In modern aviation engineering, the reliance on precision castings for critical components, especially within jet engines, has grown substantially. These aerospace castings are characterized by intricate geometries, thin-walled sections, internal cavities, and a vast variety of types. The inherent complexity of the manufacturing process often leads to the formation of various internal discontinuities, such as porosity, cracks, inclusions, and compositional segregation. Ensuring the structural integrity of these components is paramount, as any defect could lead to catastrophic failure under extreme operational conditions. Traditional non-destructive testing (NDT) methods, primarily film-based radiography and ultrasonic testing, have been the mainstay. However, ultrasonic inspection faces limitations when applied to aerospace castings due to their complex contours, significant thickness variations, coarse grain structure, and poor surface finish. While conventional film radiography provides a direct, intuitive, and effective record of internal defect morphology, size, and distribution, it suffers from significant drawbacks: long inspection cycles, high and rising costs (particularly for film), difficulties in archival storage and sharing of physical films, and environmental concerns related to chemical processing. This context creates a pressing need for a faster, more economical, environmentally friendly, and digitally manageable radiographic inspection technology. Computed Radiography (CR) has emerged as a highly promising digital alternative to film, utilizing reusable Imaging Plates (IP) to capture radiographic data. This study investigates the application of CR technology for the inspection of aerospace castings, focusing on its capability to detect machined artificial defects, its image quality compared to industrial film, and the application of advanced image processing techniques to enhance defect conspicuity.
The fundamental principle of Computed Radiography revolves around the use of a photostimulable storage phosphor within the Imaging Plate. When exposed to X-rays or gamma rays, the radiation passing through the aerospace casting interacts with the IP. High-energy photons excite electrons within the phosphor crystals, promoting them to higher energy states and trapping them in meta-stable energy levels. This process creates a latent image proportional to the radiation intensity distribution, which corresponds to the internal structure and any flaws within the casting. Following exposure, the IP is placed into a CR scanner. A focused red laser beam systematically scans the plate. The laser light provides the energy to release the trapped electrons, which then return to their ground state, emitting blue light in the process—a phenomenon known as photostimulated luminescence (PSL). This emitted light is collected by a light guide or a parabolic mirror and directed to a photomultiplier tube (PMT), which converts the light signals into electrical signals. These analog signals are subsequently digitized by an analog-to-digverter (ADC), resulting in a high dynamic range digital radiograph. The entire process eliminates the need for a darkroom and chemical processing, offering immediate digital results. The digital image, typically with a grayscale depth of 16 bits (0–65,535 levels), contains a vast amount of information, but its wide dynamic range can make subtle defect signals difficult to perceive directly with the human eye, necessitating post-processing.
The subject of this investigation was a representative aerospace casting fabricated from a high-temperature alloy. To quantitatively evaluate the detection performance of CR for typical flaws, artificial defects were machined into the casting body. The defect types and dimensions were designed based on common flaw morphologies found in aerospace castings, such as pore clusters and linear indications. The specifications for these machined defects are summarized in Table 1. The inspection standard referenced for acceptance criteria was ASTM E192, which defines permissible defect sizes for castings. For this class of casting, the maximum allowable pore diameter was established, and linear defects were considered unacceptable. The machining was performed to precise tolerances as per relevant standards to ensure the artificial flaws accurately simulate real-world defects.
| Defect ID | Type | Nominal Dimensions (Diameter × Depth or Length × Width × Depth) [mm] | Machining Tolerance [mm] | Purpose |
|---|---|---|---|---|
| 1 | Hole | Φ0.5 × 1.0 | Φ±0.03 × ±0.03 | Simulates deep, small-diameter porosity. |
| 2 | Hole | Φ1.0 × 0.5 | Φ±0.03 × ±0.03 | Simulates medium-sized, shallow porosity. |
| 3 | Hole | Φ1.5 × 0.3 | Φ±0.03 × ±0.03 | Simulates large-diameter, very shallow porosity. |
| 4 | Slot (Groove) | 5.0 (L) × 0.11 (W) × 0.5 (D) | ±0.10 (L), ±0.05 (W), ±0.05 (D) | Simulates a fine, linear crack-like indication. |
Radiographic inspection of the aerospace casting was conducted using a CR system comprising a high-frequency X-ray generator (HS320), flexible IP plates (240 mm × 300 mm), and a HD-CR35 scanner. Given the significant thickness variations across the complex geometry of the aerospace casting, the inspection was divided into multiple exposure zones to ensure optimal image quality for all regions, a practice similarly required in film radiography. Two primary inspection areas, designated Zone A and Zone B, covering the locations of the artificial defects, were defined. The exposure parameters for both CR and, for comparative purposes, conventional AGFA C7 film radiography were meticulously determined. AGFA C7 is a fine-grain, high-contrast film commonly used for high-sensitivity inspection of aerospace components. The parameters, including tube voltage (kV), current (mA), exposure time, and focus-to-film distance (FFD), were optimized for each technique and zone based on material thickness and desired sensitivity. A summary of the optimized exposure parameters is provided in Table 2.
| Inspection Zone | Material Thickness [mm] | Technique | Tube Voltage [kV] | Tube Current [mA] | Exposure Time [s] | FFD [mm] |
|---|---|---|---|---|---|---|
| Zone A | 5.5 | CR (IP Plate) | 165 | 10 | 84 | 1800 |
| Film (AGFA C7) | 140 | 10 | 180 | |||
| Zone B | 10.0 | CR (IP Plate) | 200 | 10 | 120 | 1800 |
| Film (AGFA C7) | 160 | 10 | 180 |
The CR inspection process yielded 16-bit digital images. Initial visual assessment confirmed that the wide dynamic range of these images, while capturing all radiographic information, made the subtle contrast variations associated with the shallow or small defects challenging to discern. For instance, in the image of Zone A, the thin slot defect produced a very low-contrast signal against the background. In Zone B, Hole 1 (deep and small) showed the highest contrast, Hole 2 was faint, and Hole 3 (large but very shallow) was nearly imperceptible. This underscores a key characteristic of digital radiography: the separation of data acquisition from display. The full data is preserved in the high-bit-depth image, but specialized processing is often required to optimize the presentation for human interpretation. Simple global contrast enhancement techniques like linear window-level adjustment or histogram equalization are often insufficient, as they apply the same transformation to all pixels and can fail to enhance specific localized detail or can amplify noise.
To address this, a multi-scale contrast enhancement algorithm was employed. This method operates on the principle of extracting detail information at different spatial scales, enhancing these details selectively, and then recombining them into the image. It is particularly effective for enhancing subtle features in images with a wide dynamic range, such as CR images of aerospace castings. One effective implementation uses a non-linear s-shaped mapping function applied to a processed version of the image. The core enhancement transformation can be mathematically described. First, consider an input pixel intensity ‘x’. For an 8-bit image, the enhanced output ‘y’ is given by:
$$y = m \times \left[\frac{x – 128}{m}\right]^p + 128$$
where ‘m’ defines the coordinate range. For an 8-bit image (0–255), m is typically 127. The exponent ‘p’ controls the degree of enhancement. For 16-bit CR images (0–65,535), the formula is adapted by scaling the parameters appropriately. The mapping range ‘m’ for 16-bit becomes 32767, and the midpoint is 32768. The generalized form for any bit-depth can be written as:
$$y = M \times \text{sign}(x – C) \times \left|\frac{x – C}{M}\right|^p + C$$
where:
– $x$ is the original pixel intensity.
– $y$ is the enhanced pixel intensity.
– $C$ is the center intensity value (e.g., 32768 for 16-bit).
– $M$ is the half-range value (e.g., 32767 for 16-bit).
– $p$ is the enhancement exponent, where $0.5 < p < 0.9$.
– The $\text{sign}()$ function and absolute value ensure proper handling of values below and above the center.
The choice of the exponent ‘p’ is critical. As ‘p’ decreases from 1, the function applies a stronger non-linear boost to pixel values near the mid-tone, effectively stretching the local contrast for subtle density variations while compressing the very bright and very dark regions. This behavior is summarized in Table 3, which describes the effect of different ‘p’ values on image appearance.
| Exponent ‘p’ Value | Effect on Mapping Function Shape | Resulting Image Enhancement Characteristics |
|---|---|---|
| p = 1.0 | Linear mapping (no enhancement). | Original contrast is preserved. |
| 0.9 < p < 1.0 | Mildly compressive S-curve. | Slight global contrast increase, minimal detail boost. |
| 0.7 ≤ p ≤ 0.8 | Pronounced S-curve. | Strong local contrast enhancement for mid-tone details; optimal for defect visualization in CR images. |
| 0.5 ≤ p < 0.7 | Very strong S-curve. | Aggressive local contrast boost; may introduce artifacts or excessive noise amplification. |
In this study, the raw 16-bit CR images were first converted to 8-bit format using a non-compressive scaling method to retain proportional grayscale relationships. The multi-scale contrast enhancement algorithm with a p-value of 0.7 was then applied. The results were striking. For Zone A, the previously faint slot defect became clearly visible without the aid of any digital magnifying tool. For Zone B, all three hole defects—including the very shallow Hole 3—were readily discernible to the human eye. The algorithm successfully extracted and amplified the low-contrast signal of the defects while maintaining a natural overall image appearance. This demonstrates the algorithm’s generality and utility for inspecting complex aerospace castings where defect contrast can be inherently low.
A direct comparative analysis between CR and conventional film radiography was conducted to evaluate performance metrics. The evaluation criteria included detection sensitivity (ability to image the artificial defects), inspection cost, operational efficiency, and image quality equivalence. The results are summarized in Table 4.
| Evaluation Parameter | Computed Radiography (CR) | Conventional Film (AGFA C7) | Remarks and Implications |
|---|---|---|---|
| Defect Detectability | All four machined defects (3 holes, 1 slot) were clearly imaged and identifiable, especially after contrast enhancement. | All four machined defects were imaged according to standard film interpretation practice. | CR image quality, in terms of defect detection capability, is equivalent to that of high-quality C7 film for this aerospace casting. |
| Inspection Cost per Image | ~$2.00 per scan (based on IP plate cost amortized over 5000 uses). | ~$5.20 per film (material cost for standard size). | CR offers a direct cost reduction of approximately 60% per inspection shot, not accounting for savings from eliminated chemistry and waste disposal. |
| Process Time Efficiency | IP readout time ~60 seconds. No chemical processing. Exposure times were 53% (Zone A) and 33% (Zone B) shorter than film. | Chemical processing (development, fixing, washing, drying) requires ~10-15 minutes. Longer exposure times needed. | CR significantly reduces the total cycle time from exposure to image availability, boosting inspection throughput. Shorter exposures also reduce generator duty cycle. |
| Dynamic Range & Exposure Latitude | Very high (16-bit). A single exposure can often accommodate a wider thickness range in a part, potentially reducing the number of shots needed for complex parts. | Limited (characteristic H&D curve). Requires precise exposure to achieve correct optical density, often necessitating multiple shots for parts with high thickness variation. | CR provides greater flexibility in exposure parameter selection and can forgive minor exposure errors, which is advantageous for complex aerospace castings. |
| Archiving & Sharing | Fully digital. Images are easily stored, retrieved, transmitted, and integrated into digital quality records and databases. | Physical storage of film archives is space-intensive. Sharing requires physical duplication or scanning. | CR facilitates paperless (film-less) operations, remote expert review, and long-term data mining for process improvement. |
| Environmental Impact | No wet chemicals, silver, or related hazardous waste. Lower energy consumption for processing. | Generates chemical waste (developer, fixer) and requires silver recovery systems. | CR is a “greener” technology, aligning with modern environmental regulations and sustainability goals. |
The mathematical foundation of image quality in radiography can be related to basic principles. The contrast of a defect, which is crucial for its detection, is governed by factors like the thickness difference it presents ($\Delta t$), the material’s linear attenuation coefficient ($\mu$), and the radiation energy spectrum. For a simple case under monoenergetic radiation, the contrast C can be expressed as:
$$C \propto \mu \cdot \Delta t \cdot G$$
where $G$ represents the system gradient or film contrast. In digital systems like CR, the gradient is related to the signal-to-noise ratio (SNR) and the display processing. The multi-scale enhancement effectively increases the perceived local gradient for subtle $\Delta t$ variations. The detective quantum efficiency (DQE) is a key metric for comparing imaging systems. While film has a high DQE at optimal exposure, CR systems maintain good DQE over a much wider exposure range. The performance of a CR system in detecting small defects in an aerospace casting can be modeled by considering the modulation transfer function (MTF) and the noise power spectrum (NPS). The limiting spatial resolution for defect detection is often not the limiting factor for volumetric flaws like pores; instead, contrast sensitivity is paramount. The application of the multi-scale enhancement algorithm can be thought of as applying a local, non-linear filter $F(p)$ to the image $I(x,y)$ to produce an enhanced image $I_{enh}(x,y)$:
$$I_{enh}(x,y) = I(x,y) + \lambda \cdot F(p)[D(I(x,y))]$$
where $D(I)$ represents a detail extraction operation (often involving multi-scale decomposition like Laplacian pyramids or wavelet transforms), $F(p)$ is the non-linear enhancement function parameterized by ‘p’, and $\lambda$ is a blending coefficient. This processing selectively amplifies the detail components where defect signals reside.
In conclusion, this comprehensive investigation demonstrates that Computed Radiography is a mature and highly capable technology for the non-destructive inspection of aerospace castings. The study confirms that the intrinsic image quality of CR, in terms of defect detection sensitivity for both volumetric and planar flaws, meets and in some operational aspects surpasses that of conventional high-quality industrial film like AGFA C7. The significant advantages in cost per inspection, process efficiency, exposure flexibility, digital workflow integration, and environmental footprint make CR a compelling replacement for film-based methods in aerospace foundries and maintenance facilities. The wide dynamic range of CR images, while initially presenting a challenge for direct visual analysis, is actually a strength—it preserves all captured information. Through the application of advanced image processing algorithms, such as the multi-scale contrast enhancement technique detailed here, the subtle details critical for flaw detection in complex aerospace castings can be effectively extracted and presented for reliable human or automated analysis. The transition from analog film to digital CR represents a significant step forward in ensuring the safety, reliability, and economical production of vital aerospace components. Future work could focus on the integration of CR with automated defect recognition (ADR) systems and 3D computed tomography for even more comprehensive inspection of high-integrity aerospace castings.