Mechanical and microstructural testing of wire and arc additively manufactured sheet material

• Results of tensile tests on WAAM stain- less steel coupons are presented. • Degree of anisotropy and in ﬂ uence of geometric variability are examined. • Non-contact measurement methods are usedtodeterminethegeometryandde-formations. • Microstructural analysis of the samples reveals a strong crystallographic tex- ture. • Effective mechanical properties de ﬁ ned for as-built material based on simple geometrical measures. There are, uncertainties relating to the structural performance of material, including the basic mechanical properties, the of anisotropy, the in uence of the as-built geometry and the Towards a comprehensive seriesoftensiletests onWAAM stainless conducted; the results are presentedherein. As-builtand ma- chined couponswere tested to investigate the in ﬂ uence of the geometrical irregularity on the stress-strain characteristics, while material anisotropy was explored by testing coupons produced at different angles to the printing orientation. Non-contact measurement techniques were employed to determine the geometric proper-tiesanddeformation ﬁ eldsofthespecimens,whilesophisticatedanalysismethodswereusedforpostprocessing the test data. The material response revealed a signi ﬁ cant degree of anisotropy, explained by the existence of a strongcrystallographictexture, uncovered bymeans ofelectronbackscatterdiffraction. Finally,the effectiveme-chanical properties of the as-built material were shown to be strongly dependent on the geometric variability; simplegeometricmeasureswere thereforedevelopedto characterise the key aspects ofthe observed behaviour.


Introduction
Additive manufacturing (AM), commonly known as 3D printing, is a manufacturing process whereby a component is built-up layer by layer, as defined by a 3D digital model [1]. The potential for a reduction in material consumption and wastage, geometric freedom and enhanced customisation when compared to conventional fabrication methods [2], in conjunction with the feasibility of its application across a wide range of materials including metals, polymers, ceramics and composites [3], has led to rapid growth in the use of AM across multiple industries [4]. The worldwide AM market was estimated to be worth over $7 billion in 2018, with market trends indicating further upcoming growth [5].
According to ISO/ASTM 52900 [6], the principal types of metal AM are sheet lamination, powder bed fusion (PBF) and directed energy deposition (DED), requiring little to no post-processing, with the two latter techniques being deemed the most appropriate for the construction industry [2,3]. Wire and arc additive manufacturing (WAAM) is a method of DED that uses conventional welding technology [7,8], allowing the production of large-scale metallic components [9,10] in a timely and cost-effective manner [11]. However, there are a number of prominent challenges, many of which relate to uncertainties and variability in the basic mechanical properties [12].
Studies into the microstructure and mechanical properties of WAAM stainless steel, although limited in number, have indicated the potential correlation between the employed printing strategy and the exhibited material properties [31][32][33]. Eriksson et al. [34] showed that increasing the heat input during printing results in increased ductility and reduced yield and tensile strengths, while Ji et al. [35] found that specimens loaded parallel to their printing orientation exhibited higher ultimate strength and less ductility than specimens with the load acting perpendicular to their printing orientation. However, it should be noted that other researchers did not observe such anisotropic behaviour [36][37][38].
The aims of the present paper are to broaden the pool of experimental data on WAAM stainless steel and to assess the dependency of the material properties on the employed printing strategy. The undulating as-built geometry, which is inherent to the WAAM process, can be machined smooth, but this additional operation will add time and cost and may not be feasible with certain printed geometries. If the material is left in the as-built state, an important question is: what is the influence of the inherent geometric undulation on the effective mechanical properties? This issue is also addressed herein and simple design rules for predicting the key mechanical properties are established.

Manufacturing of test specimens
Flat plates of 3.5 mm and 8.0 mm nominal thickness were cut from oval tubes (with flat sides), printed using Grade 308LSi austenitic stainless steel wire of 1 mm thickness. The chemical composition of the employed parent stainless steel wire, as well as its mechanical properties, as provided in the manufacturer's data sheet, are presented in Tables 1 and 2, respectively. The specimens were printed by the Dutch start-up company MX3D, using their proprietary multi-axis robotic WAAM technology [39]. Following fabrication, the specimens were sandblasted with glass beads, the size of which was sufficiently coarse to clean the surface, but sufficiently fine not to affect their geometry.
The as-built surface of WAAM elements is inherently undulating, resulting in geometrical variations across their profile and hence non-uniform deformations under load. In order to assess the influence of the as-built geometry on the mechanical response, tensile coupon tests were performed on both the as-built and machined material; in the case of the latter, the surface undulations of the as-built material of 8.0 mm nominal thickness were removed by an end mill, as shown in Fig. 1(a). The resulting machined coupons were prismatic and of uniform thickness; a comparison of the as-built and machined surface of a typical plate is shown in Fig. 1(b). Coupons were extracted from both the as-built and machined material at 0°, 45°and 90°to the print layer orientation, as defined in Fig. 2, in order to investigate any material anisotropy. Typical as-built and machined coupons are illustrated in Fig. 3.
Prior to testing, the machined coupons were X-ray scanned to determine the location and extent of any internal voids; a typical scan is shown in Fig. 4. Minor internal defects were identified, with void sizes of up to 1 mm (in the plane of the X-ray image), though there was no obvious correlation between the void locations and the position of fracture or noticeable influence on the measured strain field; the presence of the minor internal defects was therefore deemed not to have any significant influence on the monotonic deformation response of the coupons.
The adopted specimen labelling system begins with the letter M for the machined coupons and AB for the as-built coupons; this is followed by the nominal thickness of the plate in mm, the angle to the print layer orientation in degrees as defined in Fig. 2, the plate from which the coupon was cut and finally the number of the specimen originating from the plate. For instance, coupon AB-3.5-0°-C4 is an as-built coupon with a nominal thickness of 3.5 mm, aligned at a 0°angle to the print layer orientation and was the fourth coupon from the plate labelled "C". In total, 51 coupons of different thicknesses, orientations and finishes were tested.

Geometric measurements
The geometric measurements of the machined tensile coupons were determined prior to testing using Vernier callipers. However, the undulating surface of the as-built WAAM coupons rendered the use of conventional measuring techniques impractical. Laser scanning was therefore employed to record the full geometry of the printed specimens digitally and hence obtain detailed and accurate geometric measurements, as described in this section.

Methods and techniques
A FARO ScanARM, capable of measuring up to 600,000 points per second to an accuracy of 0.1 mm, was used to scan each as-built coupon. Laser scans were taken at several different angles of incidence to the coupon surface in order to capture all of the geometric undulations. The scans were subsequently merged using the software Geomagic  [41] as STL files, and accurately aligned with the global coordinate system using principal component analysis [42,43], with the overall centroid of the coupon coinciding with the origin of the global coordinate system and with the longitudinal axis of the coupon being parallel to the global axis x -see Fig. 6. Following alignment, contouring of each specimen at regular intervals dx along the longitudinal axis was conducted to determine the geometric properties at each crosssectional cut. The cross-sectional area A i and the eccentricities e y,i and e z,i of the cross-sectional centroid P i , along the global y and z axes respectively, relative to the overall centroid of the coupon (coinciding with the origin of the global coordinate system), were calculated at each cut. The average cross-sectional thickness t i was calculated by contouring the coupon in the transverse direction (along the global y axis) with the same spacing used for the longitudinal contouring. A schematic illustration of a typical coupon processed in Rhino 3D, as well as the resulting cross-sectional area A i , thickness t i and eccentricities e y,i and e z,i at a typical cross-section, are illustrated in Fig. 6.
A sensitivity study was undertaken to determine the most suitable contour spacing that would allow sufficient measurements to be taken to ensure an accurate replication of the geometrical variation along the length of the specimen, while achieving computational efficiency. Twelve as-built specimens were chosen (two for each build direction and thickness), the geometric measurements of which were obtained at contour spacings of 0.05 mm, 0.10 mm, 0.50 mm, 1.00 mm and 2.00 mm. Note that a square mesh was employed for all different spacings with dx = dy. It should be mentioned that the considered contour spacings were all below the typical WAAM deposition width w, shown in Fig. 7, which was found to vary between about 3 mm and 5 mm for the studied samples; a similar value of 4 mm was reported by Ding et al. [44].
A summary of the results obtained from the conducted sensitivity study is presented in Fig. 8, where the mean and minimum measurements of area (A and A min respectively) and thickness (t and t min respectively) and the mean and maximum centroid eccentricity (e z and e z,max respectively) for each contour spacing are normalised against the equivalent values corresponding to dx = 0.05 mm. As expected, it can be observed that the extreme values of all measurements (namely A min , t min and e z,max ) are more sensitive to the contour spacing compared to their equivalent mean values (namely A, t and e z respectively). Overall, the measurements obtained using a spacing of 0.10 mm were almost identical to these obtained using a spacing of 0.05 mm; hence a contour spacing of 0.10 mm was adopted.

Geometric properties
The average geometric properties of the as-built coupons are presented in Table 3, where t nom is the nominal thickness of each specimen, θ is the orientation of the coupon relative to the print layer orientation, as defined in Fig. 2, A is the mean cross-sectional area, t, t min and t sd are the mean, minimum and standard deviation values of the thickness respectively, e y , e y,max and e y,sd are the mean, maximum and standard deviation values of the eccentricity along the y axis respectively, e z , e z,max and e z,sd are the mean, maximum and standard deviation values of the eccentricity along the z axis respectively and Δ is the difference between the actual length of the as-built coupon and its corresponding developed length (i.e. the length of a line passing through each crosssectional centroid, within the parallel length).
The θ = 0°coupons exhibited the least variation in thickness along their length, as indicated by the t sd values and the values of the t min /t ratio being close to unity; this can be explained with reference to the direction of the contours relative to the print layer orientation. For the θ = 0°specimens, as illustrated in Fig. 9(a), the contour planes are perpendicular to the direction of the deposition paths, and all crosssections are similar; hence, the variation in the values of t i between different cross-sections is expected to be minor, leading to low values of standard deviation t sd . In contrast, since for the θ = 90°coupons the contour planes are parallel to the direction of the deposition paths (see Fig. 9(b)), some cross-section cuts are expected to contain only trough regions (leading to values of t min being substantially lower   than t) while others only crest regions, leading to a much greater spread in t i , and, thus, higher values of t sd .
A histogram of the measured thicknesses, grouped by coupon orientation relative to the print layer direction, is presented in Fig. 10, with each cross-sectional thickness measurement t i normalised by the corresponding average coupon thickness t. For the 0°specimens, the distribution is approximately symmetric, but, for the other two types of specimen, the distributions exhibit an increasingly positive skew (i.e. the mass of the distribution is concentrated on the left of the histogram) as θ increases, particularly for the 90°specimens.
Finally, according to the values reported in Table 3, the maximum centroid eccentricities, as well as their standard deviations, increase with increasing values of θ. This can be also observed in Fig. 11, where the normalised eccentricities along the y and z axes (e y /t and e z /t respectively) are plotted against the position along the coupon length for three typical specimens with θ = 0°, 45°and 90°. As expected, the eccentricities measured in the y-direction e y were generally significantly lower than those in the z-direction e z .

Test set-up
In order to obtain the monotonic stress-strain response of the WAAM material, both machined and as-built coupons were subjected to tensile testing at room temperature. All tests were conducted in compliance with EN ISO 6892-1 [45] in the Structures Laboratory of the Department of Civil and Environmental Engineering at Imperial College London. The cross-sectional area of the machined coupons A was measured using a mechanical micrometer of 0.001 mm accuracy, and standard gauge lengths of 5.65 ffiffiffi A p were marked onto their surfaces for the calculation of fracture strains ε f . The geometric properties of the asbuilt coupons were determined by laser scanning, as described in the previous section. Strains at fracture were again determined over the standard gauge length, calculated based on the mean cross-sectional area.
For the machined specimens, two electrical resistance strain gauges were attached at the mid-height of each coupon, one on either side, to record longitudinal strains in the early stages of testing (up to the 0.2% proof stress σ 0.2 ) while, for both the machined and as-built specimens, a four-camera LaVision digital image correlation (DIC) system was employed to provide highly accurate measurements of the surface strain field on both sides of the coupon.
Prior to testing, the parallel length of all coupons was painted white and then spray-painted with a random black speckle pattern, in order for the strains to be calculated over the full area of the parallel length. The DIC cameras were positioned around the specimen, with each camera pair monitoring opposite faces, while the angle between the pairs of cameras was kept at approximately 60°as a compromise between a larger angle to increase the stereo effect and hence the accuracy of the computed out-of-plane deformations and a smaller angle to ensure that the area-of-interest remained in sharp focus. The DIC setup is shown in Fig. 12. The use of DIC is particularly important for WAAM material since the strain field under macroscopic uniaxial loading is not as uniform as for conventionally produced material. The use of single point strain gauge measurements, which can only reflect the localised response of the sample, can lead to inaccurate results.
An overview of the tensile test setup is presented in Fig. 13. A 250 kN Instron 8802 testing machine was used for the application of the tensile load, which was measured by a load cell within the actuator. Both axial load and strain gauge measurements were recorded at a frequency of 2 Hz using an in-house developed data logger, while the DIC system recorded the tensile force through an analogue to digital converter and acquired images at a frequency of 1 Hz. The acquired images were processed in the software DaVis [46]. A longitudinal strain box was drawn over the full parallel length of both sides of each coupon, within which the relative displacements of the surface patterns were analysed. An average stress-strain curve was then produced.
Load was applied using strain control at a strain rate of 0.00007 s −1 prior to the 0.2% proof stress σ 0.2 , while a strain rate of 0.00025 s -1 was employed beyond σ 0.2 and until failure, in accordance with EN ISO 6892-1 [45]. A gradual transition from the initial to the final strain rate was achieved by using three intermediate strain rates.

General
Stainless steel exhibits a rounded stress-strain response with no sharply defined yield stress and significant strain hardening. Several material models have been devised over the years to describe the  response of nonlinear metallic materials, the vast majority of which are based on the Ramberg-Osgood expression [47,48]. Nowadays, the most frequently employed material model is a two-stage version of the original expression, as given by Eqs. (1) and (2) [49][50][51][52]: where σ and ε are the engineering stress and strain respectively, E is the Young's modulus, σ u and σ 0.2 are the ultimate and 0.2% proof stresses respectively, E 0.2 is the tangent modulus of the stress-strain curve at σ 0.2 , as given in Eq. (3), ε u and ε 0.2 are the total strains corresponding to the ultimate and 0.2% proof stresses respectively while n and m u are strain hardening exponents determining the degree of roundedness of the stress-strain curve. An alternative two-stage Ramberg-Osgood model, in which the 1% proof stress σ 1.0 features in the second stage in place of the ultimate tensile stress σ u , was proposed by Gardner and Ashraf [53]. In this model, the strain hardening exponent for the second stage is denoted m 1.0 .
The two-stage Ramberg-Osgood model given by Eqs. (1) and (2) has been employed herein to describe the measured stress-strain curves of the WAAM material. The process by which the key parameters were fitted to the measured data is described below.
Particular attention was given to the determination of the Young's modulus E for each of the examined coupons; a typical example of the adopted automated process is presented in Fig. 14. First, the numerical derivative of the stress-strain curve was computed. Since the use of a finite difference approach was deemed to be inappropriate as the stressstrain curve is expected to contain experimental noise, a moving regression filter was passed over the curve and, for each point, ordinary least squares regression (OLSR) analysis was employed to calculate E. Then, as shown in Fig. 14(b), the calculated values of E were plotted against the corresponding values of ε (calculated as the mean strains across the regression window). In Fig. 14(b), five regions can be identified: (i) noise, though barely visible, prior to the start of the test, (ii) a ramp up in E as the test begins, (iii) a plateau of constant E, corresponding to the elastic region of the material response, (iv) a ramp down in E due to the gradual introduction of plasticity and (v) a final plateau corresponding to the plastic region of the material response. Only region (iii) is required for calculating the value of the Young's modulus E; however, its identification and isolation in a consistent and automated manner is difficult. To overcome this difficulty, the Haar wavelet transform [54] was used to detect the position of the ramp up and ramp down events in regions (ii) and (iv) respectively, from which a window containing region (iii) (defined in between bands A and B in Fig. 14(b)) can be inferred. An S-shaped logistic curve, similar to Richards curve [55], was then fitted to the data between bands A and B using the  Levenberg-Marquardt nonlinear least squares regression algorithm [56,57]. The Young's modulus E was then calculated as the value of the fitted logistic curve at the strain associated with band A (point C in Fig. 14(b)).

Machined coupons
The full stress-strain curves of the machined tensile coupons are plotted in Fig. 15(a), while the elastic region and initial yielding are shown more clearly in Fig. 15(b). Note that only the curves derived using the DIC data are reported herein since they were almost identical to these calculated based on the strain gauge measurements and are deemed to be more accurate. A summary of the average material properties arranged by direction of testing relative to the print layer orientation (i.e. θ = 0°, 45°and 90°) is presented in Table 4, while the material properties of all coupons are reported in Table 5. In Tables 4  and 5, θ is the direction of testing relative to the print layer orientation as defined in Fig. 2, E is the Young's modulus, σ 0.2 and σ 1.0 are the 0.2% and 1.0% proof stresses respectively, σ u is the ultimate tensile stress, ε u is the strain at the ultimate tensile stress, ε f is the fracture strain measured over the standard gauge length [45] and n, m 1.0 and m u are the strain hardening exponents of the two-stage Ramberg-Osgood material model.
The results from the three tested orientations demonstrate an inherent material anisotropy, as apparent from Fig. 15 and Table 4. It was observed that, despite removing the surface undulations prior to testing to give prismatic coupons of nominally the same geometry, the coupon ig. 8. Results of sensitivity study on contour spacing. Table 3 Average geometric properties of the as-built coupons. surfaces exhibited deformations during testing that matched their build directionsee Fig. 16. Furthermore, the pattern of the longitudinal surface strain field clearly remained influenced by their print layer orientation; a typical example of this is shown in Fig. 17, where the longitudinal surface strain fields of machined coupons tested in the three orientations are plotted just after attainment of their 1% proof stress σ 1.0 . The reasons behind this are explained in Section 5. The coupons tested in the 90°orientation had the lowest 0.2% proof stress σ 0.2 , 1.0% proof stress σ 1.0 and ultimate tensile strength σ u , which would be expected since the material is being loaded perpendicular to its individual layers. The coupons tested in the 0°and 90°directions had the lowest Young's moduli, with the values observed being approximately 30% lower than the typically observed value of E = 200,000 MPa for stainless steel [49,58], while the coupons tested in the 45°orientation exhibited a significantly higher Young's modulus and slightly higher strength than the other tested orientations.

As-built coupons
The full stress-strain curves of the as-built tensile coupons of 3.5 mm and 8.0 mm nominal thickness are plotted in Figs. 18(a) and 19(a) respectively, with the early portion of the curves response shown in Figs. 18(b) and 19(b), respectively. The as-built material properties are described as 'effective' due to their dependence on the variability of the built geometry. A summary of the average effective material properties arranged by direction of testing relative to the print layer orientation (i.e. θ = 0°, 45°and 90°) and by nominal thickness t nom (i.e. 3.5 mm and 8.0 mm) is presented in Table 6, while the effective material   Tables 7 and 8 respectively, where the employed notation is the same as that used in Tables 4 and 5. Note that the material responses of the as-built coupons were determined based on the mean cross-sectional area over the parallel length of each coupon, as measured by the 3D laser scanning and the subsequent geometric analysis outlined in Section 3. It should be also mentioned that, other than the 0°coupons, the majority of the as-built coupons fractured at the cross-section of minimum thickness (i.e. cross-section of minimum area), as shown in Fig. 20, immediately after attainment of the ultimate stress σ u , resulting in the fracture-to-ultimate strain ratios ε f /ε u being approximately equal to unity.
The results of the tensile tests on the as-built coupons followed a somewhat similar trend to the underlying material properties of the machined coupons. The as-built coupons tested in the 90°orientation had the lowest effective values of E eff , σ 0.2,eff , σ 1.0,eff and σ u,eff , with E eff being up to 55% lower than the typically assumed value of E = 200,000 MPa for stainless steel. Although the values of strength were not substantially affected by the thickness of the coupons, the values of the Young's modulus in all orientations were found to be consistently lower for the thinner material; this can be explained with reference to the magnitude of the geometric undulations relative to the coupon thickness (i.e. the e z /t values), which were higher for the 3.5 mm than the 8.0 mm material, reflecting the fact that positional control during printing is essentially absolute, rather than relative to the thickness. The higher relative eccentricities induce bending in the coupons during loading, which has a detrimental effect on the effective material response.
In terms of ductility, the as-built coupons tested in the 0°orientation were the most ductile due to the load acting parallel to the deposition paths. Typical as-built coupons are shown before and after testing in Fig. 21, where the differences in fracture strain for the different orientations are clearly visible. A post-test inspection of the coupons revealed a local lack of fusion in the AB-8.0-90°-B4 specimen, which explains the lower strength and ductility relative to the other specimenssee Fig. 19(a).

Comparison between as-built and machined coupons
In this section, comparisons between the results of the as-built and machined coupon tests are presented in order to quantify the influence of the geometric variability that is inherent in the WAAM process on the material responsesee Fig. 22 and Table 9, where the effective material properties of the as-built coupons are labelled with the subscript 'eff'. As expected, the as-built specimens exhibited a drop in effective Young's modulus, 0.2% proof stress, ultimate stress, and ultimate strain (with the exception of the 8.0 mm 0°coupons on average) when compared to their equivalent machined specimens, with the 3.5 mm 90°coupons showing the most severe reductions (up to about 35%, 20%, 20% and 60% reductions in E, σ 0.2 , σ u and ε u , respectively). It should be mentioned that one of the machined coupons (M-8.0-0°-H4) showed lower ductility than the others and that a ε u,eff /ε u ratio of 1.0 is found if this individual result is excluded.

Correlation between microstructure and mechanical properties
Rapid solidification of additively manufactured metals is a common characteristic of all AM techniques and, although this is expected to affect the material microstructure and hence the material properties, research into its influence on different metals is not yet comprehensive [38]. Several researchers have observed variation in the mechanical properties of additively manufactured metals and linked this variation to differences in the internal microstructure [20,[59][60][61][62][63]. Hence, in order for the anisotropic mechanical performance of the tested coupons presented herein to be examined from a metallurgical point of view, the microstructure and crystallographic texture of the printed material has been investigated and compared to that of the parent welding wire. The experiments were carried out at the Harvey Flowers EM Suite in the Department of Materials at Imperial College London.

Employed techniques and methods
In order to examine the microstructure of the investigated WAAM plates, samples for metallographic examination were extracted from all surfaces of a typical plate, with the axes LD, TD and ND being parallel (i.e. θ = 0°), transverse (i.e. θ = 90°) and normal to the print layer orientation, respectivelysee Fig. 23.
In preparation for metallographic analysis, grinding and polishing of all samples was performed in line with ASTM: ACI 301 and 318 [64]. The specimens were polished using magnetic polishing pads and diluted OP-S lubricant, with ultrasonic cleaning undertaken at standard time intervals to remove any residual OP-S particles. Following polishing, all samples were etched to reveal their microstructure, providing information about the grain size and shape. Since this process relies on the chemical reaction between the metal surface and the etchant, selection of an acid solution appropriate for austenitic stainless steel is crucial. Since the use of glycerin reagent, recommended by ASTM E407-07 [65], was not effective in revealing the microstructure of the examined material, an acid solution proposed by Fernandes de Lima and Sankare [66] was employed.
The crystallographic texture of the samples was examined by means of electron backscatter diffraction (EBSD) using pole figures. EBSD is an advanced technique that can provide qualitative and quantitative microstructure measurements such as the distribution of grain sizes, grain shape and crystal orientation [67], while the resulting pole figure is a 2D stereographic projection revealing the locations and intensities of specified average crystal orientations relative to the examined surface. Since the quality of the EBSD diffraction patterns is quite sensitive to the surface quality of the top layer of the examined sample, all specimens were prepared until there were no visible scratches or contamination under an optical microscope. EBSD analyses were then performed at 20 keV on a Quanta Field Emission Gun (FEG) 650 scanning electron microscope equipped with a Bryher eFlash HD and analysed with Esprit 2.1. The morphology of the microstructure was imaged using FSD imaging [68].

Polishing and etching
After 30 min of polishing, the manufacturing characteristics of WAAM were apparent on the surface morphology of all samples, with the boundaries between the welding layers being clearly distinguishable, revealing a preferential growth orientation of the grains along  the TD direction; this observation is consistent with a study carried out by Liverani et al. [21], in which the grains were found to be orientated preferentially along the path with the greatest thermal gradient.

EBSD analysis
Prior to collecting EBSD data, lower forescatter diodes (FSD) equipped within the scanning electron microscope (SEM) were used Table 4 Average material properties of machined coupons by direction of testing relative to the layer direction.  Table 5 Material properties of machined coupons.
Coupon  Table 6 Average effective material properties of as-built coupons.  orientation per grain (as also illustrated by the overlayed unit cube orientation). Long and columnar grains can be observed, preferentially aligned along the TD (i.e. θ = 90°) axis. This alignment can be explained with reference to the arc welded plate being built layer by layer along the TD axis since, during solidification of most metals, grains tend to track the highest thermal gradient [20].

Variation in elastic modulus
Based on the obtained results, it has been concluded that the examined WAAM austenitic stainless steel has a strong crystallographic texture. Hence, in order to interpret the macroscopic elastic performance of this textured material, knowledge of the single crystal elastic moduli is required [69]. The Young's modulus of a single crystal in a specific direction depends on the number and strength of its interatomic bonds. The theoretical and experimental monocrystal elastic stiffness constants C 11 , C 12 and C 44 of Grade 304 austenitic stainless steel were provided in a study made by Ledbetter [70]. Hence, for the calculation of the monocrystal elastic moduli, since the elemental composition of Grade 308LSi stainless steel is similar to that of Grade 304 stainless steel, the single crystal elastic-stiffness constants of Grade 304 were adopted and used in conjunction with a set of formulae established by Armstrong et al.  [71]. The Young's modulus in the b100N type directions, corresponding to the θ = 0°and θ = 90°coupons was calculated to be 91.6 GPa. The Young's modulus in the b110N type directions, corresponding to the θ = 45°coupons was calculated as 195.5 GPa. These findings verify the results of the tensile coupon tests, according to which the value of the Young's modulus for the θ = 45°coupons was higher than the equivalent values for θ = 0°and θ = 90°− see Table 4. It should be mentioned that a greater value of Young's modulus of 314.2 GPa was calculated for the b111N type directions, which were not sampled by the conducted coupon tensile tests.

Variation in yield and ultimate strength
Although to a lesser extent than the Young's modulus, the yield and ultimate strengths were also observed to be anisotropic, with the 0.2% proof strength σ 0.2 and ultimate strength σ u of the θ = 45°coupons being higher than the equivalent values for the θ = 0°and θ = 90°couponssee Table 4. This variation is attributed to a difference in the effective mean free path for dislocations to travel in the different directions, where grain boundaries limit the dislocation motion and act as barriers in so called 'Hall-Petch' type strengthening mechanism [72][73][74].

Comparison with parent feedstock wire
In order to compare the microstructures of the parent feedstock wire and the welded material, the cross-section and side face of the feedstock wire were examined in the SEM. The FSD image of the cross-section and side face of the wire are presented in Fig. 26(a) and 26(b), respectively, while the b100N, b110N and b111N pole figures are shown in Fig. 26(c). It can be observed that the microstructure of the wire is entirely different from that of the arc welded sheet, with the feedstock material having a small grain size and a common wire crystallographic texture in which the grains are elongated along the rolling direction. The b100N and b110N pole figures demonstrate an intense crystallographic texture, with the grains within the wire stretched along the extrusion direction followed during the forming process. Hence, it may be concluded that the microstructure of the WAAM material is substantially altered due to the printing process, resulting in anisotropic mechanical characteristics.

Correlation between effective mechanical properties and as-built geometry
With the results of the tensile coupon tests performed on the asbuilt material showing a clear dependence of the effective mechanical properties on the geometric variability, expressions to describe the correlation between the effective mechanical response and the geometric characteristics of WAAM material are sought in this section.
Of all the geometric parameters determined from laser scanning of the coupons and examined in Section 3, the mean and standard deviation values of the coupon thickness (t and t sd respectively) were found to be the most influential and suitable for establishing a correlation between effective mechanical properties and geometric variability. One additional simple geometric parameter, denoted v max20 , was also introduced to enable effective mechanical properties to be estimated without the reliance on laser-scanning technology. The parameter v max20 is defined as the maximum deviation between the examined WAAM surface and a straight edge (e.g. a ruler of approximately 1 mm thickness) parallel to the direction of loading (i.e. along the length of the coupon), as shown in Fig. 27 (a), and measuring the maximum distance between the straight edge and the coupon over a 20 cm length, as shown in Fig. 27(b). This measurement should be undertaken using a slip gauge at a series of points over the surface.
Ordinary least squares regression analysis was employed to fit linear functions to the test data, providing predictions for the key effective material properties (i.e. E eff , σ 0.2,eff , σ 1.0,eff , σ u,eff and ε u,eff ). The results of the regression analysis are graphically illustrated in Fig. 28, where the examined as-built coupons are grouped by their angle θ relative to the print layer orientation, while the effective material properties of each as-built coupon are normalised by the material properties of the corresponding machined coupon (as reported in Table 4) and plotted against the examined geometric parameter (namely the standard deviation of  thickness t sd or the v max20 measurement) normalised by the average coupon thickness t. Note that in the absence of laser scanning data, the average coupon thickness would have to be estimated based on weight or by taking a series of individual calliper measurements.
The devised expressions are presented in Eqs. (4) to (13), while the coefficients of determination R 2 are reported on the respective graphs -see Fig. 28. Note that all the devised expressions were forced through the point (0,1) since the effective and underlying material properties should be equal when t sd or v max20 are zero.

Data availability
The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study.