Professional Documents
Culture Documents
Journal of Manufacturing Science and Engineering APRIL 2016, Vol. 138 / 041002-1
C 2016 by ASME
Copyright V
system is carried out. In Ref. [22], a finite-element based model Using Galerkin’s method and putting the approximated
has been proposed for the prediction of the deformed roll profile assumed vertical deflection ve in Eq. (1), the residual is
in a 20-high Sendzimir mill. In the present work, the working roll
is considered as a continuous system, and the FEM has been used d2 d 2 ve d 2 ve
to analyze the system as discussed in Secs. 2 and 3. Re ¼ 2 EI 2 þ qA 2 fes ke ve (2)
dz dz dt
2 Mathematical Modeling Equating the weighted residual in the element to zero, one obtains
In this section, the mathematical modeling of a four-high ðh ðh 2
e e e d d 2 ve d 2 ve s e
rolling mill is carried out using the FEM. Figure 1(a) shows the w R dz ¼ w EI 2 þ qA 2 fe ke v dz ¼ 0
0 0 dz2 dz dt
schematic diagram of a four-high rolling mill having two working
rolls and two backup rolls. The working roll is modeled as an (3)
Euler Bernoulli beam by taking beam elements with vertical dis-
where we is the weighted parameter, and h is the length of the
placement and slope as the nodal degrees-of-freedom in the finite-
element. Hence,
element formulation. The bearings at the ends of the working rolls
are modeled using spring elements. Figure 1(b) shows the upper ðh 2 ðh ðh
d d 2 ve e d 2 ve e s
half of the rolling mill with supports modeled by springs and the EI w dz þ qA w dz ¼ fe ke ve we dz
dz2 dz 2 dt 2
coordinate system used in this analysis. In Sec. 2.1, the governing 0 0 0
equation of motion for the working roll is developed by FEM, and (4)
the expression for an approximate value of the end bearing stiff-
ness is found. In Sec. 2.2, the roll bite submodel and the work roll Using integration by parts, the first part of the expression is inte-
submodel are described to calculate the sheet force and stiffness. grated twice, so that Eq. (4) is converted to the weak form as
2.1 FEM Formulation. As shown in Fig. 1(b), the working d d2 ve h dwe d 2 ve h
e
roll has been divided into three zones. In zones 1 and 3, there is w EI 2 EI 2
dz dz 0 dz dz 0
no sheet metal but in zone 2 the sheet metal width is L2. The ðh 2 e ð
2 e h
working roll is discretized in n elements with n ¼ n1 þ n2 þ n3 . d w d v e d 2 ve
Here, n1 ; n2 ; and n3 are the numbers of elements in zones 1, 2, and þ 2
EI 2 dz þ w qA 2 dz
0 dz dz 0 dt
3, respectively. It can be noted that an interesting node is taken ðh
s
between nodes n1 þ 1 and n1 þ n2 þ 1, where the working roll is ¼ fe ke ve we dz (5)
in contact with sheet metal. A typical two-node beam element of 0
the working roll is shown in Fig. 2. The governing equation of
motion of this Euler Bernoulli beam element can be written as
d2 d2 v d2 v
EI þ qA ¼ fes ke v (1)
dz2 dz2 dt2
Fig. 3 Force acting on the beam at the boundary Fig. 4 Free-body diagram of the working roll
Journal of Manufacturing Science and Engineering APRIL 2016, Vol. 138 / 041002-3
x2i
hð xÞi ¼ hci þ (20)
R
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x1i ¼ ðh1i hci ÞR (22)
Rhci h_ ci
x2i ¼ (23)
2 u1i h1i x1i h_ ci
Fig. 5 Roll bite model of a rolling process, see Ref. [18]
pffiffiffiffiffiffiffiffiffi m1 x1i
xni ¼ Rhci tan tan 1 pffiffiffiffiffiffiffiffiffi
m1 þ m2 Rhci
sy 2
fis ¼ ð2sy þ r1m Þðx2i x1i Þ þ m2 x2i x2ni þ m1 x21i x2ni m1 x 2i 1
R þ tan 1 pffiffiffiffiffiffiffiffiffi þ
m1 þ m2 Rhci 2s y ð 1 þ m2 Þ
m
h2i h1i rffiffiffiffiffiffi
þ sy Rm2 ln þ sy Rm1 ln
hni hni hci h2i
rffiffiffiffiffiffi r2m r1m þ 2sy ln (24)
R h1i
h1i R xni
þ 2sy x2i ln þ 2sy ðm1 þ m2 Þx2i tan 1 pffiffiffiffiffiffiffiffiffi
h2i hci Rhci 1 x2ni _
rffiffiffiffiffiffi! u1i ¼
h1i
vr hci þ vr
R
þ ðx1i xni Þh ci Þ (25)
pffiffiffiffiffiffiffiffiffi R x1i
þ 4sy Rhci 2sy m1 x2i tan 1 pffiffiffiffiffiffiffiffiffi
hci Rhci
u1i h1i þ ðx2i x1i Þh_ ci
rffiffiffiffiffiffi! u2i ¼ (26)
pffiffiffiffiffiffiffiffiffi x2
R x2i hci þ 2i
4sy Rhci þ 2sy m2 x2i tan 1 pffiffiffiffiffiffiffiffiffi (16) R
hci Rhci
T1
Here, R is the radius of the working roll, sy is the yield strength of r1m ¼ (27)
sheet metal in shear, m1 and m2 are the friction coefficients before h1m L2
and after the neutral point, r1m and r2m are the mean tensile
stresses at the entry and the exit, u1i and u2i are the velocities of and
the strip at the entry and the exit of the roll bite at the ith node, h1i
T2
and h2i are the thickness of the strip at the entry and the exit of r2m ¼ (28)
roll bite, hci is the gap between two working rolls, hc0 is the initial h2m L2
position of the working roll, h_ ci is the rate of change of roll gap
where T1 and T2 are the entry and exit tension, and h1m and h2m
between two working rolls along the centerline of the rolls,
are the entry and exit mean thickness of the strip. The roll gap hci
x1i ; x2i , and xni are the distances between strip entry, exit, and
will be related to the nodal displacement vi as
neutral point position from the centerline of the working roll, and
x_ ci is the horizontal velocity of the vibrating rolls. Hence, the hci ¼ hc0 þ 2vi (29)
element sheet force per unit length of the element is
fis þ fiþ1
s and
fes jith element ¼ (17)
2 h_ ci ¼ 2v_ i (30)
Due to the compressive load fis , there will be a deformation at the
contact of the working roll and the backup roll. This deformation where hc0 is the initial roll gap. To find out the stress variation
at the interface is modeled using a linear spring between them. at the entry and at the exit, following expressions are used from
Due to large size of the backup roll, one may neglect the vertical Ref. [9]:
deflection of the backup roll and hence, this force is calculated as
E u1i
the product of the stiffness of the spring and the vertical displace- r1i ¼ r1m þ 1 (31)
ment of the working roll. The contact stiffness ki per unit length 1 2 u1m
(when the roll bite is applied to the ith node) can be written as
described in Ref. [18] and
pE E u2i
ki ¼ !! (18) r2i ¼ r2m þ 1 (32)
1 2 u2m
1 0:78125EðDW þ DB Þ
2ð1 Þ þ ln
2
3 fis ð1 2 Þ where u1m and u2m are the mean entry and exit velocity of the
strip. The expression of entry stress has been simplified for this
where E is the Young’s modulus, is the Poisson ratio, and Dw model by assuming constant incoming strip width.
and DB are the diameters of the working and the backup roll. To get the desired output from the developed FEM model, one
Hence, the contact stiffness per unit length of the element is can use the flow chart shown in Fig. 6. It shows the steps involved
or
Journal of Manufacturing Science and Engineering APRIL 2016, Vol. 138 / 041002-5
N1 Number of elements in zone 1 20 Fig. 8 Time response of different nodes on the working roll for
N2 Number of elements in zone 2 60 kw 5531010 N/m and L2 50:508 m
N3 Number of elements in zone 3 20
d Constant parameter in Newmark Beta method 0.75
b Constant parameter in Newmark Beta method 0.5 is always less than 1. The value of u1 =vr is highest (0.7545) at
hc0 Initial position of working roll (m) 1:8 103 the center of working roll and depends on the friction condition
v0 Initial nodal displacement (m) 0:1% of hc0 (m1 and m2) before and after the neutral point. Figure 11(d) shows
v_ 0 Initial nodal velocity (m/s) 0 the variation in the ratio of exit velocity (u2 Þ and velocity of the
working roll (vr) and due to the forward slip the ratio is found to
be larger than 1. The value of u2 =vr is lowest (1.0227) at the cen-
ter of the working roll and is maximal at the edges (1.0229). The
difference in the velocity ratio (u2 =vr ) at the center and edge is
the same as in Ref. [9] by using a model coupling method.
Figure 12(a) shows the variation of the average entry velocity
with time. This variation in the entry velocity is responsible
for the nonlinear and parametric excitation in the rolling mill.
Figure 12(b) shows the variation in the average exit velocity with
time. In this case, the initial overshoot in the velocity is 17.82 m/s
and after 0.0065 s the velocity becomes constant (17.65 m/s). The
Fig. 7 Node selection in different zones variation in the exit velocity is similar to the variation in the roll
gap as shown in Fig. 8. This exit velocity variation is also respon-
sible for the parametric excitation in the moving strip between the
Fig. 9(b), the distribution of sheet force over the length of the two stands.
working roll is plotted for the steady state only. The sheet force is Figure 13 shows the variation of exit stress which is obtained
found to be smallest at the nodes, which is at the middle of strip by using Eq. (32). It is observed that the stress in transient state is
and maximal at edges because the reduction in thickness will be varying from 30 MPa to 130 MPa at the edges of the strip and in
higher at the edges due to bending of working roll at the center. the center it varies from 10 MPa to 90 MPa. Maximum variation
The difference of minimum to maximum value of the sheet force in the exit stress from the edge to the center is found in the very
over the length of the sheet is approximately 3 104 N/m, which beginning when the process starts and its value is approximately
is around 0.9% of the average sheet force in steady state. This 120 MPa which is found to be same as reported in the literature
sheet force distribution is found to be similar to the results [3,8,9]. Due to this variation in the stress over the strip, many
reported in Refs. [6,8]. defects, such as centerline cracking, warping edge, and wavy
Figure 10(a) shows the transient and steady-state distribution of edge, may develop at the exit. Figure 13(b) shows the distribution
relative exit thickness of sheet (strip crown corresponding to C0 of the exit stress over the strip in the steady state and it has been
locations) over the working roll length. It can be noted that the found that the difference in the exit stress from edges to the center
thickness at the center of the working roll (node C) is highest and is lesser (11 MPa) than the difference obtained in transient state,
least at the edges (node 1) due to the bending of the working roll. but even when working rolls are not vibrating there is a variation
Further it is observed that in the transient state, the relative thick- in exit stress which may cause defects. Figure 13(c) shows the
ness at the center varies from 7.5 lm to 17 lm, this implies that variation in average exit stress with time. Initially, the value of the
the strip crown C(z) corresponding to C0 locations (z ¼ 0.381 mm) average stress is about 54 MPa and similarly as the roll gap vibra-
is varying from 7.5 lm to 17 lm. Figure 10(b) shows the exit tion stops after 0.0065 s, the average stress remains constant at
thickness variation over the working roll length in steady state. 52.7 MPa.
This model predicts that for the input parameters in Tables 1–3,
the thickness (1.865 mm) at a distance of 25 mm from the edge of
the strip (z ¼ 0.406 mm) is 0.010 mm less than the center thickness 3.1 Effect of the Stiffness of the Bearings. As it has already
(1.875 mm). Hence, the C25 crown is 0.010 mm or 0.53% of the been assumed that the ends of the working rolls and backup rolls
center thickness. are placed on elastic foundations, the effect of the stiffness of
Figures 11(a) and 11(c) show the transient and steady-state var- these ends is discussed in this section. In this subsection, the
iation in entry and exit velocity over the length of the working stiffness kw is taken as 5 109 N/m which is lower than the stiff-
roll, plotted by using Eqs. (25) and (26), respectively, which is ness 5 1010 N/m considered in Sec. 3.
generated because of the change in roll gap at each node with Figure 14 shows the roll gap at different nodes A, B, and C in
respect to time. Figure 11(b) shows the variation in the ratio of the the working roll. It can be observed that due to a decrease in the
entry velocity (u1) and the velocity of the working roll (vr) in stiffness at ends, the roll gap at node A is now similar to the rest
steady state. It can be noted that due to the backward slip this ratio of the nodes in the working roll. Comparing Figs. 8 and 14, it is
Fig. 10 Exit thickness distribution of sheet over the length of working roll for kw 5531010 N/m
and L2 50:508 m: (a) both transient and steady state and (b) steady state
Fig. 11 (a) Distribution of entry velocity of strip over the length of working roll, (b)
distribution of the ratio of entry velocity of strip and working roll velocity in steady state, (c)
distribution of exit velocity of strip over the length of working roll, and (d) distribution of the
ratio of entry velocity of strip and working roll velocity in steady state, kw 5531010 N/m and
L2 50:508 m
Journal of Manufacturing Science and Engineering APRIL 2016, Vol. 138 / 041002-7
Fig. 13 (a) Exit stress distribution over the length of the working roll in both transient and
steady state, (b) exit stress distribution over the length of the working roll in steady state, and
(c) variation in average exit stress with time, kw 5531010 N/m and L2 50:508 m
observed that the initial overshoot, settling time, and the steady-
state roll gap increase with decrease in the bearing stiffness.
Figure 15(a) shows the sheet force distribution over the length
of the working roll. The value of sheet force in the transient state
varies from 2.75 MN/m to 3.9 MN/m, which is similar as found in
Fig. 9(a) for higher bearing stiffness. In case of steady state, from
Fig. 15(b), the difference in the sheet force at the edges and at the
center of the strip is found to be 2 104 MN/m which is less than
that for the bearing with higher stiffness, see Fig. 9(b). Also, the
value of average force decreases in the present case.
Figure 16(a) shows the distribution of the exit thickness of the
sheet relative to the thickness at the edges in the transient and
steady state. It can be observed that in transient state, the relative
thickness at the center varies from 4.5 lm to 14.5 lm. Hence, the
strip crown C(z) corresponding to C0 locations (z ¼ 0.3810 mm) is
varying from 4.5 lm to 14.5 lm. Figure 16(b) shows the exit
thickness variation over the working roll length in the steady state.
It is observed that the thickness (1.887 mm) at a distance of Fig. 14 Time response of nodes A, B, and C on the working
25 mm from the edge of the strip (z ¼ 0.406 mm) is 0.007 mm less roll for kw 553109 N/m and L2 50:508 m
Fig. 16 Exit thickness distribution of sheet over the length of working roll for kw 553109 N/m
and L2 50:508 m: (a) both transient and steady state and (b) steady state
than the center thickness (1.8894 mm). Hence, the C25 crown is width. Also, in this case the settling time in roll gap is 0.0045 s
0.0077 mm or 0.37% of the center thickness. From Figs. 10(b) and which is less than for the case with lower sheet width. This may
16(b), it is observed that with decrease in the bearing stiffness, the be due to the higher damping with increase in the sheet width. It
value of C25 crown decreases. It may be noted that with a smaller is observed that in this case of higher sheet width, the value of the
value of the crown, the exit strip profile is better and less material sheet force in the transient state varies from 2.2 N/m to 3.9 N/m.
has to be removed from the center of the strip to make it flat. In the steady state, the difference between the force at the edge
It is observed that with a decrease in bearing stiffness to and at the center of the strip is found to be 10 104 N/m, which
5 109 , in the transient state, the exit stress varies from 30 MPa is higher than the difference calculated for lesser width of strip
to 90 MPa at the edges of the strip and in the center it varies from (3 104 N/m) in Fig. 9(b). The value of the average force also
30 MPa to 90 MPa. Maximal variation in the exit stress from the increases for higher sheet width. In the present case, in the tran-
edge to the center is found to be approximately 122 MPa which is sient state the strip crown C(z) corresponding to C0 locations
almost the same (120 MPa) for kw ¼ 5 1010 , see Fig. 13(a). (z ¼ 0.1850) is varying from 18 lm to 68 lm. In steady state, the
Also, in the steady state, the difference in the exit stress from the thickness of the outcoming strip (1.868 mm) at a distance of
edges to the center is 8 MPa, which is less than the previously cal-
culated value of 11 MPa for kw ¼ 5 1010 . It may be noted that
the defect in the final product will be less for the less variation of
the exit stress in the strip over the working roll length.
Further it is observed that with a decrease in the value of the
bearing stiffness, the stress variation over the length of the work-
ing roll decreases and the settling time increases.
Journal of Manufacturing Science and Engineering APRIL 2016, Vol. 138 / 041002-9