wire designated D10 were clearly resolved. According to standards such as NB/T 47013.11, this level of sensitivity meets the AB grade (medium sensitivity) requirements for pressure equipment inspection, which is entirely satisfactory for most engineering assessments of precision investment castings.
Most importantly, the artificial cracks were detectable. A clear correlation between crack depth and visibility was observed. Cracks with depths of 0.9 mm and 1.2 mm were very clearly visible. The critical finding was that the crack with a depth of 0.6 mm, a width of 0.2 mm, and a length of 2 mm was also reliably identified in the CR image. This establishes the detection threshold for this specific setup applied to this precision investment casting material. Cracks shallower than 0.6 mm (0.4 mm and 0.2 mm) presented with significantly lower contrast and were not consistently distinguishable from background noise under these standard viewing conditions. This highlights both the capability and the limitation: the CR technique is effective for detecting cracks of a certain severity, but very fine, shallow defects may require enhanced techniques or different modalities.
The digital nature of CR images unlocks powerful post-processing tools that can significantly enhance defect detection and characterization. We applied and evaluated several fundamental image processing techniques to improve the interpretability of the acquired radiographs. The first and most immediate tool is windowing, also known as grayscale mapping. A raw CR image contains pixel values across a wide dynamic range (e.g., 0-65,535 for 16-bit). The display, however, can only show a limited range of grayscales at once (e.g., 256). Windowing selects a specific sub-range, defined by a center value (Window Level, WL) and a width (Window Width, WW), to map onto the full display range. The transformation is linear: pixel values below (WL – WW/2) are displayed as black, values above (WL + WW/2) as white, and values in between are linearly scaled to grays. Mathematically, the displayed intensity \( I_{display}(x,y) \) for pixel value \( P(x,y) \) is:
$$ I_{display}(x,y) = \begin{cases}
0 & \text{if } P(x,y) \leq (WL – \frac{WW}{2}) \\
255 & \text{if } P(x,y) \geq (WL + \frac{WW}{2}) \\
\frac{255}{WW} \cdot \left( P(x,y) – (WL – \frac{WW}{2}) \right) & \text{otherwise}
\end{cases} $$
By narrowing the WW and adjusting the WL to the average gray value of a region containing a crack, the local contrast around the defect can be dramatically increased. This simple adjustment proved highly effective in making the subtle 0.6 mm deep crack much more conspicuous, as it stretched the small gray-value difference between the crack and its surroundings across the entire display range.
Beyond simple contrast adjustment, we employed spatial domain sharpening techniques to enhance edge information. Cracks manifest as local discontinuities in gray value—edges. Sharpening algorithms accentuate these high-frequency components. The first method applied was spatial differentiation using a gradient filter. The gradient magnitude \( G(x,y) \) of an image \( f(x,y) \) approximates the local change and is often computed using the Sobel operators:
$$ G_x = \begin{bmatrix} -1 & 0 & +1 \\ -2 & 0 & +2 \\ -1 & 0 & +1 \end{bmatrix} * f, \quad G_y = \begin{bmatrix} -1 & -2 & -1 \\ 0 & 0 & 0 \\ +1 & +2 & +1 \end{bmatrix} * f $$
$$ G(x,y) = \sqrt{ G_x(x,y)^2 + G_y(x,y)^2 } $$
where \( * \) denotes the convolution operation. Pixels with high gradient magnitude correspond to edges. An output image can be created by adding a scaled version of the gradient magnitude back to the original image, thereby enhancing edges:
$$ f_{sharp}(x,y) = f(x,y) + \lambda \cdot G(x,y) $$
where \( \lambda \) is a scaling factor controlling the degree of sharpening. Applying this to our CR image made the crack boundaries appear sharper and more defined, aiding in length and morphology assessment.
The second sharpening approach operated in the frequency domain, based on the principle that edges and fine details are associated with high spatial frequencies. The 2D Discrete Fourier Transform (DFT) converts the image \( f(x,y) \) into its frequency domain representation \( F(u,v) \). A high-pass filter \( H(u,v) \) attenuates low frequencies (slow varying background regions) while preserving or amplifying high frequencies. A simple ideal high-pass filter is defined as:
$$ H(u,v) = \begin{cases} 0 & \text{if } D(u,v) \leq D_0 \\ 1 & \text{if } D(u,v) > D_0 \end{cases} $$
where \( D(u,v) \) is the distance from the point \( (u,v) \) to the origin of the frequency plane, and \( D_0 \) is the cutoff frequency. The filtered image is obtained by taking the inverse DFT of \( F(u,v) \cdot H(u,v) \). In practice, gentler filters like the Butterworth high-pass filter are used to avoid ringing artifacts. The resulting image emphasizes edges and fine details, effectively sharpening the overall appearance. For our crack inspection, this method also successfully enhanced the visibility of the crack lines, particularly their terminations.
To quantitatively compare the effects of these processing techniques, we can define a simple metric like local contrast-to-noise ratio (CNR) for a specific crack. If \( \mu_{defect} \) and \( \sigma_{defect} \) are the mean and standard deviation of pixel values within a small ROI on the crack, and \( \mu_{background} \) and \( \sigma_{background} \) are the same for an adjacent crack-free region, the CNR is:
$$ CNR = \frac{|\mu_{background} – \mu_{defect}|}{\sqrt{\sigma_{background}^2 + \sigma_{defect}^2}} $$
While a full quantitative analysis is beyond the scope of this summary, qualitative assessment clearly showed that both windowing and sharpening increased the CNR for the shallower cracks, pushing them above the visual detection threshold.
| Processing Technique | Primary Effect | Advantage for Crack Detection | Potential Drawback |
|---|---|---|---|
| Windowing (WW/WL Adjustment) | Linear contrast stretching of a selected gray value range. | Immediate, simple, and highly effective for boosting local contrast. Requires no complex computation. | Optimal settings are subjective and region-dependent. Can lose information outside the window. |
| Spatial Gradient Sharpening | Accentuates local pixel value gradients (edges). | Enhances crack boundary definition, making edges sharper. Works directly on the pixel matrix. | Can also amplify image noise. Requires selection of a suitable kernel and scaling factor (λ). |
| Frequency Domain High-Pass Filtering | Attenuates low-frequency background, retains high-frequency details. | Theoretically clean separation of signal components. Can be tuned via cutoff frequency. | Computationally more intensive. May introduce artifacts if filter is not chosen carefully. |
The successful application of CR to precision investment casting superalloys hinges on understanding the influence of material structure. Equiaxed grain structures, typical in many precision investment castings, present a relatively uniform attenuation path for X-rays compared to directionally solidified or single-crystal structures. However, micro-porosity, grain boundaries, and elemental segregation can introduce a subtle “texture” noise in the radiographic image. This background noise sets the ultimate limit for detecting low-contrast defects like fine cracks or micro-shrinkage. The 0.6 mm deep crack detection limit established here should be considered in this context. For precision investment castings with denser or more coarse-grained structures, or for different alloy compositions, the exposure curves would need to be re-calibrated, as the linear attenuation coefficient \( \mu \) changes.
An important consideration for inspecting coated components, like our TBC-coated sample, is the effective atomic number (Z-effective) of the composite material. The ceramic top coat (ZrO2 with Y2O3) has a higher Z than the metallic substrate, causing increased attenuation. Our exposure curve development using an uncoated step block provided a baseline for the superalloy. The good results on the coated plate indicate that for a given total thickness, the higher average attenuation of the coated section was adequately compensated for by the selected exposure parameters derived from the alloy-only curve. For highly complex coating systems or much thicker coatings, a separate calibration might be beneficial.
The choice of 170 kV as the optimal voltage represents a balance between penetration power and contrast. Lower kV provides higher subject contrast (greater difference in attenuation between defect and material) but may require prohibitively long exposure times for thick sections. Higher kV improves penetration but reduces contrast. The exposure curve effectively maps this trade-off. The general relationship between the required exposure \( E \) to achieve a constant optical density or gray value and the kV can be expressed as:
$$ E \propto V^{-n} $$
where the exponent \( n \) typically ranges from 2 to 3 for the kV ranges used in industrial radiography of metals. Our derived data fits this power-law model well, enabling reliable extrapolation for thicknesses slightly outside the calibrated range.
The integration of CR into the quality control workflow for precision investment casting offers significant logistical advantages. The elimination of chemical processing, the reusability of imaging plates, and the instant availability of digital images for review, archiving, and remote analysis streamline the inspection process. Furthermore, the digital format is perfectly suited for automated defect recognition (ADR) algorithms. While not explored in depth here, the image enhancement techniques described are often preprocessing steps for ADR systems, which use pattern recognition and machine learning to flag potential defects automatically. The consistent image quality afforded by a properly calibrated CR system, as demonstrated here, is a prerequisite for such advanced automation.
Future work in this domain could explore several avenues. First, investigating the detection capability for volumetric defects like porosity and inclusions in precision investment castings using CR, potentially employing different image quality metrics or specific processing algorithms like morphological filtering. Second, extending the study to more complex geometries representative of actual turbine blades or vanes, addressing challenges such as scatter control and thickness compensation—techniques like using masking materials or software-based thickness correction algorithms could be evaluated. Third, a rigorous quantitative analysis linking processing parameters (e.g., WW, WL, filter cutoff frequency) to measured detectability indices like the Signal-to-Noise Ratio (SNR) or the classic Contrast-Detail (CD) diagram would provide a more objective basis for technique optimization. Finally, comparing the performance and economics of CR with other digital radiography methods like Direct Radiography (DR) using flat-panel detectors for high-volume inspection of precision investment castings would be valuable for industry adoption.
In conclusion, this comprehensive investigation demonstrates that Computed Radiography is a highly effective and versatile non-destructive testing technique for the inspection of precision investment casting superalloy sheets. By establishing a material-specific exposure curve, we have provided a reliable method for determining optimal imaging parameters. The experimental results confirm that CR can reliably detect surface-breaking cracks with depths of 0.6 mm and greater in these materials, achieving an image sensitivity equivalent to the AB grade of conventional radiographic standards. The inherent digital nature of CR facilitates powerful post-processing, where simple window-level adjustments and more advanced spatial and frequency-domain sharpening techniques can significantly enhance the visibility and interpretability of critical defects. These image processing steps effectively lower the practical detection threshold by improving local contrast and edge definition. As the demand for high-integrity components manufactured through precision investment casting continues to grow in aerospace, power generation, and other high-tech industries, the adoption of robust digital NDT methods like CR, coupled with intelligent image analysis, will play an increasingly vital role in ensuring product quality, safety, and reliability.