(Computational Materials Science 87 (2014) 241-247 Contents lists available at ScienceDirect Computational Materials Science journal homepage: www.elsevier.com/locate/eommatsci oe Study the effect of martensite banding on the failure initiation in dual-phase steel ‘A. Ramazani**, Z. Ebrahimi, U. Prahl® *beparment of Mater Scene onl Engncerng, Univesity of Michigan, 500 Heyward St A Arbor, M8105, USA * beartmen of MectancalEgineing Paya Noor Unies FO BOX 19595-3097 Teta an “bxparoment of ers Mel, ROTH Aachen Universi, mas 1, 52072 Aachen, Cerna) Mews ARTICLE INFO AUSTRACT ‘rice os “This work aims to study the effect of martensite banding on the failure initiation in dual-phase (DP) steel, Received 23 August 2013 ‘A microstructure based approach using representative volume elements (RVE) is utlized to evaluate the Rec in evi frn 25 Jamar 2014 microstructure deformation and the failure initiation on the mesoscale. Mini tensile test with digital image cortelation (DIC) analysis was carried out and linked to local scanning electron microscopy (SEM) analysis to identify macroscopic failure initiation strain values. n stu analysis of bending test, in SEM combined with electron backseaterdifrction (EBSD) measurements before and after the test ‘showed that crak initiation occurs in martensite islands. Representative volume element (RVE) model- Pe ee oe ing combined with extended finite element method (XFEM) was applied to simulate martensite cracking Monee ‘on mesoscale. XFEM failure parameters have been identifed based on local and macroscopic mini tensile EBs0 ‘evaluation applying classical J-integral theory. The identified parameters were validated by comparing ‘vale oie 13 March 2014 eywords: Dual-phase (OP) ste! Represcotatve volume clement (RVE) Extended iit clement method (XFEM) the predictions with the experimental results © 2014 Elsevier BY. Al rights reserved 1. Introduction ual-phase (DP) steel is characterized by a microstructure con- sisting of hard martensite particles dispersed in a soft and ductile ferrite matrix. Due to this morphology, DP steels show a good com- bination of strength and ductility so that they are demanded extensively for the automotive industry [1-2 Based on several researchers’ reports dual-phase steels fail ina ductile manner [4-9]. These research works have described that the void initiation takes place around or inside martensite grains Afterward, growth and coalescence of voids occurs due to ferrite failure. Speich and Miller (4) have reported that at high Vj they ob- served martensite cracking as main failure mechanism. Kim and Thomas [5 have reported that the formation of voids in dual-phase steels depends on the morphology of the martensite. Maire et al. {| observed both martensite cracking and ferrte/martensiteinter- ace debonding during in situ tensile tests with three dimensional X-ray tomography. Calcagnotto et al. [9] investigated failure behavior of DP steels with different grain sizes using mini tensile test with DIC technique and showed that while in classical DP steels (of coarse grain and high impurity) the cleavage fracture © Gorespoting autho. Tel: +1734 615 8815; fox +1 734753 478% mal odes rrazsnimih as (A. Ramazan, nap door 10.1016jcommatse2014.01.051 (27 0P56/0 2014 Evie BY. Argh reserved and grain split cause failure, in modem DP steels (of fine grain 4nd low impurity) grain boundaries play the significant role. Recently, failure modeling in DP steels is a subject of interest for several research groups [10-16]. They approach the topic through different scientific methods. Ramazani et al. (10] have studied the failure of DP steels macroscopically using Gurson-Tvergaard-nee- dleman (GIN) model. Luo and Wierzbicki [11] have investigated the failure behavior of DP steel sheet during stretch-bending oper- ations using Mohr-Coulomb (MMC) ductile fracture criterion. Sun etal, [12] and Choi etal. 13] studied the failure mode and ultimate ductility of DP steels using a microstructure-based model at difer- ent loading and boundary conditions. Uthaisangsuk et al. (14) investigated the failure of DP steels using RVE approach. They lized a cohesive zone and GTN models to study the ferrite-mat tensite debonding and ferrite degradation, respectively Vajragupta eta, [15] and Ramazani et al. (16,17 have investigated the failure behavior of DP steel using XFEM. This research work aims to investigate the effect of martensite banding on the failure initiation behavior in DP steel utilizing XFEM, 2. Experimental procedure Equiaxed and banded DP. microstructures were processed through laboratory heat treatment from a DP600 steel composition that was delivered in a cold-rolled, ferrite-pearite state with a22 ‘A amas Computational Material Since #7 (2014) 241-247 cS MPs G NM Aw © thickness of 1 mm using a Trebel hot deformation simulator [18]. ‘Table | shows the chemical composition. To obtain the same per- ‘centage of martensite at different heating rates, 790 °C and 760 °C ‘were chosen as temperatures and heating rates of 1C/s and 100°C/s, respectively, were used. After annealing, the samples were quickly gas-quenched at a rate of 80 °Cjs to form martensite from austenite. Different annealing temperatures were chosen in ‘order to get the same percent of martensite in both microstruc- tures. Eventually, martensite phase fraction was identified based ‘on metallography as 35% in both cases. ‘To quantify the distribution of martensite in the laboratory-pro- duced DP steel, three parameters were taken into account: the average height, the average length, and the aspect ratio (length to-height ratio) of martensite islands in DP microstructures. Digi- mizer software was used for this purpose, and the analysis was performed automatically [19]. For an aspect ratio close to 1, the microstructure was described as an equiaxed microstructure. This ‘was obtained ata slow heating rate of | Cis, while at a fast heating rate of 100°C/s, a banded microstructure was obtained. The aspect ratio of martensite bands was 7.2 for this particular heating rate. ‘he ferrite grain size was respectively quantified as 5.6 and 6.1 jum for equiaxed and banded microstructures through the li eat intercept method (ASTM Standard E 112), Using Thermocale 'TCFE6.7 according to the approach as described in [20], carbon content in ferrite and martensite was calculated as 0.0035% and, 0.216% for equiaxed microstructure and as 0.0045% and 0.214% for banded microstructure, respectively. In order to identify the crack initiation and the correspondent local strain, three parallel mini tensile tests with digital mage cor- relation (DIC) technique were carried out parallel (0 the rolling direction (21, The geometry and dimensions of tensile test speci- ‘men are illustrated in Fig. 1(a), During the tests, images were taken from the surface in every second. So, strain distributions can be ‘captured in different deformation steps in the evaluated zone SEM Analysis was performed to observe the microstructure of the material in the centerline along loading direction in order to ‘observe where the crack initiates and to identify the responsible strain (Fig. (b)) [16.17]. In order to identify the failure initiation mechanism and to val- idate the failure initiation modeling in situ three point bending test with EBSD measurement before and after the test was carried out ‘at room temperature on the DP steel specimen with equiaxed microstructure using a 5000 N/800°C bending module by Kamm- rath & Weiss. The test procedure and the sample geometry are schematically shown in Fig. 2. Rolling ditection was considered parallel tothe major stress direction. On astrip of 11 mm x 60 mm tm (a) b= 20m ‘omm 13mm Fig. 2. Bending test procedure and sample geomety for LC-SEM in sty bending with a thickness of 1mm, 2 three-point bending test was performed, In order to better localize the area of probable crack formation, the strip was notched at the sides at the position close to the central rod ofthe bending device. During the bending test, the applied force was increased incrementally. However, the load was held stepwise after a specified deforma- tion in order to record the surface images ofthe samples in the LC- SEM. The LC-SEM was adjusted to obtain an SEM image focusing on the area between two pre-existing notches in both sides in order to ‘atch the process of crack initiation at the surface while displace- ‘ment-controlled load was applied to the samples. The LC-SEM analysis was made on the surface at the tensile loaded side of the bended specimen as the loads will be highest here. During loading, displacement and load were recorded automatically. Addi- tionally, SEM images were taken at any moments when changes in ‘mesoscale structure or damage growth occurred on the specimen’s surface, Hence, the mechanical loading causing the microcrack forma- tion could be identified precisely and it can be identified where the crack stats in the microstructure, Additionally, EBSD measure- ‘ments were made before and after the test in the same surface area in order to identify whether martensite or ferrite/martensite inter- face fails first Finally, 10, IPF map and Kernel average misorienta- tion map before and after the in situ test were compared with ‘micromechanical failure simulation in order to validate the model- ing approach, 43, Micromechanical modeling Ina former study, the acceptable sizeof the RVE was considered, as a minimum of 24 um while it should include at least 19 mar- tensite particles [19]. A 2D RVE from SEM images was developed using an in-house program (see Fig. 3). Quadratic elements with the size of 0.25 jm were used. Periodic boundary conditions were imposed on the RVE. The elastic modulus for ferrite and martensite i assumed to be 210 GPa, The flow curve of individual phases for ferrite and martensite is quantified based on a dislocation density based strain hardening model. The full approach has been given in 4 former study (20.22), (b) Fig 1 (2) Geomety and dimensions of the mini geometry inst tensile est specimens and (b) positions a SEM analysis in enter ine of broken specimen‘A Ramacani et al /Computaina Matra Scece 87 (2018) 241-207 a Fig. 3. Sletion ofthe RUE rom a) equated and () banded ncrostrutures In order to model the failure initiation in DP steels, XFEM with traction-separation law was utilized, The onset and propagation of cracking in quasi-static problems were studied using the XFEM since it allows describing crack growth along an arbitrary, solu- tion-dependent path without remeshing the FE model. The mas ‘mum principle stress criterion (MAXPS) is used for describing, damage initiation. Martensite is assumed to lose its stiffness as soon as the damage initiation criterion is reached and linear degra- dation behavior with steep slope is assigned as local damage evo- lution law on mesoscale [15-17] 2D RVE generated based on real micrographs can involve all microstructural features, but itis not able to predict the 3D mate- tial flow precisely, The predicted flow curves of 20 modeling tay have to be correlated with the 3D flow stress calculation ap by introducing a correlation factor (Eq. (1)), which was developed by Ramazani etal. (22) 2x 104 x (68)? x VR —1 x 10 4110? x VE 4 0.0218 x (28)? x V3, 0.0015 (0) x VE 79 WO x VE + 0.18 x (08)? Ven $0,007» (28) Vin + 0.0036 x Vy ( O(a x (0) x VA, +) 4. Results and discussions ‘Mini tensile test was carried out and analyzed using DIC tech- nique to investigate the microstructural evolution during tensile test and to quantify the local strain along the tensile axis. Combi ing these local strain values with the SEM analysis along the center Tine transverse section of the tensile sample) allows the matching of failure initiation position and local strain in the center of de- formed specimen. During tensile test, images were taken at the surface in every second, Then, strain distributions can be captured in different deformation steps in the evaluated zone, as shown in Fig. 4 for both equiaxed and banded microstructures. ‘After mini tensile test, the broken samples showed visibly neck- ing near the fracture surface because the width and thickness de- creased while the specimens were elongated. Thus, the facture mode in DP steel is ductile and it should be identified where the imicrocrack starts. Fiz. 5 shows the microstructure of DP steel, ‘where crack initiation occurred, by means of SEM in the central line at transverse section ofthe tensile sample along loading direc- tion in the investigated DP steel. As can be observed inthis figure, the failure initiates in martensite independent of the microstruc- ture. The local strain at failure initiation position in tensile was measured as 0.085 and 0.06 for equiaxed and banded microstruc- tures, respectively. Measurements details can be found in [16} Fig. 6 shows the load displacement curve for DP steel with eq) axed microstructure obtained in the in situ LC-SEM bending exper- Jment. The bending test was performed up toa load of about 130 N and a displacement of the central rod up to 5 mm, Small drops in the curve correspond to a pausing of the experiment in order to capture SEM images at displacements of 1,2, 3,4, and 5 mm. Cor- responding secondary-electron images for 1, 3, and 5mm are shown in this figure as well. They clearly show the development of the surface topography due to plastic deformation of the mate- rial. Microcracks (marked) are visible fr displacement of 5 mm. In ‘order not to deform the specimen too much and to maintain mea- surability with EBSD, the test was stopped after a displacement of '5 mm where first mierocrack formation has been identified EBSD measurements before and afte in situ bending test were ‘made to identify whether martensite phase or ferite/martensite interphase fails first in the equiaxed microstructure. Comparing the 10, IPF map and Kernel average misorientation map before and after the in situ test showed that the crack initiation occurs (a) © teats i Distance rom Fracture Surface (um) ° LL Distance From Fracture Surface [um] Fig. 4. Lacal strain distuton along conta line during loading dieeton on DP sec fr) equated and) banded microstructures1, KD‘A Ramacani et al /Computaina Matra Scece 87 (2018) 241-207 ms Fig. 7. (2) SEN IP and KAW analysis of quantification offal initton in DP ste with equiaxed microstructure: before (upper images), and ater ower images) the inst bending est. in martensite islands. In Fig. 7 the upper row shows the micro- structure before the test. The image quality (1Q) and the inverse pole figure (IPF) map aligned to the later direction of tensile stress (parallet to the horizontal axis) confirm the expected ferritic-mar- ‘ensitic microstructure: All points are indexed as bec-iron, but the ‘martensite shows considerably smaller IQ values and higher vatia- tion in the orientation due to the higher dislocation density in these regions. Due to its higher dislocation density, itcan be easily distinguished fom ferrite by its significantly lower image quality (1Q) and confidence index. In the present work, a grain-average IQ of less than 4800 before, and 4500 after bending as threshold for the martensite was used (however, the large precipitate in the lower left is erroneously also covered by this criterion). As grain criterion, the default settings of the software (grain boundaries 5° minimum 2 pixels per grain) were considered. Adal in the kernel average misorientation (KAM, over a radius ‘of 300nm with a maximum misorientation of 5°) plot zones of ‘ransformation-induced higher dislocation density also in the surrounding of the martensite particles can be found. The lower row shows the influence of the deformation during, bending on the microstructure. Specifically in the IQ map the developed microcracks can be seen as black regions inside the martensite particles or at bound- aries between martensite particles and ferrite grains, in the IPF ‘maps the grain-internal deformation can be seen by the variation in the color especially within the ferrite grains, which were quite homogeneous in orientation before the bending. The KAM map shows that the deformation especially concentrates along lines al- ‘most perpendicular to the direction of tensile stress, which can be ‘connected to the formation of glide steps at the surface. Calculated single phase flow curves of ferrite and martensite are shown in Fig. 8. For XFEM parameters identification for martensite in the studied DP steels, the maximum principle stress criterion ‘was defined with respect to the fracture strain of martensite. To identify the equivalent crack initiation strain in martensite, 2D VE micromechanical modeling was performed until macroscopic Ferrite (a) Von Mises stress (MPa} (b) oa 00 008 ate 04 Equivalent Plastic strain] Equivalent Paste strain (] Fig Calculated re stress-strain Mow curves fr (a eit ani (b) martensite for DP sel with 25% martensite, equine and banded icetruties8 ‘A amas Computational Material Since #7 (2014) 241-247 state of fallure initiation (z,= 0.085 for equiaxed and <= 0.06 for banded microstructures). Then, using first order homogenization strategy [23], the corresponding average equivalent plastic strain in martensite was estimated to be 1?! ~ 0.029 and «9! — 0.014 for equiaxed and banded microstructures, respectively. The corre- spondent stress for this strain ean be considered as critical stress (¢,) for martensite cracking in DP steels and it was calculated from, the flow curve of martensite (eg. = 1737 MPa and 4, = 1652 MPa for equiaxed and banded microstructures, respectively). Critical ‘energy (E,) for martensite failure was considered as the area of be- hind martensite stress-strain curve and calculated as 20.47 and 19.12 })m? for equiaxed and banded microstructures, respectively (a) Fx 8 (3) Damaye evolution (mirocrak formation in martensite) and () comparison between experimental and simulated tre tess eel (a) [24.25], The developed failure parameters will be utilized in XFEM in order to simulate the crack initiation in both DP steels. ‘Alter implementing XFEM on the martensite particles, numeri- cal tensile tests were carried out on the generated 2D RVES of 'DP600 steel Figs. S(a) and 10(a) respectively show the contour plot ‘of von Mises stress and damage evolution on microscale for the ‘global martensite failure initiation strain s=0.085 and ¢-=0.06 for equiaxed and banded microstructures. A comparison between ‘experimental and RVE calculated true stress-true strain curves is ‘demonstrated in Figs. 9(b) and 10{b). Also, the corrected 2 flow ‘curve to 3D flow curve is shown using correlation factor as ex- plained in Section 3. It can be seen, that the corrected 3D flow eo 06s 0m —00TS 0.05 cult Ponte Sint} (b) ue strain cues for equiaxed DP Sealant ant (b) Fg 10.3) Damage evfution (mirocrac formation in martensite) and (b) comparison between experimental and simulated true stess-tre tran curves fr banded DP‘A Ramacani et al /Computaina Matra Scece 87 (2018) 241-207 a curve is in good agreement with the experimental flow curve in both conditions. Comparison between Figs. 9(a) and (b). and 10(a) and (b) reveals that the simulations show roughly the same cracked positions as experiments. 5. Conclusions Martensite cracking has been identified to be the main failure initiation mechanisms in equiaxed and banded DP steels. Failure initiation takes place in the equiaxed microstructure in the higher plastic strain compared to the banded microstructure. Therefore, DP steel with equiaxed microstructure shows better fallure behav- {or than banded one. Mini tensile test with DIC technique offers the ‘quantification of martensite failure initiation as function of macro- scopic loading. Quantitative identification of parameters for XFEM ‘modeling of martensite cracking has been performed and validated bby numerical versus experimental tensile test comparison. Com- parison between experimental and simulated flow eurves shows ‘good agreement, Acknowledgments ‘This research was cartied out under Project Number MC2.07293, in te framework of the Research Program ofthe Materials Innova- tion Institute M2i (www.m2Ln. The mini tensile tests were car- ried out at MPIE Dusseldorf, the support of Stefan Zaefferer and his group is highly appreciated. References 111 compete: Hourman.O. Hun JP. Laurent, A Reynaet SAE Trans. Matt ‘Manuf. 110 2001) 37-47 [3] Assman thre Ura in eben Mus, AS Herman K Kayantash (Eds) Sacral Motels and frosses in Tronsprston ‘ey VCH 201. pp. 5-8, {41 ER Spe RL Mller Srctre ad Propet of Dua Phase Stes, TMS Sint Wendte, 870 pp. AS. Is] ki Thoma Metal Tran 1241961 489-488, fol Vi ss} Gotan, Mater Set Eng 95 (1987) 131-108 DLE fea an, CA Yar) Ato Ma EP. 9.) (a Eta, 0, outs 1D. Mil ver, Acta Mats 1(208) 4954-4968 Io] caleagmto¥ ach. Pos Rane, At Maer 39,2) (2011 638- i [00] {Ramazan MLAs Ura W lk, Comput. Mate Sl 4 (2012) 101- [ut] Miu. ei nS strc. 4 (22-2) 2010 3084-2102 {ha XSun‘es cnt WA LA ee in jst 25 (205) 85, 113] Cho WN X Sam MA Khe Met Mater Tran A 40(2009)796- . Jha] Vr biaisangsuk UP W. lc, Pc Eng 1 (1) (209) 171-178, HS]. vajragupea,V. Uihasangsl 8 SehmlingS MUnsterann, A Hare Wintec Conia Mater S54 O13) 27am 1061 Ramazan A Sch Ace Urb. lock Comput, Mae. S75 fo) 35 [07] Aitamaran, A chwes, A. Art, rab Key Eg. Mater. 586 (2004) 67-7, {hs Srkcnsnns ros Mts ate Tans a 207) 928-39, Us) Arn hss a We comp a (202) (20) A Ramazan K. Moker A. Schwa Garvan Urb W Bleck Pata (213) 1 (21g Sth ea ii ton igi ms Crstion {22 Aumann ss, , P, W lck Maer. ng ASO Gory ra {23] Ve Koen, Computations Homogenization othe Mula Anas bi vulnpsse Merge m0 Tess, Fchne!Unierty Fndhoven, The etna 202 24] Ge Daer Mechanical Mealy, scone McGraw il. 1976 [25] teers Vine Herebereefoaton and race Mcchanis bv togncenng ston tho W201