Camurri (2008) PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456

journal homepage: www.elsevier.com/locate/jmatprotec

Residual stress during heat treatment of steel grinding balls

C. Camurri a , C. Carrasco a,∗ , J. Dille b


a Universidad de Concepción, Department of Materials Engineering, Casilla 53-C, Concepción, Chile
b Université Libre de Bruxelles, Faculté des Sciences appliquées/école polytechnique, Avenue F. Roosvelt 50CP194/03, Brussels, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: This work aimed to model the temperature distribution, phase transformation and resid-
Received 24 April 2007 ual stress induced during the heat treatment of grinding balls of 3 and 5 in. diameter. The
Received in revised form temperature model considered factors such as the heating of the water and the formation
25 October 2007 of a steam layer that surrounds the balls at the start of the quenching. The model of the
Accepted 8 January 2008 residual stress field considered: the temperature distribution, the force equilibrium equa-
tions and the constitutive thermo-elastic relationships, including the expansion due to the
austenite–martensite transformation.
Keywords: A good agreement between the experimental and theoretical values for the temperature
Grinding balls distribution was obtained with the differences at the end of the quenching being no higher
Heat treatment than 0.5%. The experimental behavior of the balls in a mill simulator, as well as the residual
Martensite stress measured by X-ray diffraction, agreed satisfactorily with the theoretical predictions.
Residual stress © 2008 Elsevier B.V. All rights reserved.

1. Introduction During the heat treatment, and as a result of the ther-


mal and structural gradients, residual stresses are generated
Moly-Cop at Talcahuano, Chile, produces steel grinding balls which can jeopardize the working life of the balls and, in
mainly of 3 and 5 in. (7.62 and 14.70 cm, respectively) diameter, some cases, cause fracture before they are used. The aim of
using rolling or forging processes (see Table 1 for details of this work was to model and resolves in an analytical way
their composition). the residual stress field induced during the quenching and
Immediately after the metal working the balls, at 800 ◦ C, equalization of the steel balls and thus determine their possi-
were quenched in water at an entry temperature of 50 ◦ C and ble cracking or fracture points. Also, and for further works,
for 80 and 180 s for the 3 and 5 in. diameter 2R of the balls, the model can be useful for predict the optimal operating
respectively. After quenching, the balls were cooled in air for conditions during quenching, for instance, the initial temper-
temperature homogenization for 40 and 120 s, respectively, a ature and flow of the water, in order to reduce the residual
process known as equalization. The 3 in. diameter balls were stresses.
then stored in boxes in order to allow them to slowly reach The work was carried out in two main stages: first, the
ambient temperature. When the 5 in. diameter balls had been radial distribution of the temperature of the balls during the
equalized a four-step annealing process, each step lasting for quenching and equalization processes was modeled using
1 h, was performed: cooling in the furnace by air pressure, the explicit finite difference method and validated experi-
then heating the balls to 230 ◦ C by hot air, then slow cooling mentally. Also, the radial distribution of the martensite was
down to 220 ◦ C, and finally cooling from 220 ◦ C to the ambient calculated. Second, the residual stress field was simulated
temperature. using the data obtained in the previous stage.


Corresponding author. Tel.: +56 41 2207170; fax: +56 41 2203391.
E-mail address: ccarrascoc@udec.cl (C. Carrasco).
0924-0136/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2008.01.007
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456 451

Nomenclature

Cp caloric capacity
E elasticity modulus of the mix
austenite–martensite
h heat transfer coefficient ball ambient
hconvective convective heat transfer
hradiant radiant heat transfer
K thermal conductivity
Ms temperature of beginning of martensitic trans-
formation
R radius of the balls
r radial coordinate
ra radius of a sphere of 1 g of austenite used as
reference for calculus
rma radius of a sphere of 1 g of the mix Fig. 1 – Micrography of the center zone of a 5 in. diameter
austenite–martensite ball.
T radial temperature of the balls.
Ts surface temperature of the balls
Tam ambient (steam, water, air) temperature • The heat losses by radiation and the heats of reaction of the
u radial displacement phases formed during the process are neglected.
• The thermal conductivity K and caloric capacity of the
Greek letters steel Cp depends linearly on the temperature. Addition-
˛ thermal lineal expansion coefficient ally, the densities of the only two possible phases presents
ˇ lineal expansion coefficient due to phase trans- (see Fig. 1), austenite and or martensite, were assumed as
formation 8.03 g/cm3 and 7.75 gr/cm3 , respectively (Totten et al., 1992).
εrr , ε␪␪ radial and circumferential strain, respectively • The heat losses from the water used for quenching are neg-
 Poisson modulus ligible.
i density of “i” phase (i = a, austenite; i = m,
martensite) With this assumption, the heat equation to resolve is
 rr ,  ␪␪ ,  ef ,  0 ,  R radial, circumferential, effective or
 
equivalent, yield and rupture stress, respec- ∂2 T 2 ∂T ∂
tively K + =  (TCp ) (1)
∂R2 r ∂R ∂t
M volumetric fraction of martensite (or annealed
martensite) With the following border conditions
∂T ∂T
(0, t) = 0 and −K (R, t) = h(Ts − Tam )
∂r ∂r
where h is the coefficient of heat transfer between the surface
2. Models
of the ball at T = Ts and the media at T = Tam .
The thermal and residual stress models are as follows. In a previous work (Camurri et al., 2003; Garcı́a, 2000), the
temperature of the quenching water was assumed to be con-
stant and the vapor layer formed around the balls at the
2.1. Temperature distribution of the balls
beginning of the process was neglected. These hypotheses
yielded a poor correlation between theory and experience, so
For modeling the radial temperature of the balls T, the follow-
both effects were included in the present model.
ing assumptions were made (Camurri et al., 1997; Kreith, 1973;
Hence, in a first cooling stage during the quenching, until
Incropera and Dewit, 1999; Mills, 1995; Welty, 1979; Farlow,
the surface temperature Ts of the balls is around 220 ◦ C, a
1982):
radiation and convection heat transfer mechanism across the
vapor layer surrounding the balls is considered. In this case,
• The balls are completely spherical and homogeneous, i.e., Tam = 100 ◦ C and h = hconvective + 0.75 hradiant . The formula for
they only have radial temperature gradients. determines the coefficient hconvective as a function of the tem-

Table 1 – Composition (wt.%) of the steel of 3 and 5 in. diameter balls


Diameter C Mn P S Si Cr Mo Al Ti Nb V Cu Ni

3 1.13 0.93 0.01 0.01 0.23 0.76 0.05 0.02 <0.01 <0.01 <0.01 0.2 0.05
5 0.74 0.94 0.01 0.02 0.21 0.80 0.01 <0.01 0.02 0.04 0.01 0.06 0.05
452 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456

perature can be found in Incropera and Dewit (1999). This In general, the modeling of coupled thermo mechanical
formula involves the ball diameter and physics and thermal problems with also phase’s transformations normally use
properties of the liquid and vapor water such as density, ther- numerical techniques such as finite elements for its resolu-
mal conductivity and cinematic viscosity; also the formula tion. For example, in Hossain et al. (2004), this methodology
has a constant which depend of the geometry of the bodies, is addressed in the modeling with a posterior experimental
in this case 0.67 for spheres. For the radiation coefficient an validation by means of neutrons diffraction, of the resid-
emissivity of the steel of 0.9 was supposed. ual stresses induced during the quenching of stainless steel
In the second cooling stage during the quenching, until spheres of 30 mm diameter. In the present work, a simple ana-
Ts ∼ 105 ◦ C, the cooling of the balls is produced by vapor lytical solution for the residual stresses field in grinding balls
transport from its surface, and the temperature is considered is presented based on the following equations.
as Tam = 100◦ C. In this stage the coefficient h can be found
in Incropera and Dewit (1999) and depends of the thermo- 2.2.1. Equation of Equilibrium of radial forces
physics properties of saturated liquid and/or vapor such as ∂rr (rr − ␪␪ )
+2 =0 (2)
density, specific heat, cinematic viscosity and superficial ten- ∂r r
sion. There are also two constant, C and n, in the formula
where  rr and  ␪␪ are the normal, radial and circumferential
for obtain h. These constants are related with the system
stresses, respectively.
surface-fluid, and the values for the studied steel–water are
0.013 for C and 1.0 for n (Incropera and Dewit, 1999). Finally,
2.2.2. Equations relating deformations and displacements
the third cooling stage begins when the surfaces of the balls ∂u u
reach the boiling temperature of the water, and the heat trans- εrr = and ε␪␪ = (3)
∂r r
fer mechanism is pure convection. In this case Tam is equal to
the instantaneous temperature of the water, which is calcu- where u, εrr and ε␪␪ are the radial displacements, and the radial
lated by main of a heat balance among the heat loosed for and circumferential strain, respectively.
the ball and transfer to the water, with initial temperature of
50 ◦ C. 2.2.3. Constitutive thermo-elastic-phase transformation
During the equalization and annealing, the heat losses equations
from the surface of the balls to the air are only by convec- These include the thermal expansion coefficient of the
tion. In all cases, h is function of the surface temperature of steel and the linear coefficient of expansion due to the
the balls. austenite–martensite phase transformation.
Eq. (1) was solved by the explicit finite difference method,
using algorithms for the internal, central and superficial (rr − 2␪␪ ) (1 − ) 
εrr = + F(r) and ε␪␪ = ␪␪ − rr + F(r)
nodes. Performing with the stability conditions, in the case E E E
of 3 in. diameter balls, the radial increments were R/20 and (4)
the temporal increments were 0.02 s and 0.05 s in quenching
and equalization, respectively. For 5 in. diameter balls, those and
increments were R/40, 0.05 s in quenching and 0.025 in both
F(r) = ˛ T(r) + ˇ M(r)
equalization and annealing. The temperature distribution was
validated by inserting thermocouples in three different posi- where E and  are the elasticity and Poisson modulus,
tions in the balls: the center, R/2 and R—0.7 mm, where R is respectively of the austenite–martensite mix at temperatures
the radius of the balls. Additionally, the temperature of the below Ms , obtained from reference data for their individual
quenching water was continuously measured during the pro- modulus (Okamura and Kawashima, 1988) and, in the case of
cess in order to compare it with the model results. the Young’s modulus E, were also obtained experimentally in
this research from tensile curves of standard samples of 5 mm
in diameter and 25 mm gage length L0 . These samples were
2.2. Residual stress of the balls
quenched at temperatures below Ms and subjected to tensile
tests by an Instron machine at those temperatures. The slope
To model the residual stress field, the following hypotheses
of the linear part of the stress–strain curve corresponded to
were made:
the E values (Rodrı́guez, 2006).
The symbol ˛ corresponds to the thermal expansion coef-
ficient of this mix, ˇ corresponds to the linear expansion
• Throughout the cooling there are only two phases in the coefficient due to the phase transformation from austenite to
balls: austenite and/or martensite, as Fig. 1 shows. martensite during quenching and equalization or the linear
• The induced residual stresses are only normal stresses. contraction coefficient due to the phase transformation from
• The residual stresses induced in the austenite are negligible martensite to annealed martensite during the annealing of
due to the low yield stress of this phase at a temperature 5 in. diameter balls. Finally, T(r) and M(r) are the changes
above Ms , the temperature at which start the martensitic in the temperature and the volumetric fraction of marten-
transformation. For example, at 400 and 600 ◦ C the yield site formed at each radial position for temperatures below
stress of austenite is 120 and 80 MPa, respectively, while the from Ms , respectively. Determined the temperature of the balls
yield stress of martensite at 200 ◦ C is 800 MPa (Okamura and for each time and knowing the CCT curves of the steels for
Kawashima, 1988). obtain Ms , it is possible determines the volumetric fraction
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456 453

of martensite formed (M(r)) using the Koistinen–Marrburger


equation (Krauss, 1990). Obviously, M = 0 if the temperature
in some radial portion is higher than Ms . As the densities of
the austenite (a ) and the martensite (m ) are known as well
as the fraction of formed martensite M(r), the coefficient ␤(r)
is determined as

rma − ra
ˇ= (5)
ra

where rma and ra are the radius of the 1 g of spherical shape


(base of the calculus) of the mix austenite–martensite or only
austenita, respectively, and which are calculated as
 
3 3
ra = and rma = (6)
4a 4(1 − M)a + Mm ) Fig. 2 – Temperature distribution during quenching of 3 in.
diameter balls.
If ˇ must be calculated in the transformation martensite
to annealing martensite, the procedure is analogous, and in
this case M represents the fraction of annealing martensite
given elsewhere (Camurri et al., 2003; Garcı́a, 2000; Rodrı́guez,
and should be determined from an annealing parameter as is
2007).
shown in Shi et al. (2001).
By combining formulas (2)–(4), the radial equation of the
force equilibrium can be expressed in terms of the radial dis- 3. Results and discussion
placement u as

1 ∂  Figs. 2 and 3 show the experimental and theoretical results for


∂ 1 +  ∂F(r) the radial distribution of temperature for 3 in. diameter balls
(r2 u) = (7)
∂r r2 ∂r 1 −  ∂r during quenching and equalization, respectively.
Good agreement between the present model and the exper-
Note that F(r) is known because the radial distribution of tem-
imental data is observed. In the other hand, a poor correlation
perature T and the coefficient ␤(r) have been determined as
was obtained in the stage of quenching if the thermal model
was described previously.
not included the effect of the initial formation of vapor layer
Then by a double integration the radial displacement u is
surrounding the balls (Camurri et al., 2003; Garcı́a, 2000) which
obtained
cause a higher cooling rate in the surface of grinding balls.
 Fig. 4 compares the experimental and theoretical values
(1 + ) r2 F(r)dr r C2
u(r) = + C1 + 2 (8) for the temperature of the quenching water during the same
r2 (1 − ) 3 r
process.
Using the constitutive thermo-elastic-phase transforma- There was good agreement between the experimental
tion Eq. (4), finding the stresses in terms of the strains water temperatures during quenching. They were slightly
expressed as a function of u and ∂u/∂r, the radial normal stress lower than the theoretical temperatures at the end of the pro-
is cess because the heat losses to the ambient were neglected.

E
 ∂u u

rr = (1 − ) + 2 − (1 + )F(r) (9)
(1 + )(1 − ) ∂r r

where C1 and C2 are constants which must be determined


from the following border conditions
If r = R (radius of the ball),  rr = 0.
If T(r = 0) < Ms , then ∂ rr /∂r = 0.
If T(r = 0) ≥ Ms , then  rr (r ≤ r*) = 0, where r* is the radius for
which T(r = r*) = Ms .
Additionally, if for some radius r** the effective or equiva-
lent stress  ef defined in this case as  ef = | rr −  ␪␪ |, with  ␪␪
the circumferential stress, results equal or greater than the
yield stress of the steel  0 , then  ef(r=r**) =  0 , if  ef <  R (rup-
ture stress), or  ef(r=r**) = 0 if  ef ≥ ␴R . Due to the low ductility
of the steel of the balls, their rupture stress is closely equal
to their tensile strength. Once the radial distribution of the
radial residual stress has been obtained, the radial distribution
of circumferential stress is obtained directly from the equi- Fig. 3 – Temperature distribution during equalization of
librium Eq. (2). More details of the mathematical models are 3 in. diameter balls.
454 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456

Fig. 4 – Experimental–theoretical comparison of quenching


water temperatures.

Fig. 5 shows the numerical radial distribution of martensite


in 3 in. diameter balls at the end of quenching and equaliza-
tion.
Fig. 5(a) indicates that at the end of quenching and for a
radius smaller than 1.9 cm, the temperature of the balls was
higher than Ms (Ms = 220 ◦ C for the steel used), and conse-
quently no martensite was observed in that zone. This implies,
according to the previous discussion, that the residual stresses
at that radius are negligible.
Figs. 3 show that at the end of the equalization the centers
of the balls have a temperature closely of 150 ◦ C, which implies
that at that moment there is 54% of martensite when r = 0. As
a result, when the equalization concludes, all the balls have Fig. 5 – Numerical martensite distribution in balls of 3 in.
residual stress. diameter: (a) at the end of the quenching and (b) at the end
Fig. 6 shows the modulus of elasticity as a function of the of the equalization.
temperature of the austenite–martensite mix obtained from
the reference data (Okamura and Kawashima, 1988) and exper-
imentally in this investigation (Rodrı́guez, 2006).
Fig. 6 shows a very good agreement between both data
groups up to temperatures of around 170 ◦ C. For higher
temperatures, the experimental data obtained from this inves-
tigation are lower than the data from the literature for
martensite alone. This is probably due to the expansion effect
of the pull roads used to subject the sample during the test,
which implies a longer displacement for each tensile strength.
Since a temperature extensometer was not available, the elon-
gation of the sample was obtained from the displacement data
recorded by the Instron machine minus the effect of the elon-
gation of the equipment, including the pull roads, which was
determined from previous traction tests at room temperature
made on steel samples of known modulus of elasticity.
Figs. 7 and 8 show the theoretical radial distribution of
circumferential stress for balls of 3 and 5 in. diameter, respec-
tively, at the end of the quenching and equalization. Some
experimental measurements of residual stresses on the sur-
face of industrial 3 and 5 in. balls, i.e., after the slow cooling
in boxes after the equalization (3 in. diameter balls) or after Fig. 6 – Variation in modulus of elasticity with temperature
annealing (5 in. diameter balls) are included. These values of the austenite–martensite mix. Comparison between
were obtained by X-ray diffraction (XRD) with a diffrac- reference and experimental data.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456 455

Fig. 9 – In time distribution of circumferential stress for 3 in.


diameter balls at 0.07 mm below the surface at the end of
Fig. 7 – Theoretical radial distribution of circumferential quenching (dotted line) and equalization (continuous line).
stress for 3 in. diameter balls at the end of quenching
(dotted line) and equalization (continuous line) and
experimental (spots in black). homogenizer the temperature of the balls. It is also interest-
ing to note that the part of the ball with traction stress at the
end of the equalization coincide with the volume with mix
tometer (Siemens D 500). Also, as a reference, in depth austenite–martensite during the quenching. The circumferen-
experimental values of the residual stresses were obtained tial stresses in this volume at the end of the equalization are
by measures on 1.0 cm × 1.0 cm × 6.5 cm parallelepipeds cut positive because the transformed martensite during this treat-
from the 3 in. diameter balls, or from 1.0 cm × 1.0 cm × 12.0 cm ment has less rigidity than the martensite formed during the
parallelepipeds from the 5 in. diameter balls. quenching, and for this reason the central zones of the balls
Note from Figs. 7 and 8 that after the quenching, in zones have more trends to expand circumferentially. Additionally,
where there are only austenite, the residual stresses are cero, Figs. 7 and 8 show the good agreement between the experi-
while after the equalization, the totality of the ball as the mix mental and theoretical results at the end of equalization of
austenite–martensite and, as a consequence, all the mate- the circumferential residual stresses on the surface. Also, that
rial has residual stresses. In the other hand, it is observed those greater tensile values occur at a certain depth below the
that the absolute values of the residual stresses induced dur- surface of the balls, which coincides with the position of their
ing the quenching, are greater than the induced during the fracture when occurs during the heat treatment. The values
equalization, just because that treatment has the function of of the residual stresses inside the balls show an acceptable
agreement between the theoretical (end of equalization) and
experimental values, the latter being only a reference due to
the effect of cutting the samples from the balls for the XRD
measurements, which introduces distortions and also stress
release.
Fig. 9 shows the time evolution of the circumferential
stress of 3 in. diameter ball at 0.07 mm below the surface after
quenching and equalization. From that figure can be seen the
variation from compressive stresses at the end of the quench-
ing to traction stresses at the end of the equalization.

4. Conclusions

This simple model of the heat transfer gives the temperature


distribution of the balls during quenching and equalization,
with good experimental agreement.
The model of the residual stresses duly predicts the
observed experimental fact in a mill simulator that, when the
Fig. 8 – Theoretical radial distribution of circumferential balls fracture during the equalization, the fracture begins at a
stress for 5 in. diameter balls at the end of quenching certain depth from the surface.
(dotted line) and equalization (continuous line) and The modeled and experimental residual stresses in the
experimental (spots in black). balls compare well. It should be noted that the experimen-
456 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 450–456

tal values measured inside the balls are only intended as a Farlow, S., 1982. Partial Differential Equations for Scientists and
reference, due to the effects of distortion and stress release Engineers. Elsevier.
when cutting the samples for the XRD measurements. Garcı́a, A., 2000. Esfuerzos residuales durante el temple de
medios de molienda. Memoria de Tı́tulo, Facultad de
Future improvement to the model can be making, such to
Ingenierı́a, Universidad de Concepción.
include the release of residual stresses in the balls due to plas- Hossain, S., Daymond, M.R., Truman, C.E., Smith, D.J., 2004.
tic micro-deformations during their heat treatment and the Prediction and measurements of residual stresses in
local deformation hardening associated to this phenomena. quenched stainless-steel spheres. Mater. Sci. Eng. A, 339–349.
Finally, to use the model as a predictor of the better oper- Incropera, P., Dewit, D., 1999. Fundamentos de Transferencia de
ational conditions during the heat treatment, such as initial Calor. Prentice Hall.
temperature and flow and agitation of water for reduce resid- Krauss, G., 1990. Steels Heat Treatment and Processing Principles.
ASM Publications.
ual stresses of the balls, mayor accuracy in the in depth values
Kreith, F., 1973. Principios de Transferencia de Calor. Herreros
of the residual stresses must be obtained. For this, techniques Hnos.
such as hole drilling or neutrons diffraction must be used. Mills, A., 1995. Transferencia de Calor. Irwin.
Okamura, K., Kawashima, H., 1988. Proceedings of the 32nd Japan
Congress Materials Research, p. 323–329.
Acknowledgements Rodrı́guez, P., 2006. Esfuerzos residuales en medios de molienda:
determinación experimental del módulo de elasticidad de la
This work has been supported by The National Council mezcla austenita–martensita entre Ms y la temperatura
of Research in Science and Technology of Chile, CONI- ambiente. Memoria de Tı́tulo, Facultad de Ingenierı́a,
CYT (FONDECYT project no. 1050078). The authors gratefully Universidad de Concepción.
Rodrı́guez, D., 2007. Modelación matemática del campo de
acknowledge this support.
esfuerzos residuales inducido durante el tratamiento térmico
de bolas de molienda, M.Sc. Tesis, Universidad de
Concepción.
references
Shi, W., Liu, C., Liu, Z., 2001. Numerical simulation of internal
stresses and microstructure distribution of heavy forgings
during quenching and tempering processes. In: The First
Camurri, C., Carrasco, C., Cañete, P., 1997. Etude de la Sino-Korean Conference on Advanced Manufacturing
modification du procedé de fabrication des systemes de Technology, China.
broyage. Revue de Metallurgie 3, 949–954. Totten, G., Bates, C., Clinton, N., 1992. Handbook of Quenchants
Camurri, C., Garcı́a, A., Cañete, P., 2003. Mathematical model for and Quenching Technology. ASM International, Hardbound.
the residual stresses induced during the heat treatment of Welty, J., 1979. Transferencia de Calor Aplicada a la Ingenierı́a.
grinding balls. Mater. Sci. Forum 426–432, 3933–3938. Limusa.

You might also like