Contents lists available at ScienceDirect
International Journal of Fatigue
journal homepage: www.elsevier.com/locate/ijfatigueVariable amplitude loading of additively manufactured Ti6Al4V subjected to
surface post processes
M. Kahlin
a,b,⁎, H. Ansell
a,c, A. Kerwin
d, B. Smith
d, J. Moverare
b a Saab AB, Aeronautics, SE-58188 Linköping, Swedenb Division of Engineering Materials, Linköping University, SE-581 83 Linköping, Sweden c Division of Solid Mechanics, Linköping University, SE-581-83 Linköping, Sweden d Manufacturing Technology Centre, Coventry CV7 9JU, UK
A R T I C L E I N F O Keywords:
Additive manufacturing Ti6Al4V
Fatigue
Variable amplitude loading Post process
A B S T R A C T
The fatigue life of additively manufactured metals need to be improved before they can be introduced to flight critical aerospace structural applications. In this study, laser powder bed fusion (L-PBF) and electron beam powder bed fusion (E-PBF) Ti6Al4V material were therefore subjected to four different surface post processes. Furthermore, variable amplitude fatigue testing were performed and compared to fatigue life predictions based on constant amplitude fatigue tests using a cumulative damage approach. The predictions were in good ac-cordance with the experimental results and post processed L-PBF and E-PBF material showed an increase in fatigue life with > 5 times.
1. Introduction
Additive manufacturing (AM), also referred to as 3D-printing, in metal has gained a lot of interest within the aerospace industry in the recent years. Originating from prototyping, AM in metal can today be used for serial production of aerospace parts with the potential to re-duce fuel consumption by either reducing weight or increasing part performance, for example through improved cooling. Moreover, even though AM still is an expensive manufacturing process, the cost to produce small serial production lots can be lowered for many applica-tions [1]. There are different AM processes which all have in common that they manufacture parts by adding material layer-by-layer in con-trast to a subtractive manufacturing process in which material is re-moved. The laser powder bed fusion (L-PBF) and electron beam powder bed fusion (E-PBF) AM processes use a laser or an electron beam, re-spectively, to melt thin layers of metal powder after which a new layer of un-melted metal powder is added; this procedure is repeated until the full geometry of one or several parts is completed.
However, metals manufactured with AM do not always exhibit the same material properties as conventionally manufactured metals. This is particularly true for fatigue properties which are influenced both by different microstructure, due to repeated heating and quick cooling, and internal defects [2,3]. In addition to this, it has been found that for Ti6Al4V, manufactured by L-PBF or E-PBF, the rough as-built (AB) surface is “the single most severe factor for fatigue” [4]. The rough as-
built surface consists of both partially melted powder particles and waviness which creates hills and valleys that will act as stress con-centrations [5]. As a consequence, the fatigue strength of additively manufactured materials is reduced up to 75% compared to con-ventionally manufactured materials [4]. Critical or highly loaded ad-ditively manufactured aerospace parts therefore need improved fatigue properties. Previous studies of additively manufactured Ti6Al4V, using either L-PBF or E-PBF, have shown that fatigue properties similar to wrought Ti6Al4V can be achieved if the rough as-built surface is re-moved by machining and the internal defects are closed by hot isostatic pressing (HIP) [4,6,7]. However, the main benefit with AM which is the freedom of design, would be limited to conventional machining geo-metries if all surfaces were machined. Therefore, other surface post processes, that can handle different degree of complex geometries, are needed to increase the fatigue properties while maintaining the possi-bility of designing complex geometries. There are many conventional surface post processes for titanium that also can be used for additively manufactured parts, for example shot peening, tumbling, chemical milling and electropolishing. The effect of many of these surface pro-cesses, on additively manufactured Ti6Al4V, have been investigated in several previous studies, both with the focus on surface roughness im-provement [8–10] and with the focus on fatigue improvements [11–13] even though fewer studies have had the latter focus.
All previous fatigue investigations on surface post processed addi-tively manufactured metals, Ti6Al4V included, have to the authors
https://doi.org/10.1016/j.ijfatigue.2020.105945
Received 27 May 2020; Received in revised form 8 September 2020; Accepted 8 September 2020 ⁎Corresponding author at: Saab AB, Aeronautics, SE-58188 Linköping, Sweden.
E-mail address: magnus.kahlin@saabgroup.com (M. Kahlin).
Available online 11 September 2020
0142-1123/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
knowledge, been performed using constant amplitude (CA) loading. However, most structural aerospace parts are subjected to variable amplitude loading [14] and findings under constant amplitude loading conditions are not always directly transferable to real aerospace ap-plications. Variable amplitude loading can give a different material response due to load sequence effects and associated local plastic de-formations [15]. Moreover, surface post processed additively manu-factured Ti6Al4V has complex stress conditions with process induced residual stresses and irregularities acting as stress concentrations at the surface. This could furthermore enhance a different material response due to rare occurring peaks and troughs during variable amplitude loading. Several different methods can be used to predict fatigue life of metal alloys. There are for example methods based on Palmgren-Miner linear damage rule, on stochastic, on energy and continuum damage mechanics methods [16]. In the present study, variable amplitude fa-tigue tests of additively manufactured Ti6Al4V, subjected to surface post processes, have been performed; experimental results were com-pared to predictions from cumulative damage calculations (Palmgren- Miner) based on constant amplitude test results. A schematic overview of the analysis is illustrated in Fig. 1.
2. Materials and experimental methods
2.1. Materials and test specimens
Test specimens in Ti6Al4V material were manufactured using both L-PBF and E-PBF processes. The specimens were produced in a single build, with L-PBF and E-PBF respectively, to avoid batch to batch var-iations. All specimens had the as-built geometry according to Fig. 2 with the load direction in the vertical (Z) build direction.
The L-PBF specimens were manufactured with an EOS M280 equipment using a layer thickness of 60 µm and a powder size between 15 and 45 µm. The default build parameters recommended by EOS for this alloy and equipment were used (laser power 280 W, hatch spacing 0.14 mm and scanning speed 1200 mm/s). The L-PBF specimens were stress relieved (SR) under high vacuum (10-4 mBar) at 730 °C for 2 h, before removed from the build plate. The E-PBF specimens were built using an ARCAM A2XX equipment with a layer thickness of 70 µm and a powder size between 45 and 105 µm. The default build parameters recommended by GE Additive for this alloy and equipment were used (maximum beam current 17 mA, hatch spacing 0.2 mm, and speed function 36). The E-PBF specimens were further blasted with titanium powder, to remove loosely bound powder, and HIP:ed at 920° using 103 MPa pressure for 2 h (HIP cycle fulfilled the requirements of ASTM F2924/F3001 2012).
2.2. Post processes
Four surface post processes were investigated in this study; cen-trifugal finishing (CF), linishing (Lin), shot peening (SP) and laser shock peening (LSP). The goal with these surface post processes was to in-crease the fatigue life by reducing the surface roughness and/or through introducing compressive residual stresses in the material sur-face. Process parameter development was performed, for each post process, through an iterative procedure using both surface roughness measurements and simulations. However, fatigue testing was only performed on the specimens finished using the final set of process parameters. Different process parameters were needed for L-PBF and E- PBF material for some of the post processes due to the inherent dif-ferences of the surface roughness of these AM methods.
2.2.1. Centrifugal finishing
Centrifugal finishing uses the relative motion of parts, abrasive media and carrying agents, by rotating a barrel at a high RPM, to polish the surface of the parts. Centrifugal finishing is a non-target process, and over-polishing can result in rounded corners of the part. Therefore, process development is needed to achieve the required overall surface finish and material removal without significant loss of definition or rounding of part features. The centrifugal finishing process was per-formed in three steps with fresh media used at the start of each step: 1.) cutting, 120 min with media SFB 10 × 10 with 50 ml LQ18, 2.) smoothening, 90 min with media CFB 6 × 10 with 50 ml LQ16 and 3.) polishing, 60 min with media PTM run dry.
2.2.2. Linishing
Linishing uses high-speed abrasive brushes to remove material and smoothen the surface. The linishing process was performed in three steps: 1.) bulk removal, 2.) micro form finishing and 3.) polishing, using finer abrasive grit for each step. Robot controlled brushes with force
Nomenclature AB as-built AM additive manufacturing CA constant amplitude CF centrifugal finishing D life ratio
Dth theoretical cumulative damage sum E-PBF electron beam powder bed fusion
FALSTAFF Fighter Aircraft Loading STAndard For Fatigue HIP hot isostatic pressing
Lin linishing LOF lack of fusion
L-PBF laser powder bed fusion LSP laser shock peening
Nf cycles to fatigue failure Nf,test experimental fatigue life Nf,pred predicted fatigue life
ni number of applied type i fatigue cycles
Ni number of type i load cycles that equals fatigue failure R stress ratio
RO run out
SEM scanning electron microscope σa stress amplitude
σm mean stress σmin minimum stress SP shot peening SR stress relieved
Sz maximum height of the scale limited surface VA variable amplitude
Fig. 1. Overview of the evaluation sequences in this study. CA = constant amplitude, VA = variable amplitude.
feedback were used which gave higher precision repeatability and en-abled a maintained contact force even where the part geometry would deviate from the computer model. Linishing was performed with the rotation in the specimen loading direction leaving the linishing marks in this direction. The feed rate of the brushes varied between 125 mm/ min, at the gauge length section, down to 50 mm/min at the start of the grips. For L-PBF material, 0.3 mm of material was removed from the as- built surface while 0.5 mm was needed for E-PBF material.
2.2.3. Shot peening
Shot peening introduces compressive residual stresses to a surface by striking the surface with steel shots which will give cold working of the material. In addition to this, the shot peening will also have a smoothening effect for very rough surfaces, such as additively manu-factured materials. The shot peening process was performed according to the automatic shot peening standard AMS2430 with Ø0.279 mm cast steel shots with hardness 55–62 HRC (ASH110 media) using 200% coverage. The Almen intensity was measured to 0.0093″ A.
2.2.4. Laser shock peening
During laser shock peening, the part is covered by a thin layer of water and a pulsed laser beam is fired towards the part. The laser pulse travels through the transparent water layer and hit the surface of the part. The laser strike causes surface ablation and a rapidly expanding plasma is generated. Forces from this interaction are contained by the water (tamping layer) and prevented from expanding outward off the surface. Instead the forces are contained and spread as a compressive stress wave into the surface, which will induce cold working into the surface of the part [17]. A pulsed Nd:YAG laser with a power density of 5.3 GW/cm2 was used with a 2 mm laser beam diameter, 12 ns pulse lengths and a maximum of 10 Hz pulse rate. The overlap of beam spots was 45%.
2.3. Surface roughness investigations
The surface topography was investigated by a HITACHI SU-70 scanning electron microscope (SEM) with a field emission gun oper-ating at 15 kV. Surface texture measurements were performed using an Alicona SL focus variation microscope. The L-PBF data was analysed with the software Infinite Focus Measurement Suite by Alicona and the E-PBF with the software MountainsMap by Digital Surf. The surface texture measurements were taken over an area of 2 mm2 in the centre of the specimens gauge length. The surface texture parameter Sz (max-imum height of the scale limited surface) was determined as the average roughness from two or more specimens, per test series, with a minimum of three measurements per specimen. Curvature corrections were performed in order to calculate the surface texture parameters from the curved surfaces of the specimens; the surfaces were levelled using least-square mean plane by subtraction, then filtered with S- and L-filters with nesting index 5 µm and 250 µm respectively.
2.4. Internal porosity investigations
The internal porosity was estimated by investigation of cross-sec-tions using light optical microscopy. The images were analysed using the open source image processing software ImageJ [18] with a threshold of individual porosity > 5 µm2. Two cross-sections with X-Y (area ~ 25 mm2) and Z-X (area ~ 105 mm2) planes were investigated on both E-PBF and L-PBF specimens with as-built surfaces.
2.5. Constant amplitude fatigue design data
The constant amplitude fatigue behaviour of surface post processed additively manufactured Ti6Al4V was investigated in a previous study by Kahlin et al. [11]. The test specimens in the present study were manufactured in the same AM builds with the same geometries, see Fig. 2, as the material from the previous study. The specimens from the present and the previous study were furthermore heat treated and post processed in the same batch. The material properties for the test ma-terial in the present study can therefore be expected to be comparable to the material from the previous study.
The constant amplitude fatigue data from Kahlin et al. [11] can be used to produce Wöhler curves for stress ratio R = 0.1. However, more complete fatigue data is needed, e.g. in the form of Haigh diagrams which includes data from Wöhler curves with several stress ratios, in order to perform cumulative damage calculations according to the Palmgren-Miner rule [19,20], see Eq. (1).
Ten variants of Haigh diagrams, one for each combination of L-PBF or E-PBF and post process were constructed by using existing Haigh diagrams for investment cast Ti6Al4V as a template. The stress ampli-tude (σa) for the Wöhler curve for investment casting with R = 0.1 was scaled by a linear adjustment function, as schematically illustrated in Fig. 3, to fit the constant amplitude test data from Kahlin et al. [11]. This means that any effect from the post processes in form of residual stress or local stress concentrations due to rough surfaces is included in the scaled Wöhler curves. The scaling procedure was performed at three positions along the Wöhler curve, i.e. 1 000, 100 000 and 5 000 000 cycles. The linear adjustment function was then further applied to the full Haigh diagram, from the Saab Aeronautics company material da-tabase, for investment cast Ti6Al4V in order to produce a Haigh dia-gram for each test series in this study.
2.6. Aircraft load spectrum with variable amplitude
Fatigue predictions and experimental tests have been performed using variable amplitude loading. The load sequence used was a fighter aircraft wing bending spectrum FALSTAFF (Fighter Aircraft Loading Fig. 2. Fatigue specimen. a.) As-built specimen (machined grips), b.) As-built
nominal dimensions before post processing. All dimensions are in millimetres (mm).
Fig. 3. A schematic illustration of the adjustment procedure to fit the design data for investment cast Ti6Al4V to the test data used in this study. A and B are constants.
STAndard For Fatigue) which was originally jointly developed by NLR- Netherlands, LBF-Germany, IABG-Germany and F&W-Switzerland [21]. The FALSTAFF spectrum was further developed by CEAT [22] into a modified version, Short-FALSTAFF, that has about 50% less load cycles but still generates almost all of the cumulative fatigue damage as the original spectrum. One load sequence of this Short-FALSTAFF load spectrum corresponds to 200 simulated flights and has 18 012 turning points i.e. 9 006 rain-flow counted cycles, see Fig. 4 and Fig. 5. The Short-FALSTAFF load spectrum can be tested with different maximum tensile peak loads and all cycles in the spectrum are set as a fraction of the maximum tensile peak load, so that the inherent relationship be-tween the cycles stays the same. The Short-FALSTAFF load sequence is repeated during a fatigue test until fatigue failure occurs.
2.7. Fatigue life prediction with a cumulative damage approach
A full Haigh diagram was constructed for each test series, i.e. for each combination of AM process and post process, see section 2.5. The Haigh diagram for each test series respectively and the rain-flow counted cycles for the Short-FALSTAFF load sequence, illustrated in Fig. 6, was used for cumulative fatigue damage calculations according to the Palmgren-Miner rule [19,20], see Eq. (1).
= = D n N th ik i i 1 (1)
Dth is the theoretical cumulative damage sum, ni is the number of
applied type i load cycles, and Ni is the number of type i load cycles that
will result in fatigue failure. The rain-flow counted cycles for one load sequence of the Short-FALSTAFF spectrum are presented as data points
in a Haigh diagram format in Fig. 6 together with schematic curves for constant fatigue life (Nf = constant). Wöhler curves that fit each data point are extracted from the Haigh diagram in order to calculate the corresponding number of cycles to fatigue failure for each data point. For interpolation purposes, the Wöhler curves are extracted in three different ways:
1.) With constant minimum stress (σmin) for stress ratio R > 0, 2.) With constant mean stress (σm) for R < -1 and
3.) By interpolation of stress amplitude (σa) between different R-values for −1 ≤ R ≤ 0.
2.8. Variable amplitude fatigue tests
Variable amplitude fatigue tests were performed with a Short-FALSTAFF load spectrum using a maximum tensile peak stress of 670 MPa. The load level of 670 MPa was set based on the prediction from constant amplitude tests, to give fatigue failure for all variable amplitude test series, except for L-PBF subjected to centrifugal fin-ishing, within the range of 2 000–50 000 flights. The fatigue tests were performed in room temperature at 10 Hz using load control in a servo hydraulic fatigue test rig. An Instron ± 50 kN load cell and an Instron 8800 control system were used and the Short-FALSTAFF load sequence was repeated until the specimen fractured or until a fatigue life of 50 000 flights was exceeded. All test series that had run-out tests or fatigue failure close to the run-out limit were further tested with a maximum tensile peak stress of 910 MPa. Run-out specimens were not re-tested, instead spare specimens were used for the 910 MPa level. Each test series at 670 MPa level consisted of 3 tests (E-PBF AB surface had only 2 tests due to limited amount of test specimens) and each series at 910 MPa of 2 tests. The fracture surfaces for all fatigue spe-cimens were further studied to determine crack initiation locations by stereomicrography. The majority of the specimens were then further studied by SEM for a more detailed view of the crack initiation locations using a HITACHI SU-70 with a field emission gun operating at 15 kV.
3. Results
3.1. Surface roughness
The investigations of the material cross-sections show that both the E-PBF and L-PBF material were almost fully dense. Based on the cross- Fig. 4. One load sequence, 200 flights, of the short-FALSTAFF load spectrum
[21,22]. This sequence is repeated until failure.
Fig. 5. Peaks and troughs distribution for one sequence, 200 flights, of the short-FALSTAFF load spectrum [21,22].
Fig. 6. A schematic Haigh diagram for one Short-FALSTAFF load sequence, with rain-flow counted cycle data points and interpolation procedure. The size of each data point is proportional to the logarithmic number of cycles, i.e. a large point corresponds to a large number of cycles in the load sequence. Figure reprinted with permission from Elsevier [23].
section investigations, the E-PBF and L-PBF material were estimated to be 99.99% and 99.98% dense respectively which is similar to levels found in previous studies [24–26].
The surface roughness for the four post processes are compared to the roughness of as-built surfaces in Fig. 7. The E-PBF material has higher roughness than L-PBF material, even after post processing, which indicates that all the as-built E-PBF roughness could not be re-moved during the surface post processing. This is confirmed by the surface topographies presented in Fig. 8 in which E-PBF generally has a rougher surface generally compared to corresponding L-PBF surface. 3.2. Constant amplitude fatigue design data
Constant amplitude test results from Kahlin et al. [11] were used to produce constant amplitude design data for R = 0.1 which are pre-sented in Fig. 9 and Fig. 10. An existing fatigue Wöhler curve for Ti6Al4V investment casting, from the Saab Aeronautics company ma-terial database, was scaled with a linear adjustment function to fit to the test data points from each test series. All test series were very well fitted to the scaled investment casting R = 0.1 Wöhler curves except for E-PBF material subjected to linishing which had a couple of test points with deviant behaviour, see Fig. 9.
3.3. Experimental variable amplitude fatigue life
Variable amplitude fatigue tests were performed with a maximum tensile peak stress of 670 MPa for all test series, and with 910 MPa for those test series that had run-out tests (> 50 000 flights) or fatigue failure close to the run-out limit at 670 MPa. The results are presented in Fig. 11.
3.4. Predicted variable amplitude fatigue life
The full Haigh diagrams, scaled with the same linear adjustment function as the R = 0.1 Wöhler curves presented in section 3.2, were used to predict the fatigue life for the Short-FALSTAFF load spectrum with a theoretical cumulative damage sum Dth = 1.
A life ratio [27], see Eq. 2, is calculated to illustrate how well the predicted fatigue life corresponds to the experimental fatigue life. A life ratio of D = 1 would then indicate that the predicted and the experi-mental fatigue life are equal.
D = Nf,test / Nf,pred (2)
in which Nf,test is the experimental fatigue life and Nf,pred the pre-dicted fatigue life.
The predicted fatigue life, for maximum tensile peak stress 670 MPa, is compared to the average experimental results in Fig. 12 and to all experimental test points in Fig. 13. Furthermore, the scatter band is illustrated with the maximum and minimum life ratio D for
average fatigue life in Fig. 12 and for all test points in Fig. 13. 3.5. Fatigue crack initiations
An equal amount (50%) of the test specimens had the main crack initiating at the surface respectively at a near-surface position just below the surface. The main crack starting position for each test series is presented in Table 1. The average number of crack initiation points is presented in Table 2 and it can be seen that the test series with the main initiations in the near-surface region generally have fewer initiation points whereas series with the main initiations at the surface have more crack starting points. Images of typical crack initiations for each test series are presented in Fig. 14. Beach marks can furthermore be seen in the overview images in Fig. 14; these were not found after constant amplitude testing [11] and are therefore most likely an effect of high peak loads during the variable amplitude loading.
4. Discussion
4.1. Fatigue improvement after post processing
The variable amplitude loading fatigue behaviour for surface post processed L-PBF and E-PBF Ti6Al4V has been investigated and an overview of the fatigue life is presented in Fig. 11. The fatigue life in-creased considerably after surface post processing both for L-PBF and E- PBF material, see Table 3.
As illustrated by Fig. 11, the shot peened or centrifugal finished L- PBF material had by far the longest fatigue life of all the test series. However, it is notable that even though all centrifugal finished L-PBF samples outperformed the shot peened L-PBF material and were stopped for run-out (> 50 000 flights) at maximum tensile peak stress 670 MPa, the centrifugal finished L-PBF test series were equal to shot peened L-PBF at the higher load level 910 MPa. This indicates that even a direct comparison between post processes, as performed in this study, should be regarded with some degree of caution since loading or geo-metrical aspects could influence the performance of various test series in different ways.
The fatigue life of E-PBF material was increased considerably after surface post processing as illustrated by Fig. 11 and Table 3. But even though the increase in fatigue life is large for both E-PBF and L-PBF material, see Table 3, the fatigue life of as-built E-PBF is much lower than the fatigue life for as-built L-PBF material, see Fig. 11. This means that even if the E-PBF fatigue life is increased considerably after post processing it is still inferior to the corresponding post processed L-PBF material. The reasons for this will be further discussed in the succeeding sections.
4.2. Fatigue crack initiations
The average number of crack initiation locations on the fracture surfaces are presented in Table 2. All test series were tested with the same loading, a maximum tensile peak stress of 670 MPa, only L-PBF subjected to centrifugal finishing and shot peening were further tested at 910 MPa. However, even though tested at the same maximum load level, this maximum load of 670 MPa can be considered to be a very high fatigue load for some of the test series, such as L-PBF or E-PBF with as-built or laser shock peened surfaces (see the constant amplitude Wöhler curves in Fig. 9 and Fig. 10), while the maximum load of 670 MPa is a considerable less severe fatigue load for other test series. Moreover, test samples subjected to high stresses that have defects acting as stress concentrations usually have multiple cack initiations [28–30]. E-PBF and L-PBF test series with as-built surface and E-PBF test series subjected to laser shock peening had the most initiation lo-cations (> 8), see Table 2. These three test series had the highest sur-face roughness, see Fig. 7 and Fig. 8, and the lowest constant amplitude Wöhler curves, see Fig. 9 and Fig. 10, of all the test series. This could Fig. 7. Surface roughness. Average values of areal surface texture parameter, Sz
give these test series both high stress concentrations at the surface and a high maximum load relative seen on the Wöhler curve. The combina-tion of these factors can then promote multiple crack initiacombina-tions. 4.3. Effect of material removal
Two of the post processes, linishing and centrifugal finishing
removed a considerable amount of material from the surface, see Table 4. It is not likely that any surface valleys remain after linishing post processing, for neither E-PBF nor L-PBF material. This is because the average removed material during linishing exceeds the average Sz (maximum height) surface roughness, see Table 4, which is the sum of the largest peak and the deepest valley within a measurement area (2 mm2 in this study). The valleys at the linished surfaces from which Fig. 8. SEM images of surface topography of as-built and post processed conditions. Images are of the rounded side of the fatigue test specimen gauge length. a.) E- PBF as-built, b.) L-PBF as-built, c.) E-PBF centrifugal finished, d.) L-PBF centrifugal finished, e.) E-PBF linished, f.) L-PBF linished, g.) E-PBF shot peened, h.) L-PBF shot peened, i.) E-PBF laser shock peened, j.) L-PBF laser shock peened.
Fig. 9. E-PBF constant amplitude test data and design curves for R = 0.1. Test data from ref [11]. AB = as-built, CF = centrifugal finishing, Lin = linishing, SP = shot peening, LSP = laser shock peening.
Fig. 10. L-PBF constant amplitude test data and design curves for R = 0.1. Test data from ref [11]. AB = as-built, CF = centrifugal finishing, Lin = linishing, SP = shot peening, LSP = laser shock peening.
Fig. 11. Average fatigue life for test with maximum tensile peak stress 670 MPa and 910 MPa. Standard deviation is presented as error bars. RO = run out tests. Fig. 12. Average experimental fatigue life versus predictions for tests with maximum peak stress 670 MPa. Lines indicate the scatter band of the life ratio D. Run-out test are not included.
Fig. 13. Experimental fatigue life versus predictions for tests with maximum peak stress 670 MPa. Lines indicate the scatter band of the life ratio D. Run-out tests are not included.
the cracks start from, see Fig. 14i-l, are therefore likely to be internal LOF that have been opened to the surface during the post processing. In the case of E-PBF material, which was HIP:ed, the internal LOF:s need to have had an open channel to the surface which prevented them from being closed during the HIP process.
L-PBF and E-PBF material subjected to linishing had similar fatigue life, as illustrated by Fig. 11, which most likely is a consequence of the full removal of the rough as-built surface. The linished fatigue life was a top result for the E-PBF material but only a mediocre result for L-PBF which had even longer fatigue life after shot peening or centrifugal finishing. In contrast to linishing, it is instead very likely that the deepest valleys have not been fully eliminated for the E-PBF material that was subjected to centrifugal finishing since the average Sz, see Table 4, is larger than the amount of removed material. This corre-sponds well with the shape of the rounded valley, presented in Fig. 14e- f, which is the major crack initiation point for centrifugal finished E- PBF material. The average removed material during centrifugal fin-ishing of L-PBF material is slightly higher than the maximum surface roughness height Sz, see Table 4. This indicates that it possible but unlikely that any surfaces valleys will remain after the centrifugal finish process. The main crack initiations for the L-PBF centrifugal finished material were found to be near-surface LOF:s, see Fig. 14g-h, which support the idea that all of the roughness was removed during post processing.
4.4. Effect of surface roughness and surface residual stress
The variable amplitude fatigue life from the present study has been compared to surface residual stress measurements and surface rough-ness, see Fig. 15 and Fig. 16. The residual stress was measured in a previous study [11] which was performed with material from the same AM builds and the same heat treatment and post processing batches as the present study. The residual stress measurements were performed at the surface in the loading direction (Z-direction) of the fatigue speci-mens using X-ray diffraction. For E-PBF material it seems like the sur-face roughness is the dominating factor as illustrated in Fig. 15, as all test series fit well to a linear trendline. The residual stress on the other hand seems to have less impact on the fatigue life of E-PBF material, see Fig. 16, since material subject to either centrifugal finish, shot peening or linishing has a similar fatigue life even though the surface residual stress varies between −550 MPa to + 50 MPa.
The L-PBF material on the other hand shows a clear trend of in-creasing fatigue life with increased compressive surface residual stress
which is illustrated with a linear trendline in Fig. 16. Furthermore, reduced surface roughness has a positive effect on fatigue life for L-PBF material, as illustrated in Fig. 15. This means that the most favourable combination of residual stress and surface roughness will give the longest fatigue life. For example, shot peening can introduce beneficial compressive residual stresses but at the same time induce surface or near surface irregularities that can act as local stress concentrations that is unfavourable for the fatigue life. The effect of residual stress therefore depends on the competitive relationship between compressive residual stress and local stress concentrations. E-PBF material has the dis-advantage of a rougher surface and larger powder particles which can create larger remaining defects acting as local stress concentrations even after surface post processing, see Fig. 14. The local stress con-centrations of post processed E-PBF material would then have a larger negative effect on the fatigue life than any positive effect from the compressive residual stresses in this study. It should however be noted that only surface residual stress investigations have been performed. The residual stress profile in the depth direction could have a large effect on the fatigue life in which both a large depth and a large mag-nitudes of the compressive residual stress are important to increase the fatigue life [31,32]. The trends discussed in this section for surface roughness and residual stress were also found for constant amplitude fatigue testing [11] even though the trends were more distinct after variable amplitude loading.
4.5. Effect of microstructure and defects
The microstructures of as-built L-PBF and E-PBF material are dif-ferent due to the inherent differences in the two processes and the subsequent heat treatments. The microstructure of both the L-PBF and E-PBF material, used for this study, with as-built and post processed surfaces was characterised in a previous study [11], see Fig. 17. It was found that the microstructure of L-PBF material, without surface post processing, had a Widmanstätten structure with needle like α’-phases together with a mixture of α + β phases. The E-PBF material had a coarser α + β Widmanstätten microstructure which can be expected due to the more extensive heating of the material both during the E-PBF process and HIP. Furthermore, no changes to the microstructures of the as-built L-PBF and E-PBF material could be observed after any of the post processes investigated in the present study. The microstructure has most likely only a minor influence on the fatigue life of the investigated samples due to the presence of very large surface and near surface defects, presented in Fig. 14, that act as stress concentrations [5] and therefore will promote fatigue crack initiations.
The surface and near-surface crack initiations of the surface post pro-cessed test series were found to originate from different defects, see Table 1, which can be summarized to three major type of flaws: 1.) surface defects that have been compressed and pushed down below the surface during surface post processing, which is the case for shot peened and laser shock peened test series seen in Fig. 14m-t, 2.) remaining valleys from the as-built surface which have not been fully removed during post processing, which is the case for E-PBF material subjected to centrifugal finishing presented in Fig. 14e-f and 3.) internal LOF defects that have become surface or near- surface defects after the post process material removal which is the case for the linished test series, see Fig. 14i-l.
Table 1
Main fatigue crack initiation locations.
Post process E-PBF L-PBF
Surface Near-surface Position Surface Near-surface Position
None (as-built) X Rough surface X Rough surface
Centrifugal finishing X Remaining surface valleys X LOF
Linishing X Prior subsurface LOF X Prior subsurface LOF
Shot peening X Embedded prior surface features X Embedded prior surface features Laser shock peening X Embedded prior surface features X Embedded prior surface features Table 2
Average number of crack initiation points on fracture surface for variable am-plitude loading with maximum tensile peak stress 670 and 910 MPa. All tests with L-PBF CF at 670 MPa were run out tests.
Few (1–2) Medium
(3–8) Many (> 8) E-PBF SP (near-surface) E-PBF CF E-PBF AB
E-PBF Lin L-PBF Lin E-PBF LSP (near-surface) L-PBF CF (near-surface)910 MPa L-PBF AB
L-PBF SP (near-surface)670 and 910 MPa L-PBF LSP (near-surface)
The defects are generally larger for E-PBF material due to the ori-ginally rougher as-built surface, larger powder particles and larger layer thicknesses used compared to L-PBF which can be seen for each post processed condition in Fig. 14. The larger surface and near-surface defects of E-PBF material, compared to L-PBF material are most likely Fig. 14. Fatigue crack initiation locations for each material and post process. Arrow indicates main crack initiation location. a-b.) E-PBF as-built, c-d.) L-PBF as-built, e-f.) E-PBF centrifugal finished, g-h.) L-PBF centrifugal finished, i-j.) E-PBF linished, k-l.) L-PBF linished, m-n.) E-PBF shot peened, o-p.) L-PBF shot peened, q-r.) E- PBF laser shock peened, s-t.) L-PBF laser shock peened.
Table 3
Average fatigue life after variable amplitude loading with maximum tensile peak stress 670 MPa. The fatigue life is compared to fatigue life for as-built E- PBF material and as-built L-PBF material respectively.
Post process E-PBF + HIP L-PBF + SR
None (as-built) Reference level2 588 flights Reference level7 950 flights Centrifugal finishing x 3.3 > x 7.1 (run-out tests)
Linishing x 5.4 x 2.3
Shot peening x 4.0 x 5.0
Laser shock peening x 1 x 1.1
Table 4
Surface roughness and amount of removed material during post processing. E-PBF L-PBF Surface roughness Sz (maximum height) 236 µm 165 µm Removed material during linishing 520 µm [11] 270 µm [11]
one of the major reasons why the post processed E-PBF, and as-built E- PBF (if the rough surface is defined as surface defect) test series gen-erally have a lower fatigue life.
The internal porosity level was only measured for the test series with as-build surface but observations of the cross-sections of the sur-face post processed test series indicate that there are no difference in the amount of internal pores for the surface post processed test series. Furthermore, even if the total amount of internal porosity was esti-mated to be below 0.02%, see section 3.1, occasional internal flaws can
still have an effect on the fatigue properties if the as-built surface de-fects are completely removed as discussed above for samples subjected to linishing.
4.6. Comparison between prediction and experimental results
In this study, the fatigue life predictions with the cumulative da-mage approach were based on Haigh diagrams derived from the con-stant amplitude material behaviour for each post processed test series at the stress ratio R = 0.1. The FALSTAFF load sequence is a tensile dominated spectrum, which is illustrated in Fig. 4, and the Haigh dia-grams derived from the material behaviour of R = 0.1 Wöhler curves should therefore give relevant results. The comparison between pre-dicted, theoretical cumulative damage sum Dth = 1, and experimental variable amplitude test results are presented in Fig. 12 and Fig. 13. The average experimental results show a life ratio, D, for all test series within 0.7 ≤ D ≤ 1.8 which should be considered to be very consistent with the predicted Dth = 1 cumulative damage sum. This shows that even though the various surface post processes give different surface and near-surface features as well as different residual stress situations, the fatigue test results from constant amplitude loading can be used to predict the fatigue life of a variable amplitude loaded sample, at least for tensile dominating load spectrums. The individual test results, presented in Fig. 13, show a somewhat larger scatter 0.5 ≤ D ≤ 2.8 for all test series which also should be considered to be a relative low scatter compared to the predicted Dth = 1 cumulative damage sum. In a previous study [23] on Ti6Al4V with Short-FALSTAFF load spectrum, the experimental results had a life ratio of 0.8 ≤ D ≤ 1.3 for E-PBF and L-PBF material with rough as-built surfaces, and 0.8 ≤ D ≤ 1.6 for wrought material which is in the same order of magnitude as the life ratio found in the present study.
Furthermore, the fatigue life of the vast majority of the test points were underestimated, D > 1, by the predictions, see Fig. 13, which means that the experimental fatigue life exceeded the predicted fatigue life. Only four test points had a life ratio D < 1, which were all three of the L-PBF tests samples subjected to shot peening and one L-PBF sample subjected to linishing. One reason for this situation could be that the L- PBF samples subject to shot peening or linishing showed the largest scatter of all test series both after constant amplitude loading, see Fig. 10, and variable amplitude loading, see Fig. 13. This can be at-tributed to large differences in the size of the critical surface or near- surface flaws from which the cracks start from. Both the linished and shot peened L-PBF series had one single test point, with a shorter variable amplitude fatigue life compared to the rest of the test series, which had a considerably larger surface/near-surface defect compared to the other samples.
Fig. 15. Effect of surface roughness, maximum height of the scale limited surface (Sz), on fatigue life from the present study with max peak stress 670 MPa.
Fig. 16. Residual stress at surface, from Kahlin et al.[11], compared to fatigue life from the present study with max peak stress 670 MPa.
Fig. 17. Microstructure of Ti6Al4V material without surface post processing, B.d. = build direction. a.) L-PBF b.) E-PBF. Figure reprinted with permission from Elsevier [11].
4.7. Comparison to previous variable amplitude loading investigations There are very few published studies on variable amplitude fatigue loading on additively manufactured materials. Even for conventionally manufactured materials, there are few published studies, which could be due to the fact that variable amplitude loading investigations are often performed in-house by end-user companies rather than by aca-demia. However, there is one previous variable amplitude loading study by Kahlin et al. [23] on L-PBF and E-PBF Ti6Al4V with as-built surfaces compared to wrought Ti6Al4V, and one study by Sarkar et al. [33] on L- PBF stainless steel 15–5 PH with machined surfaces. In these previous studies the cumulative damage from variable amplitude fatigue loading was predicted by Palmgren-Minor’s rule, see Eq. (1), using material data from constant amplitude loading in the same way as in the present study. The predictions were then further compared to experimental variable amplitude loading test results which are presented in Table 5 together with data for wrought Ti6Al4V. The experimental results from these studies show that symmetric tensile-compression load spectrums give lower life ratio, D < 1, which means that the constant amplitude data overestimates the variable amplitude fatigue life, in contrast to the tensile dominating spectrums which generally show D > 1. One reason for this difference could be that the compressive stresses in a symmetric load sequences may relax existing compressive residual stresses which would then be unfavourable for the fatigue life.
Overall it seems, based on the limited available studies, that the variable amplitude fatigue response for additively manufactured ma-terials, with as-built, machined and surface post processed surfaces, can be predicted using a cumulative damage approach with constant am-plitude fatigue data in the same ways as for conventionally manu-factured materials. Both tensile dominated load sequences and sym-metric tensile-compression load sequences give a life ratio in the same range, see Table 5, for both additively manufactured materials and conventional materials.
5. Conclusion
The primary goal for this study was to investigate which effects different surface post processes have on additively manufactured Ti6Al4V material, both L-PBF and E-PBF, when subjected to variable amplitude loading. Additively manufactured materials with rough as- built surfaces were subjected to either centrifugal finishing, linishing, shot peening or laser shock peening.
Furthermore, it was investigated how well the fatigue life for vari-able amplitude loading could be predicted with the use of constant amplitude test results and cumulative damage calculations.
•
A considerably longer fatigue life could be achieved by surface post processing for both E-PBF (x 5.4 fatigue life by linishing) and L-PBF (> x 7.1 fatigue life by centrifugal finishing) material compared to as-built conditions. The increase in fatigue life depends both on the surface roughness and the amount of residual stresses.•
L-PBF Ti6Al4V achieved generally superior fatigue properties com-pared to E-PBF in both the as-built and in surface post processedconditions. This is due to the inherent larger surface roughness of the as-built E-PBF material which still has a large effect on the properties after surface post processing.
•
L-PBF Ti6Al4V subjected to either shot peening or centrifugal fin-ishing showed superior fatigue life compared to all other material conditions.•
The experimental variable amplitude fatigue life (Short-FALSTAFF load spectrum) was predicted with good agreement using a cumu-lative damage approach with data from constant amplitude testing. The cumulative damage approach can therefore be considered to accurately predict the fatigue life, at least for tensile dominating load spectrums, for variable amplitude loaded additively manu-factured and surface post processed Ti6Al4V material.Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influ-ence the work reported in this paper.
Acknowledgments
The authors are grateful to Saab AB and the Swedish Foundation for Strategic Research, project ID 14-0060, for financial support. Moreover, part of this study was performed within the project AddMan funded through the Clean Sky 2 joint undertaking under the European Union’s Horizon 2020 research and innovation programme under grant agree-ment No 738002. Dr Lewis Newton at Nottingham University and Dejan Basu at MTC are furthermore acknowledged for surface roughness measurements, and the staff at Sandwell-UK for their contribution to the shot peening process.
References
[1] Frazier WE. Metal additive manufacturing: a review. J Materi Eng Perform 2014;23(6):1917–28.
[2] Rafi HK, Karthik NV, Gong H, Starr TL, Stucker BE. Microstructures and Mechanical Properties of Ti6Al4V Parts Fabricated by Selective Laser Melting and Electron Beam Melting. J Materi Eng Perform 2013;22(12):3872–83.
[3] Liu QC, Elambasseril J, Sun SJ, Leary M, Brandt M, Sharp PK. The Effect of Manufacturing Defects on the Fatigue Behaviour of Ti-6Al-4V Specimens Fabricated Using Selective Laser Melting. Adv Mater Res 2014;891–892:1519–24. doi:10. 4028/www.scientific.net/AMR.891-892.1519.
[4] Kahlin M, Ansell H, Moverare JJ. Fatigue behaviour of notched additive manu-factured Ti6Al4V with as-built surfaces. Int J Fatigue 2017;101:51–60. [5] Chan KS. Characterization and analysis of surface notches on Ti-alloy plates
fabri-cated by additive manufacturing techniques. Surf Topogr Metrol Prop 2015;3:44006. doi:10.1088/2051-672X/3/4/044006.
[6] Greitemeier D, Palm F, Syassen F, Melz T. Fatigue performance of additive manu-factured TiAl6V4 using electron and laser beam melting. Int J Fatigue 2017;94:211–7.
[7] Günther J, Krewerth D, Lippmann T, Leuders S, Tröster T, Weidner A, Biermann H, Niendorf T. Fatigue life of additively manufactured Ti–6Al–4V in the very high cycle fatigue regime. Int J Fatigue 2017;94:236–45.
[8] Boschetto A, Bottini L, Veniali F. Surface roughness and radiusing of Ti6Al4V se-lective laser melting-manufactured parts conditioned by barrel finishing. Int J Adv Manuf Technol 2018;94(5-8):2773–90.
[9] Urlea V, Brailovski V. Electropolishing and electropolishing-related allowances for powder bed selectively laser-melted Ti-6Al-4V alloy components. J Mater Process
Table 5
Mean life ratio for the present and previous variable amplitude loading studies.
Material Surface condition Load spectrum Mean life ratio, D Reference
Ti6Al4V, L-PBF and E-PBF As-built Tensile dominated (Short-FALSTAFF) 1.6–1.7* Present study Ti6Al4V, L-PBF and E-PBF As-built Tensile dominated (Short-FALSTAFF) 1.0–1.5* Kahlin et al. [23]
Ti6Al4V, L-PBF and E-PBF Surface post processed Tensile dominated (Short-FALSTAFF) 0.7–1.8* Present study Ti6Al4V, wrought machined Tensile dominated (Short-FALSTAFF) 1.2 Kahlin et al. [23]
15–5 PH, stainless steel machined Tensile (stress ratio R = 0) 0.7–1.2* Sarkar et al. [33]
15–5 PH, stainless steel machined Symmetric (stress ratio R = -1) 0.5–0.6* Sarkar et al. [33]
Ti6Al4V, wrought machined Symmetric (Gaussian narrow band random loading) 0.3 Cardrick et al. [34]
Technol 2017;242:1–11.
[10] Longhitano GA, Larosa MA, Munhoz ALJ, Zavaglia CAdC, Ierardi MCF. Surface Finishes for Ti-6Al-4V Alloy Produced by Direct Metal Laser Sintering. Mat. Res. 2015;18(4):838–42.
[11] Kahlin M, Ansell H, Basu D, Kerwin A, Newton L, Smith B, Moverare JJ. Improved fatigue strength of additively manufactured Ti6Al4V by surface post processing. Int J Fatigue 2020;134:105497. https://doi.org/10.1016/j.ijfatigue.2020.105497. [12] Denti L, Bassoli E, Gatto A, Santecchia E, Mengucci P. Fatigue life and
micro-structure of additive manufactured Ti6Al4V after different finishing processes. Mater Sci Eng, A 2019;755:1–9.
[13] Persenot T, Buffiere J-Y, Maire E, Dendievel R, Martin G. Fatigue properties of EBM as-built and chemically etched thin parts. Procedia Struct Integrity 2017;7:158–65. [14] Marsh G, Wignall C, Thies PR, Barltrop N, Incecik A, Venugopal V, Johanning L.
Review and application of Rainflow residue processing techniques for accurate fa-tigue damage estimation. Int J Fafa-tigue 2016;82:757–65.
[15] POMMIER S. Cyclic plasticity and variable amplitude fatigue. Int J Fatigue 2003;25(9-11):983–97.
[16] Santecchia E, Hamouda AMS, Musharavati F, Zalnezhad E, Cabibbo M, El Mehtedi M, Spigarelli S. A review on fatigue life prediction methods for metals. Adv Mater Sci Eng 2016;2016:1–26.
[17] LSP Technologies, Inc. 2019. https://www.lsptechnologies.com/laser-peening-pro-cess.php (accessed July 24, 2019).
[18] ImageJ, https://imagej.net (accessed August 20, 2020).
[19] Palmgren AG. Die Lebensdauer von Kugellagern. Zeitschrift Des Vereines Dtsch Ingenieure 1924;68:339–41.
[20] Miner MA. Cumulative damage in fatigue. J Appl Mech 1945;12:159–64. [21] Joint publication of Flugzeugwerke Emmen. Switzerland; Laboratorium für
Betriebsfestigkeit (LBF), Germany; National Aerospace Laboratory (NLR), Netherlands; and Industrie-Anlagen-Betriebsgesellschaft mbH (IABG) G. FALSTAFF: Description of a fighter aircraft loading standard for fatigue evaluation; 1976. [22] CEAT Report M7681900, Centre d’Essais Aeronautique de Toulouse, Toulouse.
1980.
[23] Kahlin M, Ansell H, Moverare JJ. Fatigue behaviour of additive manufactured Ti6Al4V, with as-built surfaces, exposed to variable amplitude loading. Int J Fatigue 2017;103:353–62.
[24] Krakhmalev P, Fredriksson G, Yadroitsava I, Kazantseva N, Plessis A du, Yadroitsev I. Deformation Behavior and Microstructure of Ti6Al4V Manufactured by SLM. Physics Procedia 2016;83:778–88.
[25] Ahlfors M, Bahbou F, Ackelid U. Optimizing HIP and printing parameters for EBM Ti-6Al-4V. Quintus Technol: Whitepaper; 2018.
[26] Vandenbroucke Ben, Kruth Jean‐Pierre. Selective laser melting of biocompatible metals for rapid manufacturing of medical parts. Rapid Prototyping Journal 2007;13(4):196–203.
[27] Schütz W, Heuler P. Miner’s rule revisited. Bordeaux, France: An assesment fatigue damage crack growth Predict. Tech; 1993.
[28] Zerbst U. Application of fracture mechanics to welds with crack origin at the weld toe—a review. Part 2: welding residual stresses. Residual and total life assessment. Weld World 2020;64(1):151–69.
[29] Zhu Shun-Peng, Hao Yong-Zhen, Liao Ding. Probabilistic modeling and simulation of multiple surface crack propagation and coalescence. Appl Math Model 2020;78:383–98.
[30] Mishnaevsky Jr Leon, Fæster Søren, Mikkelsen Lars P, Kusano Yukihiro, Bech Jakob Ilsted. Micromechanisms of leading edge erosion of wind turbine blades: X‐ray to-mography analysis and computational studies. Wind Energy 2020;23(3):547–62. [31] Wang Haidou, Zhu Lina, Binshi Xu. Residual Stresses and Nanoindentation Testing
of Films and Coatings. Singapore: Springer Singapore; 2018.
[32] Farrahi GH, Lebrijn JL, Couratin D. EFFECT OF SHOT PEENING ON RESIDUAL STRESS AND FATIGUE LIFE OF A SPRING STEEL. Fat Frac Eng Mat Struct 1995;18(2):211–20.
[33] Sarkar Sagar, Kumar Cheruvu Siva, Nath Ashish Kumar. Investigation on the mode of failures and fatigue life of laser-based powder bed fusion produced stainless steel parts under variable amplitude loading conditions. Addit Manuf 2019;25:71–83. [34] Cardrick, A. W. , Perrett BHE. Fatigue tests on plain specimens of titanium 6Al-4V
