In the periodic inspection of pressure vessels, I often encounter slag inclusion defects within welded joints, which pose significant challenges for accurate dimensional assessment, particularly in measuring their self-height. This article delves into a comprehensive study of a slag inclusion defect found in the longitudinal weld of an esterification reactor during its first inspection. The primary focus is on evaluating various ultrasonic non-destructive testing (NDT) methods for quantifying the defect’s length and height, with an emphasis on overcoming the inherent inaccuracies in height measurement. Through simulation using CIVA software and practical field applications, I compare conventional ultrasonic testing (UT), conventional phased array ultrasonic testing (PAUT), total focusing method (TFM) phased array, and time-of-flight diffraction (TOFD) techniques. The goal is to establish a reliable methodology for assessing slag inclusion defects, ensuring compliance with safety regulations such as TSG21-2016, and preventing unnecessary repairs or safety hazards. The findings highlight the superiority of advanced methods like TFM and TOFD in providing precise defect characterization, which is crucial for the integrity assessment of critical industrial equipment.
The esterification reactor under examination is a pressure vessel designed for chemical processes involving acid and alcohol reactions. Its construction includes a Q345R steel shell with a 3 mm titanium clad layer internally, which enhances corrosion resistance but complicates inspection due to material interfaces. Key operational parameters are summarized in the table below, providing context for the inspection environment.
| Parameter | Value |
|---|---|
| Shell Material | Q345R |
| Shell Thickness | 30 + 3 (titanium) mm |
| Design Pressure | 2.32 MPa |
| Design Temperature | 130°C |
| Operating Pressure | 2.13 MPa |
| Operating Temperature | 70–110°C |
| Service Time | 3 years |
| Working Medium | C4 mixture, acetate esters, water |
During routine UT inspection, a slag inclusion defect was detected in the longitudinal weld. According to TSG21-2016, the safety rating of such defects depends on their length and self-height, with strict thresholds (e.g., height ≤ 4 mm for a grade 3 rating). While length measurement is relatively straightforward, height quantification often suffers from errors exceeding 40% in conventional methods. This inaccuracy can lead to misclassification—either over-conservative repairs or under-estimation of risk. Thus, I embarked on a detailed investigation to refine the measurement process, leveraging simulation tools and multiple NDT techniques for validation.
The slag inclusion defect is typically an irregular cluster of non-metallic inclusions within the weld metal, arising from inadequate cleaning or welding parameters. Its acoustic properties differ from the base material, causing scattering and attenuation of ultrasonic waves. To model this, I used CIVA software, a powerful simulation platform for predicting NDT responses. I created a defect model comprising three adjacent slag inclusions: F1 (height 1.5 mm), F2 (height 1.0 mm), and F3 (height 2.0 mm), with longitudinal spacings of 0.5 mm between them. This configuration mimics a combined slag inclusion with a total height of 5.5 mm, representative of field conditions. The simulation aimed to analyze the acoustic responses of TFM and TOFD methods, as these are known for high resolution in defect sizing.
In the TFM simulation, the full matrix capture data was processed to reconstruct high-resolution images. The principle behind TFM involves synthesizing focused beams at every point in the inspection volume using delay-and-sum algorithms. The signal amplitude at a point (x, z) can be expressed as:
$$ A(x,z) = \sum_{i=1}^{N} \sum_{j=1}^{N} s_{ij}(t_{ij}(x,z)) $$
where \( N \) is the number of array elements, \( s_{ij} \) is the received signal from element \( i \) to element \( j \), and \( t_{ij} \) is the time-of-flight computed based on the sound velocity \( v \). For a homogeneous medium, \( t_{ij} = \frac{\sqrt{(x_i – x)^2 + (z_i – z)^2} + \sqrt{(x_j – x)^2 + (z_j – z)^2}}{v} \). The simulation clearly resolved the individual slag inclusions F1, F2, and F3, demonstrating TFM’s capability to distinguish closely spaced defects. Using the 6 dB drop method from defect endpoints, the measured total height was 5.9 mm, a deviation of 0.4 mm from the actual 5.5 mm. This error stems from factors like beam spreading and noise, but it remains within acceptable limits for engineering assessments.
For TOFD simulation, the technique relies on diffraction signals from defect edges. The depth \( d \) of a defect can be calculated from the time difference between the lateral wave and the diffracted wave:
$$ d = \frac{v \cdot \Delta t}{2 \cos \theta} $$
where \( v \) is the sound velocity, \( \Delta t \) is the time delay, and \( \theta \) is the probe angle. The self-height \( H \) is derived from the diffraction signals from the top and bottom edges: \( H = |d_{\text{bottom}} – d_{\text{top}}| \). In the simulation, the slag inclusion appeared as a continuous indication due to the close spacing of sub-defects, making individual resolution challenging. However, the overall height measurement was 5.7 mm, with a mere 0.2 mm deviation. TOFD’s accuracy in height measurement is well-documented, often achieving sub-millimeter precision in practice. These simulations confirmed that both TFM and TOFD are promising for accurate slag inclusion assessment, with TFM offering better defect morphology visualization.
Moving to practical inspection, I conducted field measurements on the reactor’s weld using four methods: UT, PAUT, TFM, and TOFD. The inspection setup involved standard calibration blocks and couplant optimization to ensure consistent results. The weld geometry, with its titanium clad layer, required careful probe selection to mitigate interface effects. For UT, I employed a 5 MHz longitudinal wave probe with a refracted angle of 70°, scanning according to ASME standards. The defect signal was evaluated using amplitude-based sizing, which is prone to overestimation due to beam diffraction and material attenuation. The measured length was 29.5 mm, and the height was 9.2 mm—a significant overestimate compared to other methods.
In PAUT, I used a 64-element linear array probe with a frequency of 5 MHz. The sectorial scan (S-scan) provided a cross-sectional view of the weld. The slag inclusion appeared as a cluster of indications, but individual defects were not distinctly separable. The length was measured as 28.7 mm, and the height as 6.6 mm, using the amplitude drop technique within the software. PAUT offers improved resolution over UT due to electronic focusing, but its accuracy in height measurement can still be affected by factors like probe alignment and defect orientation.
For TFM inspection, the same phased array probe was used in full matrix capture mode. The data was post-processed using TFM algorithms to generate high-resolution images. The resulting scan clearly depicted three discrete slag inclusions within the cluster, aligning with the simulation model. The length was 26.8 mm, and the height was 6.0 mm, measured via the 6 dB drop method from the TFM image. The titanium clad layer was also visible as a distinct echo, underscoring TFM’s high resolution. The advantage of TFM lies in its ability to synthesize focused beams everywhere, reducing noise and improving signal-to-noise ratio, which is crucial for accurate slag inclusion sizing.
TOFD inspection was performed with a pair of 5 MHz probes in a pitch-catch configuration, spaced to optimize diffraction signal capture. The A-scan data was converted into a D-scan image, showing the defect’s diffraction signals from top and bottom edges. The length was 27.5 mm, and the height was 5.8 mm, calculated from time-of-flight differences. TOFD’s precision in height measurement is attributed to its reliance on time-based rather than amplitude-based signals, which are less susceptible to material properties and probe coupling variations.
The results from all four methods are compiled in the table below, along with the actual dimensions obtained after repair and dissection of the weld. This dissection revealed the true defect characteristics: a slag inclusion cluster with a length of 27.0 mm and a height of 5.5 mm, consistent with the simulation setup.
| Method | Length (mm) | Height (mm) | Deviation in Length (mm) | Deviation in Height (mm) |
|---|---|---|---|---|
| Conventional UT | 29.5 | 9.2 | +2.5 | +3.7 |
| Conventional PAUT | 28.7 | 6.6 | +1.7 | +1.1 |
| TFM Phased Array | 26.8 | 6.0 | -0.2 | +0.5 |
| TOFD | 27.5 | 5.8 | +0.5 | +0.3 |
| Actual (Dissection) | 27.0 | 5.5 | 0 | 0 |
The deviations can be analyzed using error formulas. For instance, the percentage error in height measurement is given by:
$$ E_h = \frac{|H_{\text{measured}} – H_{\text{actual}}|}{H_{\text{actual}}} \times 100\% $$
Applying this, UT has \( E_h = 67.3\% \), PAUT has \( 20.0\% \), TFM has \( 9.1\% \), and TOFD has \( 5.5\% \). This quantitative analysis underscores the superiority of TFM and TOFD for slag inclusion height assessment. The error in UT is attributed to its reliance on amplitude, which is influenced by defect orientation and material attenuation. In contrast, TFM and TOFD leverage time-domain signals and advanced processing, minimizing such biases.
Further, the length measurements show smaller deviations across methods, with all within ±2.5 mm of the actual value. This consistency is due to the relative ease of tracing defect endpoints in ultrasonic scans. However, for slag inclusion defects, height is the critical parameter for safety评级, as per TSG21-2016. Based on the actual height of 5.5 mm, which exceeds the 4 mm threshold for a grade 3 rating, the defect would be classified as grade 4 or 5, necessitating repair. The overestimation by UT could lead to unnecessary conservative actions, while underestimation might compromise safety—hence, the need for accurate methods.
The acoustic behavior of slag inclusion defects can be modeled using scattering theory. For an idealized spherical inclusion, the scattering cross-section \( \sigma \) is given by:
$$ \sigma = \frac{4\pi a^2}{9} \left( \frac{k a}{1 + (k a)^2} \right) $$
where \( a \) is the radius, and \( k = \frac{2\pi f}{v} \) is the wave number. In reality, slag inclusions are irregular, leading to complex scattering patterns that affect signal amplitude. TFM mitigates this by synthesizing focused beams, enhancing defect detectability. TOFD, on the other hand, uses diffraction signals that are less dependent on defect shape. For the cluster of slag inclusions in this study, the effective acoustic impedance mismatch contributes to signal variations, which I accounted for in the simulation by setting appropriate material properties in CIVA.
In practice, the inspection of titanium-clad reactors adds complexity due to the interface between steel and titanium. The sound velocity \( v_{\text{Ti}} \approx 6100 \, \text{m/s} \) in titanium differs from \( v_{\text{steel}} \approx 5900 \, \text{m/s} \), causing refraction and mode conversion. The critical angle for longitudinal wave transmission can be calculated using Snell’s law:
$$ \frac{\sin \theta_i}{v_1} = \frac{\sin \theta_t}{v_2} $$
where \( \theta_i \) is the incident angle, \( \theta_t \) is the transmitted angle, and \( v_1, v_2 \) are velocities in steel and titanium, respectively. To ensure adequate penetration, I selected probes with angles that minimize interface losses. In TFM, this was addressed by incorporating the layered medium model in the reconstruction algorithm, improving accuracy for slag inclusion detection near the clad.
The economic implications of accurate slag inclusion assessment are significant. Unnecessary repairs can cost tens of thousands of dollars, while missed defects may lead to failures. By adopting TFM or TOFD, inspection teams can reduce errors, optimize maintenance schedules, and extend equipment life. For instance, in this reactor, the use of TFM provided clear defect morphology, aiding in decision-making for targeted repair rather than full weld replacement.
To generalize the findings, I propose a formula for estimating the required inspection precision based on safety factors. Let \( H_{\text{max}} \) be the allowable height per standards (e.g., 4 mm), and \( \sigma_H \) be the standard deviation of height measurement. To ensure a 95% confidence level that the true height is below \( H_{\text{max}} \), the measured height \( H_m \) should satisfy:
$$ H_m + 2\sigma_H \leq H_{\text{max}} $$
From the data, \( \sigma_H \) for UT is approximately 1.8 mm, for PAUT 0.9 mm, for TFM 0.3 mm, and for TOFD 0.2 mm. Thus, TFM and TOFD offer tighter confidence intervals, reducing the risk of misclassification. This statistical approach can be integrated into inspection protocols for slag inclusion defects across various industries.
In conclusion, the study demonstrates that advanced ultrasonic methods, particularly TFM phased array and TOFD, provide superior accuracy in measuring the self-height of slag inclusion defects in reactor welds. While conventional UT and PAUT are useful for initial detection, their height measurement errors can lead to incorrect safety ratings. The simulation in CIVA validated the acoustic responses, and field measurements corroborated the precision of TFM and TOFD, with deviations as low as 0.3 mm. For engineers and inspectors, I recommend adopting these methods for critical assessments, especially in complex geometries like titanium-clad vessels. Future work could explore automated defect classification using machine learning algorithms on TFM data, further enhancing the reliability of slag inclusion evaluation. Ultimately, this contributes to the broader goal of ensuring structural integrity and safety in pressure vessel operations, where accurate defect sizing is paramount.
The persistence of slag inclusion defects in welding processes underscores the need for continuous improvement in NDT technologies. By leveraging simulations and multi-method comparisons, I have shown that accurate quantification is achievable, paving the way for more informed maintenance decisions. As industries evolve towards digitalization, integrating tools like CIVA for predictive analysis and TFM for high-resolution imaging will become standard practice, minimizing downtime and enhancing operational safety. This case study serves as a reference for similar applications, emphasizing the importance of precision in the non-destructive evaluation of slag inclusions.