XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere

AITEM 2019
Padova, 09-11 Settembre 2019

Analytical modelling of residual stresses in

sheet metal tension-levelling
T. Magroa,*, E. Simonettoa, A. Ghiottia, S. Bruschia, A. Ferraiuolob
a
Department of Industrial Engineering, University of Padova, Via Venezia 1, 35131, Padova, Italy
b
Marcegaglia s.p.a., Via Baiona 141, 48123, Ravenna, Italy

Abstract

Tension-levelling is used to minimize flatness imperfections and residual stresses of metal strips by means of plastic
deformation. During the process, the material is subjected to cyclic tension-compression stresses, obtained with
repeated and alternate bends under superposed tension. Consequently, the desired material properties and product
standards are achieved. This paper presents an analytical model for predicting the residual stresses during the tension-
levelling. The backward and the forward loads, given by the brindles, were taken into account and superposed to the
bending moment given by the rollers. The results in terms of residual stresses and neutral layer position were
evaluated.

Keywords: sheet metal forming, rolling, levelling

1. Introduction

The increasing competition in the steel market concerning the quality and steel prices, as well as the need of
compliance with the international environmental regulations regarding greenhouse gas emissions have pushed, and/or
even forced, steel industry towards significant process improvements in terms of efficiency, online monitoring and
energy consumption. To achieve these goals, many efforts are being spent by the main market players to integrate and
develop advanced manufacturing technologies, often aligned with the Industry 4.0 technological pillars. At the same
time, the growing demand from OEMs of improved uniformity of the strip tensile properties pushes metallurgists
towards the real-time knowledge of the main physical characteristics of the raw material (namely the hot rolled coil)
and the evolution of the strip mechanical properties during the annealing process. Nowadays, it is well recognized that
uniform microstructural properties under tensile stress conditions can be obtained only through a tight control of the
annealing furnace temperature, which depends on the material emissivity and on the strip surface conditions (such as
roughness, cleanliness). Other shape defects in strip products may be caused by various factors, such as: differential
reductions applied across the material width at rolling mills, non-uniform distribution of the internal residual stresses

* Corresponding author. Tel.: +39 049 8276819

E-mail address: [email protected]
 2 T.Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere

Nomenclature

ri nominal radius of i-th work roller of rolling mill [mm]

N number of work rollers [-]
xi absolute coordinate in longitudinal direction of the centre of the i-th work roller [mm]
yi absolute coordinate in transversal direction of the centre of the i-th work roller [mm]
E material Young modulus [MPa]
Y material yield stress [MPa]
K material strength coefficient [MPa]
n material strain hardening exponent [-]
t strip thickness [mm]
w strip width [mm]
α contact angle between the work roller and the strip [deg]
y distance of the m-th fibre of material from the cross-section axis of symmetry of the strip [mm]
ei neutral layer distance from the cross-section axis of symmetry of the strip after the i-th work roller [mm]
m,i longitudinal strain of the m-th fibre of the strip after the i-th work roller [-]
m,i-1 longitudinal strain of the m-th fibre of the strip after the (i-1)-th work roller [-]
m,i longitudinal stress of the m-th fibre of the strip after the i-th work roller [MPa]
m,i-1 longitudinal stress of the m-th fibre of the strip after the (i-1)-th work roller [MPa]
el,m,i elastic deformation of the m-th fibre after i-th work roller of rolling mill [-]
el,m,i-1 elastic deformation of the m-th fibre after (i-1)-th work roller of rolling mill [-]
el,m,i elastic stress of the m-th fibre after i-th work roller of rolling mill [MPa]
res,m,i residual stress of the m-th fibre after i-th work roller of rolling mill [MPa]
θi,ent angle at the entrance between the cross-section axis of the strip and the vertical at the roller i-th [rad]
θi,exit angle at the exit between the cross-section axis of the strip and the vertical at the roller i-th [rad]
Fv,i vertical force generated by the i-th work roller [N]
Fi,ax external force generated by the i-th and (i-1)-th work rollers [N]
Fi resulting force generated by the bending moment at the i-th work roller [N]
Fi,ax,ent axial force generated at the entrance of the i-th work roller [N]
Fi,ax,exit axial force generated at the exit of the i-th work roller [N]

along the thickness direction, non-uniform mechanical properties due to uncontrolled microstructural transformations,
and improper staking during transportation1. In Roberts2 different types of shape defects are described as well as the
main influencing factors that lead to defects in cold-rolled strip. Concerning the strip flatness, the most frequent defects
are the coil set, generated by the re-coiling process, and the crossbow, created by an uneven cooling of the rolling
stock across the strip width or slitting. Edge wave, centre buckle, camber and twist are other typical defects that degrade
the sheet or make it unusable.
The processes commonly used to remove such defects are: (i) pure stretch levelling, (ii) multi-roller levelling (i.e.
roller levelling), and stretch-bend levelling (i.e. tension-levelling)3,4. In the former, high brindle drum tensions are
required to overcome the yield stress point and to flatten the incoming material. In multi-roller levelling multiple
rollers, arranged in a zig-zag position, deform the strip by repeated bending, and, decreasing the rollers gap while the
strip moves along the process, the curvature radius is increased and higher flatness is obtained5. In the latter a
longitudinal tension is superposed to the metal sheet bending through the rollers, which can be equal up to the 70% of
the initial yield stress6. Small rollers’ radii are commonly used in tension-levelling to reduce the magnitude of the front
and the back tension required since material yielding is due mainly to the bending contribution rather than the axial
tension7. Nowadays, most of the process parameters (i.e. roller-intermesh, tensile load, brindle-rollers velocities, roller-
pitches and roller-radii, etc.) are still determined by trial-and-error approaches8 and often do not allow a strict respect
of the tolerances with flatness errors up to 20 mm9. Industrial guidelines have revealed strong limitation for accurate
 T. Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere 3

process control: as example it can be mentioned that the suggested strip elongation may range from 0.3% up to 5% so
to make impossible to assure an acceptable process repeatability10,11.
A review of the scientific literature shows that both empirical and numerical approaches have been developed about
the selection of the optimal process parameters for sheet metal levelling. Sachs and Klinger12 and Tanaka et al.13
carried out pioneering experimental studies about the curling of metal sheets in asymmetric rolling, assuming that the
reason of this defect was correlated to the frictional forces. Based on these studies, Misaka and Masui14 developed an
analytical method to calculate quantitatively the sheet curl depending on the elongation. Later, Pospiech15 investigated
the influence of the main process parameters in asymmetric rolling, such as the lubrication, the strip thickness and the
rollers’ roughness. Many analytical models today-used are based on the investigations of Kinnavy16, in which the
neutral layer position was described as function of the strip tension under superposed bending by the forward and
backward loads. Doege et al.17 developed an analytic model to calculate the optimal rollers position considering the
residual stresses along the sheet thickness. Guericke18 presented a material model to describe the cyclic elastic-plastic
stress-strain generated by a cyclic tension-compression load during the levelling. Liu et al.19 analysed the curvature
of the strip after the levelling without taking into account the residual stresses and proposed a method for the fast
optimization of the roller position. With respect to the tension-levelling process, very few studies can be cited from
literature. Morris et al.6 studied the residual stresses and the Bauschinger effect, comparing analytical and FE
modelling. A FE model was developed by Yoshida and Urabe8 by using a sophisticated constitutive model based on
cycling plasticity and it was found that the roller intermesh has little influence on the product properties, while Huh 19
by using numerical analyses showed the correlation between the strip elongation and the number of rollers and
forward/backward tensions.
The present paper reports the results of an analytical model developed to compute the residual stresses in the metal
sheet thickness in tension-levelling, with the aim of allowing a reliable and prompt control of the process parameters,
considering a work-hardening material behaviour. The backward and the forward loads given by the brindles are
superposed to the bending moment given by the rollers. The results in terms residual stresses and neutral layer position
are evaluated.

2. Case study

2.1. Tension leveller unit

The configuration of the tension leveller is shown in Fig. 1. The tensile load applied to the strip is provided by the
brindle-rollers, whose speed is made to vary in order to maintain the constant elongation of the sheet, before and after
the process unit of the tension leveller. The so-called process unit is made of 5 working rollers that are grouped into:
(i) two couples of small-radius rollers followed by large-radius rollers (the latter adjustable and un-driven), and (ii) a
couple of large-radius rollers plus a single small-radius roller. The first group imposes the alternate bending to the
strip, while the latter is positioned at the exit of the process unit for the elimination of the crossbow defect.
The tension-levelling is a parameter-sensitive process: small variations of some parameters, such as the thickness
or the width of the strip, cause many differences in the final shape of the strip. It is for this reason that generally the
only parameter that is made to vary is the load generated by the brindles, thus keeping the intermesh between the work
rollers and their radius constant.
Because of the superposed tension that can expand the tensile zone and minimize the compression zone of the strip,
according to Marciniak21, the strip Neutral Layer (NL) moves towards the centre of curvature and changes its position
along the process due to the alternate direction of the bending moment. The deformation imposed to the strip can be
divided into three stages according to the loads that are applied, having respectively:
 deformation induced by the tension generated by the brindle rollers;
 elastic bending deformation and elastic-plastic deformation generated by the work rollers;
 elastic recovery during the unloading.
 4 T.Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere

Fig. 1. Tension-levelling process configuration.

3. Modelling

3.1. Hypotheses and assumptions

The following assumptions are given:

 the material of the sheet is homogeneous and isotropic;

 the value of Young modulus E is the same in tension and compression, and the Bauschinger effect is not
considered;
 no residual stresses and strains at the entrance of the tension leveller;
 the metal sheet cross-section remains plane during bending;
 the curvature radius is much larger than the sheet thickness and the initial and final curvature negligible;
 the thickness and the width deformations of the sheet can be neglected;
 no deformation impressed by the guide rollers;
 no difference in stress and strain values along the width of the sheet.
A work-hardening behaviour is assumed for the material of the sheet, according to the following equations:

𝐸𝜀 𝑤ℎ𝑒𝑛 𝜎 < 𝑌
𝜎={ (1)
𝐾𝜀 𝑛 𝑤ℎ𝑒𝑛 𝜎 ≥ 𝑌

where σ and are the stress and the strain of the material during the deformation.
 T. Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere 5

3.2. Longitudinal deformation

Fig. 2 shows the model of the sheet bending with superposed tension at the i-th roller. Since the radius of the roller
ri is more than four times the sheet thickness t, a plain strain condition can be assumed, according to Marciniak21. The
dash-dot line (in black) represents the axis of symmetry of the metal sheet cross-section, while the dashed line (in red)
corresponds with the NL of the sheet after the application of the axial load.

Fig. 2. Deformation of a longitudinal fibre in sheet bending with superposed tension.

The longitudinal strain at the first roller εm,1 of the m-th fibre of the metal sheet (dash-dot-dot line in blue in Fig. 2)
generated by the bending moment M at the first work roller is approximately equal to:
𝑡 𝑡
(𝑦+𝑟𝑖1 +2)𝛼 𝑦+𝑟1 +2
𝜀𝑚,1 = ln [ 𝑡 ] = ln ( 𝑡 ) (2)
(𝑟1𝑖 + +𝑒1 )𝛼 𝑟1 + +𝑒1
2 2

where r1 is the radius of the first work roller, α is the contact angle between the roller and the sheet, y is the distance
of the fibre from the cross-section axis of symmetry, and e1 is the distance of the NL from the axis of symmetry after
the first work roller. Under the assumptions of absence of residual strains at the entrance, according to Marciniak21, it
is possible to combine eqs. (1) and (2) to calculate the longitudinal stress σm,1 of the m-th fibre of the material generated
by the first work roller:
𝑡
𝑦+𝑟1 +2
𝐸 ∙ 𝑙𝑛 ( 𝑡 )
𝑟1 + +𝑒1 𝑤ℎ𝑒𝑛 𝜎𝑚,1 < 𝑌
2
𝜎𝑚,1 = 𝑡 𝑛 (3)
𝑦+𝑟1 +
2
𝐾 [𝑙𝑛 ( 𝑡 )] 𝑤ℎ𝑒𝑛 𝜎𝑚,1 ≥ 𝑌
{ 𝑟1 + +𝑒1
2

Considering the residual strains and stresses, eqs. (2) and (3) allow computing the strain (εm,i) and the stress (σm,i)
of a fibre m-th after the N-th work roller, that respectively become:

𝑡
𝑦+𝑟𝑖 +2
𝜀𝑚,𝑖 = ∑𝑁
𝑖=1 ln ( 𝑡 ) + ∑𝑁
𝑖=2 𝜀𝑒𝑙,𝑚,𝑖−1 (4)
𝑟𝑖 + +𝑒𝑖
2
 6 T.Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere

𝑡
𝑦+𝑟𝑖 + 𝑤ℎ𝑒𝑛 𝜎𝑚,𝑖 < 𝑌
𝐸 ∙ [∑𝑁
𝑖=1 ln ( 𝑡
2
) + ∑𝑁
𝑖=2 𝜀𝑒𝑙,𝑚,𝑖−1 ]
𝑟𝑖 + +𝑒𝑖
2
𝜎𝑚,𝑖 = 𝑡 𝑛 (5)
𝑦+𝑟𝑖 +
𝐾 [∑𝑁
𝑖=1 ln ( 𝑡
2
)+ ∑𝑁
𝑖=2 𝜀𝑒𝑙,𝑚,𝑖−1 ]
𝑤ℎ𝑒𝑛 𝜎𝑚,𝑖 ≥ 𝑌
{ 𝑟𝑖 + +𝑒𝑖
2

where the first term of eq. (4) represents the strain impressed by the rollers and the second term is the elastic strain
that the fibre m-th has recovered at the exit of the (N-1)-th work roller. Varying the value of y it is possible to compute
the stress and the strain along the thickness of the strip.

4. Analytical model to calculate residual stresses along the thickness

Fig. 3 shows the processing unit of the tension-leveller. The first and the second group can apply the same
deformation and they work alternately; this configuration allows maintenance to be performed without having to stop
the strip rolling process. So, three vertical forces are applied on the sheet metal, namely Fv,1, Fv,2 and Fv,3.
Due to the static equilibrium:

𝐹𝑖 = 𝐹𝑖,𝑎𝑥 (6)

By an iterative procedure it is possible to determine the shift ei of the NL from the axis of symmetry so that the
equilibrium condition between internal axial force Fi and total external axial forces Fi,ax is respected.
The internal axial force can be calculated as:
𝑤 𝑡
𝐹𝑖 = ∫ 2𝑤 ∫2𝑡 𝜎𝑚,𝑖 𝑦𝑤𝑑𝑦𝑑𝑤 (7)
− −
2 2

where σm,i is the longitudinal stress in the m-th fibre of the material after the i-th work roller, calculated with the eq.
(5). Since the stress σm,i is constant in the width, the previous equation can written as:
𝑡
𝐹𝑖 = w ∫2𝑡 𝜎𝑚,𝑖 𝑦𝑑𝑦 (8)
−
2

To calculate the value of the external axial forces, it is possible to use the data that the sensors placed on the shafts
of the brindles and the work rollers that are continuously monitored during the process. In particular, the vertical forces
Fv,1, Fv,2 and Fv,3 recorded by the sensors are used, whose components along the strip axis are calculated according to
the scheme shown in Fig. 4.

Fig. 3. Vertical loads imposed by the work-rollers.

 T. Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere 7

Fig. 4. Components of the vertical force along the strip.

Knowing the horizontal and vertical positions of two consecutive rollers, respectively xi, yi, xi+1, yi+1, or xi-1, yi-1, xi, yi,
it is possible to calculate the angles with which the vertical force Fv,i impressed by the roller generates a load along
the sheet axis, according to eqs. (9) and (10), while with eqs. (11) and (12) it is possible to calculate the force along
the strip axis before and after the i-th work roller:

𝑥𝑖 −𝑥𝑖−1
𝜃𝑖,𝑒𝑛𝑡 = arctan ( ) (9)
𝑦𝑖 −𝑦𝑖−1

𝑥𝑖+1 −𝑥𝑖
𝜃𝑖,𝑒𝑥𝑖𝑡 = arctan ( ) (10)
𝑦𝑖+1 −𝑦𝑖

𝐹𝑖,𝑎𝑥,𝑒𝑛𝑡 = 𝐹𝑣,𝑖 ∗ cos(𝜃𝑖,𝑒𝑛𝑡 ) (11)

𝐹𝑖,𝑎𝑥,𝑒𝑥𝑖𝑡 = 𝐹𝑣,𝑖 ∗ cos(𝜃𝑖,𝑒𝑥𝑖𝑡 ) (12)

Having found the value of ei such that the equilibrium condition between internal forces Fi and external forces
Fi,ax,ent expressed in eq. (6) is respected, it is possible through eqs. (4) and (5) to obtain the value of the strain εm,i and
stress σm,i of the m-th fibre. The total distribution of the stresses and of the strains in the thickness can be obtained as
function of the distance y between a fibre and the cross-section axis of symmetry of the strip. The elastic stress (σel,m,i)
at the exit of the i-th roller, is obtained from the balance:
𝑤 𝑡 𝑤 𝑡

∫−2𝑤 ∫2𝑡 𝜎𝑚,𝑖 𝑦𝑤𝑑𝑦𝑑𝑤 = ∫−2𝑤 ∫2𝑡 𝜎𝑒𝑙,𝑚,𝑖 𝑦𝑤𝑑𝑦𝑑𝑤 (13)

2
− 2
−
2 2

Since the stresses σm,i and σel,m,i are constant in the width, and the elastic stress has a linear behaviour along the
thickness, as shown in Fig.5, it is possible to rewrite the previous equation as follows:
𝑡 𝑡

∫2𝑡 𝜎𝑚,𝑖 𝑦𝑑𝑦 = ∫2𝑡 𝜎𝑒𝑙,m,𝑖 𝑦𝑑𝑦 (14)

− −
2 2
By adding the springback component σel,m,i and combining eqs. (5) and (14) the residual stresses and the elastic
strain along the thickness can be calculated as:
 8 T.Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere

𝜎𝑟𝑒𝑠,𝑚,𝑖 = 𝜎𝑚,𝑖 + 𝜎𝑒𝑙,𝑚,𝑖 (15)

𝜎𝑒𝑙,𝑚,i
𝜀𝑒𝑙,𝑚,i = (16)
𝐸

5. Results

Fig. 5 shows the distribution of the stress σm,i ,the elastic component σel,m,i, the residual stresses σres,m,i obtained by
combining all the components in eq. (15), and the distribution of the strain at the i-th roll εm,i. During the process the
NL shifts from the axis of symmetry position, to permit the equilibrium between the external and the internal forces.

Fig. 5. Stresses and strains distribution along the process.

Table 1. Model results.

Work roll 1 Work roll 2 Work roll 3
Vertical force [N] (experimental) 89289 82372 127871
Entrance angle [deg] (experimental) 32 22 51
Exit angle [deg] (experimental) 22 14 49
Forward axial force [N] (experimental) 75721 - -
Backward axial force [N] (experimental) - - 80952
Entrance axial force [N] (calculated) 75721 76374 80472
Exit axial force [N] (calculated) 96301 84894 194908
Friction force [N] (calculated) 20580 8520 114436
NL shift [mm] (calculated) -0.081 0.226 -0.412
Maximum stress [MPa] (calculated) 374.99 400.95 418.63
Maximum residual stress [MPa] (calculated) 251.36 186.10 188.14
Minimum residual stress [MPa] (calculated) -251.45 -347.75 -406.87
 T. Magro et al./ XIV Convegno dell’Associazione Italiana Tecnologie Manifatturiere 9

Table 1 shows the results of the calculation of the analytical model using the input parameters experimentally
measured during the tension levelling of a carbon steel at the Marcegaglia S.p.A. Ravenna plant, namely the forward
and the backward forces measured along the strip rolling direction at the entrance and at the exit of the process unit,
the vertical forces on the working rollers and the strip inclination during the process. By using the analytical model it
was possible to estimate (i) the shift of the NL from the mid-layer, (ii) the maximum stress during the load, and (iii)
the maximum and minimum values of the residual stresses after the unloading at the i-th work roll.

6. Conclusions

The paper describes an analytical approach to predict the stress-strain distribution during and after the tension-
levelling process. Depending on the geometric parameters of the coil, namely the thickness, the width, and the material
constitutive parameters, the model allows calculating both the residual stresses and strains in the metal sheet thickness
during the whole process. The model allows the computation of the vertical loads as function of the positions of the
rollers, as well as the influence of the rollers intermesh on both the residual stresses and the axial load.

Acknowledgements

The authors wish to thank the company Marcegaglia S.p.A (Italy), for the support provided in the definition of the
industrial case and in the activities carried during the creation of the analytical model.

References

1. Tomita S, Fujita Y. Recent trend and problems of straightening technologies of thick plates. J. Jpn. Soc. Technol. Plast. 1999; 40:408-11
2. Roberts WL. Cold rolling of steel. Marcel Dekker, New York, 1978
3. Park KC, Hwang SM. Development of a finite element analysis program for roller levelling and application for removing blanking bow defects
of thin steel sheet. ISIJ Int. 2002; 42:990-99
4. Sheppard T, Roberts JM. Shape control and correction in strip and sheet. Int. Met. Reviews. 1973; 18:1-18
5. Lee JB, Kang SS. Numerical modelling of roller leveller for thick plate levelling. Int. J. Precis. Eng. Manuf. 2018; 19:425-30
6. Morris JW, Hardy SJ, Lees AW, Thomas JT. Formation of residual stresses owing to tension levelling of cold rolled strip. Ironmaking
&Steelmaking. 2001; 28:44-52
7. Huh H, Lee HW, Park SR, Kim GY, Nam SH. The parametric process design of tension levelling with an elasto-plastic finite element method.
J. Mater. Proc. Tech. 2001; 113:714-19
8. Yoshida F, Urabe M. Computer-aided process design for the tension levelling of metallic strips. J. Mater. Proc. Tech. 1999; 89:218-223
9. Morris JW, Hardy SJ, Thomas JT. Some fundamental considerations for the control of residual flatness in tension levelling. J. Mater. Proc. Tech.
2002; 12:385-96
10. Zhou C, Jiang Z, Wang G, Xie D, Zhang Y. Analysis of tension and deformation curvature in strip tension levelling. Adv. Mater. Research.
2010; 97-101:71-74
11. Huh H, Heo JH, Lee HW. Optimization of a roller levelling process for Al7001T9 pipes with finite element analysis and Taguchi model. Int. J.
Mach. Tool. Manuf. 2003; 43:345-50
12. Sachs G, Klinger L. The flow of metals through tools of circular contour. J. Appl. Mech. 1947; 14:88-98
13. Tanaka E, Tsunokawa K, Fukuda T. Curling and bowing of rolled strips. Trans. Jpn. Inst. Met. 1963; 4:124-33
14. Misaka Y, Masui T. Shape correction of steel ship by tension leveller. Iron Steel Inst. Jpn. 1971; 18:475-84
15. Pospiech J. A note on the influence of some factors affecting curvature in the flat rolling of strip. J. Mech. Work. Tech. 1987; 15:69-80
16. Kinnavy MG. Methods of permanently elongating strip. United States Patent Office 3552175
17. Doege E, Menz R, Huinink S. Analysis of the levelling process based upon an analytic forming model. CIRP Annals – Manufacturing
Technology. 2002; 51:191-94
18. Guericke W. Material model describing cyclic elastic-plastic deformation of roller levelling and straightening processes. Steel Research Int.
2009; 80:281-87
19. Liu Z, Luo Y, Wang Y, Liao TW. A new curvature analytical method in plate levelling process. ISIJ Int. 2018; 58:1094-101
20. Huh H, Lee HW, Park SR, Kim GY, Nam SH. The parametric process design of tension levelling with an elasto-plastic finite element method.
J. Mater. Proc. Tech. 2001; 113:714-719
21. Hu J, Marciniak Z, Duncan J. Mechanism of sheet metal forming. Elsevier, 2002