EXPERIMENTAL STUDY OF THE EFFECT OF TRIAXIALITY RATIO ON THE FORMABILITY LIMIT DIAGRAM AND DUCTILE DAMAGE EVOLUTION IN STEEL

AND HIGH PURITY COPPER

G. Chiantoni1*, N. Bonora2, A. Ruggiero3
1

Tenaris-Dalmine - Dept. of R&D p.zza Caduti 6 Luglio 1944, 1 24044 Dalmine (BG) Italy [email protected] 2 University of Cassino Dept. of DiMSAT Italy [email protected] 3 University of Cassino Dept. of DiMSAT Italy [email protected]

ABSTRACT: In this contribution, the effect of the stress state (by mean of the triaxiality ratio) on the equivalent plastic strain at fracture is experimentally investigated for different steels and high purity copper in the range of triaxiality from 0 (pure torsion) up to 1.2. This range of stress state is particularly meaningful for the majority of the mechanical applications. The prediction on the formability limit diagram for proportional loading paths (i.e. with constant triaxiality) is verified in the framework of Continuum Damage Mechanics (CDM) using the LemaitreChaboches damage model. Furthermore the influence on triaxiality on the threshold plastic strain for damage initiation and for the critical damage at rupture is experimentally studied. The results are compared with the prediction of the aforementioned damage model. A very good agreement between experimental results and CDM prediction is achieved. KEYWORDS: damage, triaxiality, formability limit, torsion test

1 INTRODUCTION
The first concern of this work is about the fracture locus of ductile materials. This is usually intended as the diagram of stress triaxiality ratio (Tx, minus hydrostatic pressure divided by von Mises equivalent stress) versus equivalent von Mises plastic strain (PEEQ). The importance of triaxiality is due to the fact that it strongly affects the fracture behavior of ductile materials. Therefore it is important to know how Tx influences the obtainable fracture strain in a wide triaxiality regime (from 0 up to 1.2. This range includes many of the mechanical processes of industrial interest). This question is the subject of numerous papers. Bao and Wierzbicki [1] performed test series on aluminum alloy sheets and found the fracture ductility is strongly dependent on the average value of Tx. In particular they observed that the value of PEEQ at fracture for triaxiality between 0 and 0.4 is less then that found for values greater than 0.4 (there is a cuspid for Tx equals 0.4). This result is in contrast with the prediction made by Continuum Damage Mechanics (CDM) models (Lemaitre and Chaboche [2], Bonora [3]) largely used, even in commercial FEM code, to predict ductile failure. The dependency of the fracture locus on the stress triaxiality ratio has to be experimentally investigated in order to evaluate the reliability of these damage models
____________________ * Corresponding author: postal address, phone, fax, email address

and improve the understanding of ductile fracture. This was done by comparing the experimental results of austenitic steels and high purity copper specimens with a strong variation in their stress triaxiality ratio. The second concern is about the effect of damage on the elastic properties of ductile materials. It is well known (Lemaitre [4]) that the damage has a direct influence on the Youngs modulus of metals. In fact the measurement of the degradation of the mechanical elastic response is one of the most used methodologies to experimentally quantify the evolution of damage during a deformation step. Recently Voyiadjis and Kattan [5] studied the possible influence of damage on Poissons ratio. This item has a particular importance on the experimental investigation of damage performed with torsion tests, because the damage is calculated from the Youngs modulus which depends on both the elastic tangential modulus (G) and the Poissons coefficient. The third concern of this paper is about the experimental measurement of the equivalent plastic strain threshold for the onset of damage. Recently Celentano and Chaboche [6] proposed a mixed numerical and experimental method based upon traction tests. But in a traction test the damage assumes a significant relevance after the necking has been started. We propose a different experimental methodology based on torsion tests performed on thin hollow specimens.

2 EXPERIMENTAL DATA
2.1 MATERIALS Experiments were carried out with three ductile materials: Ferritic high alloy steel ASTM A335 P91 grade (9% of Chromium 1% Molybdenum), which is used in the energy industry as a line pipe material. The 0.2% offset yield strength is 620 MPa. The austenitic grain size varies between 20 and 50 m. The specimens were prepared from a rolled rod with a diameter of 120 mm Carbon Manganese steel Vanadium microalloyed, grade X65 API 5L PSL1/PSL2. The 0.2% offset yield strength is 550 MPa. The austenitic mean grain size is about 15 m. The specimens were prepared from hot rolled tube. As received high purity 99.9% copper. The 0.2% yield strength is 107 MPa. The specimens were prepared from rod with diameter of 18 mm. 2.2 TORSION AND TRACTION TESTS The torsion tests were performed using the plastometer STD810 from Baehr Company. It is equipped with a hydraulic drive that is able to serve a maximum torque of 100 Nm. The torsion tests were done in quasi-static condition with a strain rate of 0.001 s-1 and at 0.0005 s-1 for the hollow specimen torsions The machine has a continuous laser measurement of gauge length diameter to enable precise calculation of transversal deformation. It can perform axial deformation both in tension and compression up to 20 kN. The traction tests were done with a standard MTS 810 servo-hydraulic universal testing machine in strain control using an extensometer with a gauge of 20 mm at strain rate of 0.00001 s-1. All the experiments were carried out at room temperature. 2.3 SPECIMEN GEOMETRY Axisymmetric specimens were used for both torsion and traction tests. Figure 1 shows the geometry of the hollow torsion specimen with a thickness (in the gauge length) of 0.5 mm. This thin wall is needed to ensure an almost uniform penetration of the plasticized zone avoiding the axial elastic core of the traditional cylindrical geometry. This configuration is particularly well suited for studying the threshold of damage, but is not able to resist up to material rupture mainly due to the buckling instability. To reduce this undesired phenomenon a graphite rod was introduced in the hollow during torsion to get a better geometrical stability. The graphite was chosen because is a dry lubrificant. In order to expand the range of triaxiality a round notched geometry was studied. This configuration permits to experimentally realize an almost proportional loading path with a triaxiality ratio of about 1.2.

Figure 1: Torsion hollow specimen

3 FINITE ELEMENT SIMULATIONS

Experiments were planned for a wide range of triaxiality. To find the parameters critical for fracture, finite element simulations of traction tests with smooth and round notched specimen and of torsion test with hollow specimen were performed in ABAQUS/Standard. The necessary Youngs modulus and flow stress curve were taken from the experimental torsion data torque vs. twist angle using a methodology described in Wu e Xu [7]. The lack of geometric instability (necking) in torsion test has enabled to use the measured true strain-true stress curve up to high plastic strain without the need to extrapolate or numerically manipulate the curve obtained by the standard torsion test (up to necking). An isotropic hardening material model and an eight node biquadratic elements with reduced integration were used for all simulations. Specifically traction tests were simulated with the axisymmetric CAX8R element while torsion tests used the axisymmetric CGAX8R element with twist. The modelled specimens were deformed in displacement control as in the experiments.

4 DISCUSSION
4.1 FORMABILITY DIAGRAM The three ductile materials have been selected for their very different alloy compositions (from pure copper to 9% chromium alloy). As far as the damage is concern, copper is considered the reference material. As illustrated in the figure 2, the triaxiality in traction tests (in smooth and notched specimens) can not be considered a constant through the loading path. So the mean value of triaxiality as calculated by the FEM simulations at the centre of the specimen was used to plot the formability diagram. The strain at failure too was extracted by the FEM at the specimens centre. The failure condition was determined by comparing the shape of the actual broken specimen with the shape of the corresponding simulation.

Figure 5: Formability diagram for 99.9% copper Figure 2: Triaxiality in traction test (notched specimen)

The gauges of the torsion specimens were accurately polished in order to reduce the initial roughness. However, the deformation of the torsion specimens lead to an increased surface roughness and consequently to a localization of the deformation. Due to this phenomenon, small cracks were initiated and lead to further localized deformation. The measured global shear strain can be lower then the local one, which is responsible for the fracture. Thus the reported values of the strain at failure for the torsion test could be an underestimation of the true values.

From the formability diagram reported (fig. 35) the aforementioned discontinuity observed by Bao and Wierzbiki [1] in an aluminium alloy could not be confirmed for these three materials. It is worth mentioning that the same result has been found by Tratting [8]. In the diagrams it has been sketched the fitting obtained using the expected locus as predicted by a CDM damage model proposed by Bonora [3]. A qualitative agreement between the models predictions and experimental results is confirmed. 4.2 DAMAGE VS POISSONS RATIO The hereafter investigations were carried out on P91 steel. In order to detect an influence of damage on Poissons coefficient it was planned a torsion test broken up into loading-unloading cycles alternating with compression steps within the elastic range. Every torsion step can be used with two purposes: the elastic response as a measure of the damage, while the plastic deformation as the damaging agent. After every torsion step (i.e. after the complete unloading) the specimen was subjected to an elastic compression during which were recorded both the axial and diametric strains. The Poissons ratio is then given, in absolute value, as the ratio between the transversal and axial strains. The Youngs modulus was calculated using the measured values of the tangential modulus and Poissons ratio as: E = 2(1 + )G . Given step by step the values of E and the damage is calculated as D = 1 E / E0 where

Figure 3: Formability diagram for steel P91

E0 is the first step value. The results obtained are summed up in figure 6: while the damage is increasing as the material accumulates plastic strain, the Poissons ratio seems to be unaffected. The values were spread out around the mean value of 0.28 without showing any clear trend.

Figure 4: Formability diagram for steel X65

increasing trend. The threshold (indicated in red with a value of 0.07) has been chosen as the half value (in strain) between the two sets.

Figure 6: Poissons ratio and damage vs. PEEQ

4.3 DAMAGES THRESHOLD The aforementioned CDM models are very sensitive to the damage threshold (plastic strain at which damage starts). It can be measured with loading-unloading tensile cycles. [2] The drawback of this method lies in the geometrical instability (necking) suffered by the material pulled up to fracture. As shown in [6] the necking induces a fictitious increasing of the Youngs modulus, thus causing a great experimental difficulty in the threshold measurement. The method proposed in this paper is based on loadingunloading cycles of torsion of a hollow specimen. FEM simulation was carried out to get the plastic strain accumulated in a single torsion step. A snapshot is shown in figure 7 where the maximum equivalent plastic strain is plotted. It is worth noting that the actual rupture of the specimen runs exactly along the red line calculated by the FEM as the zone of maximum plastic strain.

Figure 8: Damage threshold by torsion test

5 CONCLUSIONS
An extensive experimental campaign using three ductile metals with different ductility was carried out in order to get the materials formability diagrams (equivalent plastic strain to failure vs. triaxiality) and to validate the prediction of a CDM model. A good agreement between measurements and model was found. Moreover it was verified that (within the experimental accuracy) the Poissons ratio was not affected by the materials damage. Finally it was proposed a new method to measure the damage strain threshold based upon torsion test on thin walled hollow specimen. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ This paper is published under authorization of Tenaris Dalmine R&D, directed by M.A. Rossi

REFERENCES
[1] Bao Y., Wierzbicki T.: On fracture locus in the equivalent strain and stress triaxiality space. Int. Journal of Mech. Sci. 46:81-98, 2004 [2] Lemaitre J., Chaboche J.M.. Mechanics of Solids Materials. Cambridge Academic Press, 1985. [3] Bonora N., A Non Linear CDM Model for Ductile Failure. Eng. Fracture Mech., 58:11-28, 1997. [4] Lemaitre J., A Course on Damage Mechanics, Springer Verlag , Berlin, 1992 [5] Voyiadjis G., Kattan I., A Comparative Study of Damage Variables in Continuum Damage Mechanics Int. Journal of Damage Mechanics, online edition, Nov. 2008. [6] Celentano D. Chaboche J.M., Experimental and Numerical Characterization of Damage Evolution in Steels, Int. Journal of Plasticity, 23:1739-1762, 2007. [7] Wu H., Xu Z., The Shear Stress-Strain Curve Determination from Torsion Test in the Large Strain Range, Journal of Testing and Evaluation, JTEVA 20:396-402, 1992 [8] Trattnig G., Antretter T., Pippan R., Fracture of Austenitic Steel Subject to a Wide Range of Stress Triaxiality Ratios and Crack Deformation Modes, Eng. Fracture Mech. 75:223-235, 2008.

Figure 7: FEM of hollow specimens torsion

In every torsion step the mean increasing of the plastic strain was 0.016. Using the result reported in the previous paragraph, the damage was calculated measuring only the degradation of the tangential elastic modulus and considering constant (0.28) the Poissons coefficient. The diagram of damage vs. PEEQ is reported in figure 8. The damage measurements are clearly divided in two groups: the first one has a nearly vanishing mean value, while the second shows an