Sie sind auf Seite 1von 8

Computational Materials Science 34 (2005) 274281 www.elsevier.

com/locate/commatsci

Modeling heat treatment of steel parts


B.L. Ferguson *, Z. Li, A.M. Freborg
Deformation Control Technology, Inc., 7261 Engle Road, Suite 105, Cleveland, OH 44130, USA Received 4 August 2004; received in revised form 15 February 2005; accepted 24 February 2005

Abstract The ability to achieve high strength and toughness by heat treatment is a primary advantage of steel alloys. However, the development of internal stress and geometric distortion accompanies these hardening processes. Simulation of heat treatment processes must include the evolution of microstructural phases in order to calculate the mechanical behavior of the composite microstructure as the alloy changes phase. This paper discusses an optimization method to derive the phase transformation kinetics parameters from dilatometry experiments. A discussion of a method based on the lattice parameters of individual phases and a method based on a lever rule for building the bridge between phase transformations and dilatometry strains is oered. The determination of kinetics parameters using an optimization algorithm was implemented into a commercial heat treatment simulation software package, DANTE. Using Pyrowear 53 steel as an example, the kinetics parameters were t for various carbon contents. The heat treatment process steps for a 3-D test bar model with a notch on the top surface were simulated, with steps including furnace heat up, carburization, air transfer from the furnace to quench tank, quenching in heated oil, cryogenic treatment, and tempering. Because of a high amount of retained austenite after oil quenching, a cryogenic treatment is used to complete the martensite transformation in the high carbon case of the test bar. After the deep freeze, the test bar was tempered. The predicted distortion and residual stresses were veried by the experimental testing. 2005 Elsevier B.V. All rights reserved.
PACS: 02.70.Dh; 05.70.a; 07.05.Tp Keywords: Finite element analysis; Heat treatment; Phase transformation; Residual stress; Distortion

1. Introduction Many critical parts, such as helicopter transmission gears are made from alloy steels that are carburized, quenched and tempered in order to achieve a combination of high strength and high toughness.

* Corresponding author. Tel.: +1 440 234 8477; fax: +1 440 234 9140. E-mail address: lynn.ferguson@deformationcontrol.com (B.L. Ferguson).

0927-0256/$ - see front matter 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.commatsci.2005.02.005

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

275

Carburization of steel parts reduces the diusive transformation rates in the carbon case during quenching. This allows more austenite to transform to lower bainite and martensite. The increase of carbon content in the case also reduces the martensitic transformation starting temperature, which leads to a delayed martensite transformation on the surface. This well-known phenomenon is the reason compressive residual stress is generated for carburized component. However, the quantitative eects of carburization on the phase transformation and internal stress evolution have been rarely reported. Quenching, one of the most important heat treatment process steps, produces high thermal gradients and rapid phase transformation rates. These transient conditions make it dicult to experimentally investigate the sequence of events during quenching, such as the evolution of phases and changes in internal stress state. For this reason, heat treatment simulation software has been developed; including HEARTS [2], TRAST [3], SYSWELD [4], DEFORM-HT [5], and DANTE [6]. Heat treatment processes usually involve phase transformations, and due to the material volume change associated with a phase transformation, a specied point of the component may go through loading, unloading, and reverse loading. The complicated internal stress evolution makes it dicult to use traditional mechanical modeling for the calculation of the material response. For this reason, the mechanical model implemented in DANTE uses an internal state variable approach that is stress based rather than strain based to capture the material response during heat treatment [7]. Accurate modeling of the phase transformations is key to predicting internal stress and distortion during heat treatment. Sources for characterizing phase transformations include isothermal transformation (TTT) diagrams, continuous cooling transformation (CCT) diagrams, Jominy hardness data, and dilatometry data. TTT diagrams are mainly used to characterize the diusive phase transformations, and CCT diagrams document both diusive and martensitic transformations. Jominy hardness test data, while plentiful and a well documented method for assessing an alloys hardenability, is dicult

to use as a basis for quantitatively characterizing transformation kinetics because a bridge between hardness and the phases formed at dierent temperatures is required and not readily available. Dilatometry data is considered an eective method to characterize both diusive and martensitic phase transformations, and ASTM Standard A1033-04 has been adopted pursuant to these methods. Dilatometry data includes length (or diameter) change and temperature as functions of time. The dimensional change is typically converted to strain. To t the kinetics parameters from dilatometry data, a bridge between strain and phase fractions is required. The lever rule and a lattice parameter based method are the two main methods to build this bridge. The steel alloy Pyrowear 53 is used in helicopter transmission gear applications for the US Army, primarily because of its resistance to tempering at high temperatures and excellent fatigue strength. In this paper, Pyrowear 53 dilatometry data is used for tting the kinetics parameters. The advantage and disadvantage of both lever rule and lattice parameter methods are discussed. Using the tted parameters, the heat treatment process of a test bar with a modied V notch was simulated. The V notched test bar in this case was designed to simulate the stress state developed in a typical gear root area due to heat treatment and subsequent bending fatigue tests. That work is reported in Ref. [8]. The eect of carburization on the phase transformation history and internal stress evolution is discussed in detail in this paper.

2. Kinetics parameters tting The chemical composition of Pyrowear 53 steel is listed in Table 1. The alloy content for this steel is specically designed to achieve resistance to softening at high temperatures and retain hot hardness in the carburized case, while maintaining high core impact strength and fracture toughness. The high alloy content in Pyrowear 53 retards the diusive phase transformations, and consequently Pyrowear 53 has excellent hardenability. For Pyrowear 53, quenching of austenite produces martensite [1].

276

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

Table 1 Chemical composition of Pyrowear 53 C 0.1 Mn 0.35 Si 1.0 Cr 1.0 Ni 2.0 Mo 3.25 Cu 2.0 V 0.1

To capture the eect of carburization on phase transformations and internal stress evolution during quenching, kinetics parameters for dierent carbon levels were required. In this study, four carbon levels of Pyrowear 53 were tested at Oak Ridge National Laboratory under Department of Energy M-PLUS program using the high speed quenching dilatometer. The four carbon levels were 0.1%, 0.3%, 0.5%, and 0.8%, respectively. In Fig. 1, the curves with hollow markers are length strains from the dilatometry experiments. For the dilatometry experiments with dierent carbon levels, the strain increased slightly right before or at the start of martensite formation. This is somewhat dierent than observations of martensitic formation for carbon and low alloy steel. The cause for this phenomenon is unknown. The DANTE software package has several utilities for tting kinetics parameters based on sensitivity analysis. The data sources that can be used for parameter determination include TTT diagrams, CCT diagrams, Jominy hardness data, and dilatometry data, which is the favored data source. Using dilatometry data to t the kinetics parameters, either a method based on unit cell lattice parameters for each phase or a simple lever rule can be used to build the relation between the

Fig. 1. Experimental and predicted pyrowear 53 dilatometry curves with dierent carbon contents.

dilatometry strain and phase fractions present as functions of time at temperature. The lever rule assumes that the material expands linearly with the fraction of austenite transformation during cooling. Instead of using phase volume fraction, the lattice parameters method uses atomic fractions of phases. The lattice parameters of dierent phases for specied steels are available in Ref. [9]. For a specied steel, the lattice parameters are functions of carbon concentration, temperature, and phase. However, the lattice parameters available in references such as [8] are limited. Also, the published lattice parameters generally have not been determined over the range of conditions appropriate for heat treatment processes. Therefore, the eects of deformation and internal stress have not considered. For these reasons and others, an extensive experimental program is required to determine accurate lattice parameters for specied steel phases under specied testing conditions. The lack of specic data and the expense of the experiments needed to generate the data has limited the use of a tting method reliant on lattice parameters for tting phase transformation kinetics parameters. An additional consideration is the nature of the steel material used to generate the lattice parameter data and that of a production component. A lattice parameter study seeks to characterize an ideal material, and a steel component does not have an ideal microstructure. For example, the existence of grain boundaries inuences the coecient of thermal expansion of the steel so that a strain vs. temperature curve from a dilatometry test does not match the lattice parameter change with temperature. Therefore, the lattice parameters have to be adjusted to consider the polycrystalline nature of actual parts. By adjusting the lattice parameters for tting the kinetics using dilatometry experiments, the lattice parameters method and the lever rule essentially become the same. By applying lever rule, the Pyrowear 53 dilatometry experiments for four carbon contents were t individually using the DANTE utility. The objective of the tting routine is to minimize the area dierence between the experimental and predicted curves. The experimental and the predicted

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

277

results for martensite formation are shown in Fig. 1. The martensitic transformation starting temperature (Ms) drops signicantly with the increase of carbon content, as expected, with the Ms of 0.1% carbon being about 415 C and the Ms of 0.8% carbon being about 85 C. The significant drop in Ms with carbon is the benecial source of high surface compressive residual stress in carburized parts. By cooling the Pyrowear 53 to room temperature, there is no predicted retained austenite for either 0.1% carbon or 0.3% carbon levels. For a carbon level of 0.8%, the retained austenite is predicted to be almost 40%.

3. Case study Rectangular bar samples with a V notch were used to investigate the eect of carburization on the residual stress and distortion. The dimension of the test bar was 55 10 10 mm. The heat treatment process included furnace heat up, carburization, oil quench to 65 C, air cool to room temperature, deep freeze to liquid nitrogen temperature (196 C), and tempering at 230 C. The notched test bar was carburized at 900 C for 4 h, with a carbon potential of 0.8% applied only to the top surface. The other surfaces were not carburized. Because of the high thermal gradients in the surface of the part during quenching, as well as the carbon gradient in the top surface of the

test bar, ne surface elements were used in the nite element model, with four layers of ne surface elements being applied. The average thickness of the surface element was about 0.3 mm, as shown in Fig. 2. There were totally 9045 nodes and 7672 hexahedral elements in the nite element model. The contour of carbon distribution after carburization predicted by nite element simulation is shown in Fig. 3. The carbon content at the top surface is 0.8%, and the depth of the carburized case is about 0.5 mm. The testing bar is symmetric with both plane ADE and plane BAC, as shown in Fig. 3. The point A is located at the center of the top surface, and the line AB is normal to the top surface. The nite element nodes located on line AB are used to investigate the simulation results in detail. From the notch surface, the depths of the nodes on line AB are 0.0, 0.2, 0.4, 0.7, 1.0, 1.0, 2.5 mm, respectively. The carbon distribution predicted by simulation along line AB is shown in Fig. 4. As shown in Fig. 1, the carbon content had a signicant eect on the martensitic transformation starting temperature and the volume fraction of retained austenite. After cooling the test bar to room temperature, the simulation predicted about 37% retained austenite at the surface, with no retained austenite at the depths over 0.7 mm. To obtain maximum material strength at the surface of the test bar, minimum retained austenite is required. By cooling the test bar

Fig. 2. Test bar dimension and nite element mesh.

278

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

Fig. 3. Predicted carbon distribution contour after carburization.

Fig. 4. Carbon distribution and retained austenite distributions.

Fig. 5. Martensite formation histories at dierent depths during oil quenching.

in liquid nitrogen, the retained austenite in the top surface was transformed to martensite. After deep freezing, the predicted retained austenite at the surface was reduced to 7%, as shown by the line with hollow marks in Fig. 4. In Fig. 5, the curves without markers are for temperatures at dierent depths during oil quenching. The predicted temperature curves with respect to quench time show that the temperature dierence between the surface and the core is less than 30 C. The curves of martensite volume fraction in Fig. 5 show that the predicted martensitic phase transformation starts rst at sub-surface locations and then progresses to the surface location, and this dierence in martensite formation is due to

the carbon gradient. The dilatometry curves in Fig. 1 indicate a signicant Ms temperature reduction with carbon. Therefore, locations beneath the surface reached the Ms temperature before surface locations, even though the temperature at the surface was 30 C lower. In Fig. 5, at a depth of 0.4 mm, the martensitic transformation started after about 10 s, and only 40% of martensite was obtained at 20 s. At the depth of 0.7 mm, the transformation to martensite started at about 7 s, and 95% of martensite was predicted after 20 s. In terms of quench time, the martensite formation curves at depths of 1.5 mm and 2.7 mm were almost the same. From Fig. 4, the carbon content at depths of 1.5 mm and 2.7 mm is the base carbon. The small dierence in phase transformation

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

279

timing at these two points is because of the small thermal gradient between them. Because of high volume fraction of retained austenite in the high carbon top surface, the test bar was held in liquid nitrogen to continue martensitic transformation. After deep freezing, a tempering process at 230 C was applied to transform the as-quenched martensite to tempered martensite. Fig. 6 shows the volume fraction change of tempered martensite with tempering time. Fully tempered martensite was obtained after 4000 s. The eect of carbon on the tempering process was neglected during the tempering simulation, meaning that the tempering kinetics were assumed to be independent of carbon level; this assumption is most likely not accurate, but it was used due to lack of experimental data for this steel. After tempering, the test bar was cooled to room temperature. Distortion and residual stress were measured and compared with the simulation results. During quenching, the thermal gradients and phase transformations both contribute to the stress evolution. Fig. 7 shows the internal stress changes for the length direction with quenching time. As in Fig. 4, the X coordinate is the depth of the seven selected nodal locations. At the start of quenching, the internal stress is low, but it is not zero. The stress before quenching originates from the 12 s air transfer from the furnace to quench tank, during which time a small temperature gradient is developed in the test bar. The simulation predicts that there is no martensitic transformation before 4.0 s of quenching. At

Fig. 7. Evolution of lengthwise stress during quenching.

Fig. 6. Predicted volume fraction of tempered martensite with tempering time.

2.4 s, tensile stress is generated at the surface because of the high surface cooling rate. At 3.1 s, the core temperature is high and the core cooling rate exceeds the cooling rate of the colder surface, which causes more thermal contraction in the core. As shown in Fig. 7, as the time advances from 2.4 s to 3.1 s and then to 3.8 s, the core stress becomes tensile and the surface stress moves from tension to compression. At 3.8 s, the tensile stress in the core and the compressive stress at the surface are strictly due to changes in temperature as no phase transformation has begun. Between 3.8 s and 8.0 s of quenching, the simulation predicts a large change in the internal stress distribution, and this large change is mainly caused by martensite formation. At 8.0 s, the martensitic transformation at the core is almost complete, while the transformation at the surface is not yet started, as shown in Fig. 5. The material volume expansion caused by martensitic formation has generated a compressive stress of approximately 300 MPa in the core while the austenitic surface must endure a tensile stress of over 200 MPa. At 35.7 s, formation of martensite in the carburized case has started. As shown in Fig. 7 for this time, the peak compressive stress is predicted to occur at a depth of 0.4 mm, the compressive stress in the core and the tension in the case being reduced. With continued martensite formation in the carburized case, high compressive stresses in the surface are developed. The retarded phase transformation in the top surface of the test bar is the main reason of obtaining compressive stress in the surface.

280

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

Predicted and measured lengthwise residual stress distributions in the carburized case are shown in Fig. 8. The solid line with circular markers is the predicted residual stress after oil quenching and air cooling to room temperature. The predicted surface compressive stress is about 380 MPa, and the depth of the compressive stress is about 1.0 mm. The as-quenched test bar is predicted to have about 40% retained austenite at the top surface, as shown in Fig. 4. After deep freezing and tempering, the predicted residual stress is shown as the solid line with triangular markers in Fig. 8. The reduction in retained austenite in the high carbon surface layer and the associated volume expansion with additional martensite formation generates additional compressive stress in the top surface, with the magnitude of surface compressive stress increasing by 50 MPa to 420 MPa. The dotted line is the residual stress calculated from X-ray diraction measurements for the specimen after deep freeze and tempering. The measured residual stress is slightly less compressive than the predicted values. The overall dierence between lengthwise stress calculated from X-ray measurements and the predicted stress less than 15%. The predicted stress was found to not be balanced biaxial stress as was assumed in the case of measured stress calculations. The stress state at the root of the notch was predicted to be 380 MPa normal to the notch direction and 310 MPa parallel to the notch direction. This may account for the dierence between predicted stress and stress calculated from X-ray diraction measurements.

Fig. 9. Displacements at dierent stages of the heat treatment process.

Distortion at dierent stages of heat treatment is shown in Fig. 9. The X coordinate is the length direction of the test bar, with the center of the notch being zero. The Y coordinate is the normal displacement of the top edge from the initial planar position; this essentially shows bow along the length of the test bar. Fig. 9 shows that the test bar top surface expanded after heating and remained relatively at. The predicted overall expansion was about 0.05 mm along the top edge. During heating, the slight distortion was caused mainly by austenite formation. The distortion at the notch location is more severe than other locations because of the section geometry change. The displacement curve for after quenching, the line with square marks in Fig. 9, shows that signicant bowing has occurred, with the notch remaining higher than the ends of the bar. Further shape changes caused by deep freezing and tempering are minor. The dotted line with circular markers is the measured displacement for one half of the test bar. The comparison shows a good match between the nal predicted and measured shapes. 4. Summary The purpose of this paper was to demonstrate the signicant eect of phase transformations on both residual stress and distortion in a carburized and hardened part. The part in this case was a notched test bar, but it adequately documented the importance of the timing of phase transformations on the magnitude and nature of residual stress and the associated distortion.

Fig. 8. Comparison between predicted and measured residual stress.

B.L. Ferguson et al. / Computational Materials Science 34 (2005) 274281

281

Dilatometry tests for four carbon levels of Pyrowear 53 steel were conducted and used to characterize the phase transformation behavior of this steel as a function of carbon level. The link between dilation and fractions of phases was based on the lever rule, and it was shown to be suciently accurate for heat treat process simulation. The method used to determine phase transformation kinetics parameters relied on minimization of error between measured and predicted length strains using an optimization method to drive the computer simulation trials. Finally, the carburization and hardening process simulation using a commercial nite element based package, DANTE, showed that agreement could be achieved between predicted and measured values of distortion and residual stress. This is signicant because distortion has been identied as a major source of cost for hardened steel components, and fatigue life is known to be improved by the presence of compressive residual stress in part surfaces.

[3]

[4] [5]

[6]

[7]

[8]

References
[1] Product Technical Data for Pyrowear 53 Steel, Carpenter Technology Corporation, 2001. [2] T. Inoue, D.Y. Ju, K. Arimoto, Metallo-thermo-mechanical simulation of quenching processTheory and implementa[9]

tion of computer code HEARTS, in: G.E. Totten (Ed.), 1st International Conference on Quenching and Distortion Control, ASM International, 1992, pp. 205212. N. Jarvstrat, S. Sjostrom, Current status of TRAST: A material model subroutine system for the calculation of quench stresses in steel, in: ABAQUS Users Conference Proceeding, 1993, pp. 273287. SYSWELDA Predictive Model for Heat Treat Distortion, Southwest Research Institute, 1992. K. Arimoto et al., Development of heat treatment simulation system DEFORMTM-HT, in: R.A. Wallis, H. Walton (Eds.), Proceedings of the 18th Conference on Heat Treating, ASM International, Materials Park, OH, 1998, pp. 639 654. B. Lynn Ferguson, A. Freborg, G. Petrus, Software simulates quenching, Advanced Materials and Processes (August) (2000) H31H36. D. Bammann, V. Prantil, A. Kumar, J. Lathrop, D. Mosher, M. Callabresi, H.J. Jo, M. Lusk, G. Krauss, B. Elliott, G. Ludtka, T. Lowe, B. Dowling, D. Shick, D. Nikkel, Development of a carburizing and quenching simulation tool: A material model for carburizing steels undergoing phase transformations, in: G.E. Totten et al. (Eds.), 2nd International Conference on Quenching and the Control of Distortion, ASM International, Materials Park, OH, 1996, pp. 367375. B.L. Ferguson, A.M. Freborg, High strength, aordable helicopter gears, REDCOM TR 04-D-29, Final Report on SBIR Contract W911W6-040C-0022, US Army Research, Development and Engineering Command, 30 June 2004. M. Onink, C.M. Brakman, F.D. Tichelarr et al., The lattice parameters of austenite and ferrite in FeC alloys as functions of carbon concentration and temperature, Scripta Metallurgica 29 (1993) 10111016.

Das könnte Ihnen auch gefallen