1. Introduction
Wire-arc additive manufacturing (WAAM) is a direct energy deposition process that has been increasing in popularity in the past few years as it can be used to create large metal components for different applications. Moreover, it also has a low equipment cost, a high deposition rate, and low material wastage, which also results in a lower impact on the environment [
1].
There is great importance in understanding residual stress (RS) in WAAM Ti-6Al-4V as RS can play a role in causing and hastening damage within materials. This is caused by fatigue, stress-corrosion cracking, and RS-driven creep cracking, as well as by RS directly contributing to the driving force for fractures. Therefore, it is important to measure, evaluate, and attempt to mitigate RS, especially when considering components that are safety-critical [
2].
The heat input required for arc sources during WAAM leads to a high buildup of RS in the component. RS is also associated with shrinkage during the cooling of the component and is highest along the direction of deposition [
3]. Additive manufacturing, particularly WAAM, relies on the part being constructed layer by layer using melted material. However, as the material cools down, it then increases in temperature again once another layer is deposited on it, therefore leading to the material being heated and cooled several times. The unique nature of the thermodynamics of WAAM means that residual stresses (RSs) can form within the part in question, which can lead to distortions and microcracks, which in turn can lead to a failure of the manufactured part. Therefore, RS must be measured and mitigated as much as possible when manufacturing Ti-6Al-4V parts using WAAM.
For the last two decades, the contour method (CM) has been a well-utilised method for evaluating RS, particularly for material processing, where the evaluation of RS at a macro scale is necessary [
4]. Originally introduced by Prime [
5], CM is based on solid mechanics and can determine RS within an object using an experiment that is carried out by cutting a specimen into two pieces and then measuring the deformation caused by the RS distribution. The displacement data are then measured and used to create a finite element (FE) model of the sample, which computes the RS. This model takes into consideration the geometry and material stiffness of the sample, which provides a result unique to the specimen used. However, the main limitation of this method is that it is destructive and requires cutting the sample, rendering it non-functional.
Many non-destructive methods can be also used to measure strain and then stress, therefore being able to evaluate RS, with ultrasonic (US) testing being an important technique. This method of measuring RS was discussed by Noronha I and Weft [
6]. Conventional US measurements are often carried out using single-element probes. The advantages of this method are its simplicity, affordability, and little required equipment with its ability to evaluate RS at different depths [
7]. However, phased array ultrasonics offers an increased quality of inspection with a reduced inspection time. Also, phased array ultrasonics provides the ability to perform several inspections with one array and can provide instant images of the inspection [
8].
Phased array ultrasonics has previously been underutilised within industries for RS measurements. Phased array ultrasonics testing (PAUT) RS measurement was first discussed by Javadi et al. [
9]. However, there is a sizeable gap in research on ultrasonic phased arrays as a method for RS measurement.
Ultrasonic Phased Arrays for Residual Stress Measurement
In their paper on the development of a PAUT-LCR system for RS measurement, Javadi et al. carried out a feasibility study for measuring residual stress in WAAM samples [
9]. In their method, the Longitudinal Critically Refracted (LCR) ultrasonic technique was used as the goal was to measure stress in the bulk areas of the samples. However, this paper will expand on this feasibility study and use phased array ultrasonics to successfully measure RS in WAAM samples for the first time.
Although PAUT RS measurements were introduced by Javadi et al. [
9], they did not go further than a feasibility study and results were not produced as only time-of-flight (ToF) variations caused by RS were discussed, not final RS measurements. While Javadi et al. [
10] and Mills et al. [
11] used similar principles for the measurement of bolt stress using phased array ultrasonics, they did not use them for RS measurement in WAAM components. The microstructures of the weld and WAAM, as well as the heat-affected zone (HAZ), bring more challenges in RS measurement than in the stress measurement of bolted connections. In this paper, phased array ultrasonics testing will be used for the RS measurement of WAAM components for the first time.
In the feasibility study carried out by Javadi et al. [
9], they anticipated that the increase in the number of acoustic paths generated from their setup would therefore increase the measurement accuracy when compared to a traditional setup that generates two acoustic paths from three single-element transducers. As this was only a feasibility study, a comparison between the two methods was not made, but this estimated increased measurement accuracy was one of the focus areas of this study.
However, one of the main disadvantages of PAUT RS measurement, described by Javadi et al., is the issue of average data measurement. The average of the RS affected by the wave travel path, rather than point-based measurements, is measured. Therefore, when measuring RS at specific depths, it can include both bulk and surface stress data, and due to the rapid change in RS in WAAM components, the overall measurements are affected. Increasing the number of measurement frequencies in the method is one way to mitigate this issue. Another important factor to consider is the size of the probe. Larger arrays, due to the averaging issue, have problems measuring RS due to the sharp gradients within WAAM components. As a result, small-footprint arrays are the best alternative for WAAM RS measurements using PAUT.
In their paper on the development of the PAUT RS measurement system [
9], Javadi et al. describe that another one of the main disadvantages of the ultrasonics method is that the ultrasonic wave is influenced by both the material texture as well as RS, making it difficult to differentiate between them. Also, in comparison to other RS measurement methods, only the average of RS is measurable using the ultrasonics method, limiting the method selectivity.
3. Manufacturing of Sample
In this study, a plasma WAAM-deposited Ti-6Al-4V wall was utilised as the experimental sample, measuring a substrate of 250 mm × 60 mm × 7.3 mm in size. The samples were prepared using Ti-6Al-4V filler wire with a diameter of 1.2 mm. Ten layers were deposited with a wire feed speed of 2.2 m/min and a current of 180 A. During the deposition process, the plasma torch travel speed was adjusted, with the first two layers deposited at speeds of 4 mm/s and 4.5 mm/s, respectively, while a constant speed of 5 mm/s was kept for the subsequent layers. The layer height and width were measured to be 0.98 mm and 9.23 mm on average, respectively.
The Plasma Transferred Arc (PTA) was generated with pure argon serving as both the plasma and shielding gas. The flow rates for the plasma and shielding gas were set to be 0.8 L/min and 8 L/min, respectively. A local shielding device with a gas flow rate of 68 L/min was also integrated into the system.
It is important to supply a local shielding gas during this process to protect the area that is undergoing both solidification and melting to prevent oxidation. This often is in the form of a shroud that trails behind the beam and carries out the shielding. Argon is commonly used for shielding due to its density, which provides an improved efficiency of shielding [
30]. The oxidation of the component can create slag inclusions and the evaporation of nitrogen during solidification can cause pores and nitrides, which can cause brittleness [
31].
The experimental setup, depicted in
Figure 2, positioned the plasma torch at a fixed distance of 8 mm from the substrate during deposition while the wire was inclined at a 25° angle to the travel direction. To maintain stability, the substrate was securely fastened using six clamps. The controlled movement of the PTA during deposition was facilitated by a six-axis KUKA robot operating in an alternating travel direction.
Figure 3 shows a schematic of the bottom of a WAAM Ti-6Al-4V sample and a photo of the physical sample itself and illustrates the different zones that were expected to have different RS measurements. Looking at the bottom of the sample, it can be visually seen that there was a distinct heat-affected zone (HAZ) in the centre of the sample. To help with choosing increments for the US RS measurement, different zones were specified, as shown in
Figure 3. These related to the base metal area, the HAZ (Zone 1 and Zone 2), and the rest of the surface of the WAAM sample bottom. These zones were specified so that larger increments of measurements could be taken in the base metal area whilst smaller increments were measured in Zones 2 and 3. The centre of the HAZ of the WAAM sample was located at 0 mm, so
Zone 1 was measured to be from ~−5 mm to 5 mm, whilst
Zone 2 was measured at ~−10 mm to 10 mm at either side of the centre. Using these zones as a basis allows for a comparison with the different types of RS measurements to understand the accuracy of the RS measurements.
6. Results and Discussion
As described previously, RS measurements were collected using the CM and PAUT methods as seen in the FE model in
Figure 12, which was created using the CM. Scan lines were determined that can be used to measure the RS in specific areas of the sample using the phased array ultrasonics method. As the purpose of this study was to demonstrate the ability for WAAM RS measurement using the phased array setup, the scan area for the comparison of data for these results was focused on the Lc scan line. In the figure showing the FE model, the Lc scan line at the bottom of the sample can be seen to contain a high amount of RS. Therefore, by choosing this area, an easy comparison can be made between the phased array and CM results. The Lc scan line was also chosen as the penetration depth of the 5MHz transducers used was measured to be 1.5 mm, which was equivalent to the Lc line in the CM results, as that line also showed the data at 1.5 mm.
Towards the centre of the bottom of the sample, the highest amount of RS was present where RS remained high but was slightly lower within Zone 2.
As seen in the CM results shown in
Figure 13, the stress gradually increased to a peak of 400 MPa at 30 mm along the substrate. The stress then began a gradual decline to the lowest point of −150 MPa at 55 mm along the substrate and then increased again to 100 MPa between 55 and 60 mm.
There were a few anomalies found in the CM data. As seen in
Figure 13, between 0 mm and 1 mm there was a slight spike in stress from −200 MPa to −150 MPa, which then decreased again to −200 MPa before increasing gradually along the substrate as mentioned before. A similar occurrence happened between 45 and 60 mm, where stress increased and decreased several times towards the end of the substrate.
Another factor that stood out from the CM results was the overall shape of the data. We would expect, from RS measurements carried out on the WAAM sample, that the shape of the graph would be symmetric due to the overall manufactured shape of the sample. There were a few possibilities to explain this unexpected graph shape. One was that the shapes of the WAAM samples that were being tested had a slight bow to them and were not completely flat; a visual representation of this bowed shape can be seen in the cross-sectional geometry in
Figure 14, created using a laser scan where the bottom of the WAAM sample can be seen to not be flat. The slight distortion during the production process was due to the stress on the sample during the manufacturing process, and this could have affected the stress values as the distortion of the sample relieved some stress. It is believed that this distortion did not interfere with the ToF measurement for two reasons. (I) the distortion resembled angular shrinkage rather than buckling, which could have affected surface flatness. This was studied by Satarri-Far and Javadi [
32], who found that distortion on welded pipers did not affect the final RS measurements when comparing the finite element results and the experimental data. It should be noted that surface flatness can influence ToF measurement, especially if the width of the PAUT wedge is too wide to account for surface deformations. (II) The width of the PAUT wedge was 8 mm, which made it significantly narrower than any potential surface deformation on this sample.
Efforts were carried out to prevent the distortion of the sample as much as possible. The substrate was flat before deposition and six clamps were used with each side of the substrate having three clamps placed evenly in the longitudinal direction, creating a symmetrical clamping condition.
Shown in
Table 1 are the results from the Snell’s-law experiment to find the correct angle for the wedge for use with Ti-6Al-4V when collecting phase array ultrasonics measurements. From these results, the average difference in gain from the Snell calculated angle wedge could be compared against the wedge angle for titanium. These wedge angles were tested between 22.2 and 23.2 degrees as these are the approximate angles for the first critical angle level when using LCR waves with titanium alloys as a material. Therefore, by measuring within this area, a specific wedge angle degree could be found that was most suitable for the RS measurements.
From these results, the most important thing to consider is that the lower the average difference in gain is, the more suitable the wedge is for titanium, as less gain is required to reach the required amplitude necessary to collect the ToF measurements, ensuring that they are accurate.
As seen in
Table 1, the lowest values were found at 22.6 and 22.7 degrees, and therefore, both wedge angles were suitable for ultrasonic phased array testing in Ti-6Al-4V samples; ultimately, a 22.7-degree wedge was chosen for this paper.
To measure the acoustoelastic constant, tensile testing was carried out, with stress increments in both loading and unloading modes, and the results are shown in
Figure 15 and
Figure 16, respectively. Since two eight-element arrays were used, the results have been presented per element. For example, Element 4 shows the ToF is measured for the acoustic path between Element 4 of the transmitter array and Element 4 of the receiver array. The vertical axis shows the Exdt/t0, so the slope of each of these graphs represents the acoustoelastic constant, L
11, in Equation (7). However, the final acoustoelastic constant can be calculated by considering all 16 graphs presented in
Figure 15 and
Figure 16 using Equation (8). Since it increases with an increasing stress and decreases with a decreasing stress, the loading and unloading graphs are ascending and descending, respectively. However, several points do not follow this general trend. For instance, in Element 6 in
Figure 15, there is a spike at around 350 MPa, followed by an unexpected decrease at the next point. This issue can be attributed to one of the ultrasonic RS-measurement-system errors discussed by Javadi et al. [
9] such as in the couplant film thickness, a triggering error, or a tensile-testing machine error. However, it is important to note that the acoustoelastic constant would have been measured using all 200 (the number of data points in each graph) × 16 data points collected during the tensile testing process, which would have helped minimise the system error. This is one of the main advantages of a phased array over a single-element ultrasonic system, where only one of these graphs (instead of sixteen) would have been generated. Although a single averaged graph can be generated by using the ultrasonic phased array system, by generating 16 individual graphs, any anomalies/errors can be spotted and considered for the final RS results. However, with a single-element ultrasonic system, the singular graph could potentially contain errors without the ability to compare them to the average trend of the results to spot them.
Using the acoustoelastic constants, the final phased-array WAAM RS measurements could be created. Once the RS measurements were collected for each element of the phased array transducer, these measurements could then be averaged to produce a single set of RS measurements. However, it was important to ensure the accuracy of the results and to see whether there were any specific outliers or incorrect measurements from any elements. To find these outliers, the variance between the RS measurements of each element and the averaged results were calculated and are shown in
Figure 17. Any drastic difference between the RS measurements of each element and the averaged results could be visually seen and removed to improve the accuracy of the improved averaged results. To ensure the consistency of the chosen outliers, any measurements with a variance of 5 or above were identified as outliers as measurements above this variance were found to be commonly outside the normal RS measurements, as can be seen at −8 mm on Element 3. As can be seen in
Figure 17, Element 2 most notably had the most outliers, with five measurements identified as outliers, which shows that this element could have been problematic. A variance of above 40 was also calculated at 8 mm for Element 5, which would have greatly affected the final results if not removed. Element 4 also had numerous outliers, with three identified. These outliers for the individual element results could be attributed to the normal ultrasonic stress measurement errors that were reported by Javadi et al. [
9] when using the conventional single-element technique for LCR RS measurements.
In
Figure 18, the stress distribution measurements per element are shown. The results show eight separate elements that have all been used to measure the stress distribution of the same WAAM sample. These results present the measured residual stress at a depth of 1.5 mm from the bottom of the sample, with a depth of 1.5 mm being the most effective area for the measurement of the RS in the sample.
These results have been presented displaying the RS measurements for each element. However, there are a few outliers in these results. The points displayed in red are the RS measurements that were identified as outliers when calculating the variance and using
Figure 17. When considering the outliers and overall shape of the results of each element, Element 2 is most noteworthy as it displayed a very different RS distribution when compared to the rest of the elements. Element 3 and Element 7 seemed to show the most accurate RS measurements with few outliers and similar RS distributions. In all the measurements, the overall RS distribution was as expected as represented by the zones illustrated in
Figure 3. Outside of the HAZ (−10 to 10 mm), a low RS was measured at ~0 to 20 MPa, and when measured at the start of the HAZ, the RS spiked and was measured at ~200 to 300 MPa in all elements, which occured on both sides of the HAZ. In all elements, peak RS was measured at 0 mm, which had been expected. However, within the HAZ, all the elements displayed a similar drastic drop in the RS at both sides of the HAZ from ~0 to −150 MPa.
These drops in the RS and other outliers could have been due to similar issues during the measurement process mentioned previously for the acoustoelastic results discussed by Javadi et al. [
9] such as the triggering of errors, the couplant film thickness, or small angle variances with the array wedge. These factors are difficult to control during an experimental process, but future work can be controlled to mitigate their impacts. For example, one may conduct a phased array experiment using arrays with higher numbers of elements to have better control and reassurance in identifying the potential outliers.
To create the final averaged results for the RS of the WAAM Ti-6Al-4V sample using the phased array ultrasonics method, ToF measurements were used alongside the calculated acoustoelastic constant, and these results are shown in
Figure 19. These measurements have been shown at a depth of 1.5 mm due to the effective depth of the phased array ultrasonics equipment. These results are averages of the per-element results taken from each of the eight elements in the phased array transducers. The red line represents the actual stress measured, and includes the outliers that were previously identified in the per-element results in
Figure 18, and the black line removes these outliers. As can be seen, the removal of these outliers results in more symmetrical results in comparison and represents an improvement in accuracy.
These results follow a similar trend to the per-element results, as expected, with lower RS measurements outside of both Zone 1 and Zone 2, where less heat is present during the manufacturing of a sample. However, the RS increased from −10 mm until Zone 1, where the sample was most affected by heat, and there was a vast increase in the RS. From −5 mm, RS increased to ~220 MPa and peaked at 0 mm to ~310 MPa, where heat was most prevalent during fabrication. As discussed previously for the per-element results, there was a drop in RS measured on both sides of the HAZ at both between −7 and −4 mm, where RS dropped to ~60 MPa, and between 3 and 5 mm, where RS dropped to -27 MPa. Another significant drop in RS was measured at −2 mm at ~−80 MPa. These drops in RS in the HAZ could sometimes be expected within this area of the sample or could be attributed to the issues during the measurement process discussed for the per-element results. Also, the overall trend of the data was not perfectly symmetric. However, this had been expected, as the shape of the sample was slightly distorted, as shown in
Figure 14.
To create the averaged phased array results from the per-element measurements, it was important to consider the outliers found in the per-element measurements. To average correctly, we needed to divide the sum of the black points (measured RS) by the number of points as shown in Equation (9). However, instead of simply considering the number of elements used, we needed to consider the number of elements minus the red points shown in the per-element results (outliers in HAZ) for the specific point being considered. For example, some of the measurements only had a single outlier/red point, such as Elements (3), (5), and (7), so we summed the black points and divided the resulting value by seven.
Doing this mitigated the outliers affecting the RS measurement for the averaged results and was an advantage of using phased array ultrasonics to average the measurements rather than relying on single-element measurements. As previously mentioned in
Section 1, Javadi et al. [
9] anticipated that the increased acoustic paths from the LCR approach with phased arrays would increase the measurement accuracy when compared to using just two acoustic paths from the three single-element transducers in the traditional method. Although not used in this study, this phased array ultrasonic setup can also be utilised with FMC to create many more data. By using the two eight-element arrays for this paper, a potential 8 × 8 matrix of 64 LCR wave paths could be generated, increasing measurement accuracy. Also, with the ability to produce eight sets of RS measurements, the amount of data was increased, and if arrays with a higher number of elements were used, this same setup could be used to produce even more data. This is especially important if problematic elements, such as Element 2, are identified, which would have affected the accuracy of the averaged results. With a higher-element array, the impact of these problematic elements would be reduced.
7. Comparison of Results
When looking at the CM and the per-element results, the overall shape of the graph was more symmetrical to the RS measurements when considering the CM FE model depicted in
Figure 6. As shown in the FE model, RS was low in areas outside of the HAZ whilst towards the centre of the sample where the HAZ was present, from −5 to 5 mm, the RS was much higher, which matched the RS measurements using ultrasonics testing.
Figure 20 shows a comparison between the CM results and the phased array ultrasonic results. Comparing the averaged phased array ultrasonics results in
Figure 19 with the CM results in
Figure 13, shown in
Figure 19, there was an improvement in the overall measurement of the RS distribution within the sample due to the ability to remove known anomalies, which was not possible with the CM. The CM results showed a measurement of −200 MPa on one end of the sample at 0 mm and 100 MPa on the other end of the sample, showing that there was a large disparity in the measured RS outside of the HAZ. On the other hand, the phased array ultrasonics results showed a more symmetrical distribution of stress with RS measured similarly on both ends of the HAZ at ~20 MPa on each side. Both the averaged phased array and the CM results measured a similar peak RS at 0 mm, with the CM measuring it at ~420 MPa and the phased array measuring it at ~310 MPa.
Within the HAZ, the distributions of stress were similar in both results, with a slightly lower RS on one side of the area. These low RS measurements were flipped for the averaged results; however, this was due to the CM being scanned the opposite way. Although both results displayed a slightly asymmetric trend due to the shape of the WAAM sample, the non-symmetry was much more apparent in the CM results.
Set side by side, the phased array ultrasonics results agree better with the CM results compared with the per-element results. As previously discussed, the CM results represent a clear increase in RS at the HAZ between −10 and 10 mm with the highest RS expected to be between −5 and 5 mm. When comparing the shapes of the RS distributions, the phased array ultrasonics results align well with this increase in RS, which can be presumed to reaffirm the accuracy of the results.
Comparing the per-element results with the averaged phased array ultrasonics results, the latter shows a noticeable improvement in the accuracy of the overall RS measurements. By having the ability to compare the RS measurement of each element with the averaged results, outliers can be identified and removed to produce averaged phased array ultrasonics results with improved accuracy. Problematic elements, such as Element 2 in
Figure 17 and
Figure 18, can also be recognised when one is able to compare the RS measurements of the element with an averaged set of data.
Although not carried out in this study, FMC can also potentially be utilised with this phased array ultrasonics testing method, and with the two eight-element phased arrays used for this study, a possible 64 results can be produced. When capturing the acoustoelastic data, a large number can be taken to generate 200 (data points) × 16 (data points collected during tensile testing to reduce the effect of system error) sets of measurements. Therefore, eight graphs can be averaged, with each one of them being generated through the acoustoelastic constant, which itself has been generated using 200 × 16 sets of data. By averaging the eight individual element results with a much larger dataset available compared to a measurement with a single-element transducer, a single graph can be produced, which can improve accuracy and is a notable advantage of using phased array ultrasonics testing.