Professional Documents
Culture Documents
www.elsevier.com/locate/ijmactool
Abstract
A model of the bending process in Press Brakes is established using Timoshenko beam theory. Expressions for the workpiece bending
error are derived that explicitly consider the influence of shape, dimensions and initial deformation of the machine structural components on
its bending accuracy. The minimization of the bending error is formulated in terms of optimisation problems that are solved numerically
using a genetic algorithm. The methodology presented in this paper permits the analysis of existing Press Brake design solutions, the
optimisation of their performance and the development of new solutions.
q 2005 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: C351 212948567; fax: C351 212948531. The bending model is shown in Fig. 2a and is assumed to
E-mail address: pgc@fct.unl.pt (P.G. Coelho). be symmetric. The ram and bed are modelled as simply
0890-6955/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. supported beams and will be denoted, respectively, by
doi:10.1016/j.ijmachtools.2005.01.030 Upper and Lower beam.
1452 P.G. Coelho et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1451–1460
Fig. 1. Influence of ram and bed deflection on the angular precision of the
workpiece. (a) The ‘boat belly’ effect. (b) Uniform bending angle.
In Eq. (2) M is the bending moment and q the applied using Eq. (2):
load and in Eq. (3) j is the rotation of the cross-section due
Mu Kq M q
to bending, b the shear deformation and V the shear force. C Z lC (7)
Although Eq. (1) reduces to the technical theory equation EIu ku GAu EIl kl GAl
for uniform load q, the boundary and interface conditions
Solutions for this equation are possible only for Press
are different. Relevant to this study are the following two
Brakes with sandwich Lower beams, 0%d!L/2, because
interface conditions:
when dZL/2, MuZKMl.
qðL K 2aÞ For an illustrative example a rectangular cross-section
bjK
xZL=2Kd K bjxZL=2Kd Z
C
(4) with the same width for both beams is assumed with kuZ
2kl Al G
klZ0.85, a constant cross-section for the Upper beam with
VjxZt 1 1 height of 1400 mm and bending length of 3200 mm. The
bjK
xZt K bjxZt Z
C
K (5) Lower beam has variable height h(x).
ku G Au2 Au1
The shape optimisation problem of Eq. (7) consists in
The indexes l and u refer to Lower and Upper beam, finding the Lower beam shape h(x) that makes the
respectively. In Eq. (5) A takes the constant values Au1 for curvatures of the Upper and Lower beams match for this
0!x!t and Au2 for t!x!L/2. Eqs. (4) and (5) express the single bending length.
shear strain discontinuity related to the discontinuity of V at Substituting the above values in Eq. (7) yields an implicit
the supports of the Lower beam and to the variation of the function of h in variable x. This function is shown in Fig. 3
Upper beam cross-sectional area. According to Eq. (3) these for dZ0 and dZ400 mm.
discontinuities in the shear strain will originate slope Although the curvatures of the Upper and Lower beams
discontinuities. are the same in the above solutions, parallelism will only
Since the Upper and Lower beams are statically occur for the case in which the effect of shear deformations
determinate their deflections may be determined by is disregarded. As expressed in Eq. (4), the shear force
integrating twice the curvature (Eq. (2)) with the adequate discontinuities that occur at the supports will generate slope
boundary and interface conditions: discontinuities in the Lower beam deflection.
ðð In order to guarantee slope continuity the Lower beam
M q
wZ C dxdx C C1x C C2 (6) may be supported by a uniformly distributed reaction, like
EI kGA
the one shown in Fig. 4a. In this case shear force and shear
angle are continuous and q2 is determined to satisfy static
equilibrium.
As shown in Fig. 4b the load discontinuity applied to the
Lower beam implies, from Eq. (7), an abrupt variation of h
3. Shape optimisation for curvature continuity, but the deflections of both beams
become parallel.
In order to obtain parallelism between the deflected What kind of support for the Lower beam will produce a
Upper and Lower beams their curvature must be equal. For distributed reaction? A technical solution existent in the
each bending length this condition may be expressed as, market and patented [3] (see Fig. 5) resembles the computed
4. Dimensional optimisation
p Z pb C ps
qL4 24ð1 C nÞ rl
2
Z f r ; z; j C f ðr ; r ; z; jÞ
24EIl b I ku L s k A
(15)
The Eq. (15) for the penetration is given by the product
of a factor proportional to the Lower beam maximum
displacement and a non-dimensional factor. This includes
the bending contribution fb and the shear contribution,
proportional to fs. The influence of the shear term
Fig. 5. Patented solution for Lower beam. grows proportionally to (rl/L)2, as it is well known.
P.G. Coelho et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1451–1460 1455
The cross-sectional shape of the beams and its influence in was used instead of Timoshenko’s theory, ‘shear only’ as if
shear stiffness is present through rk and ku. the bending stiffness of the beams was very high and all the
contribution to the bending error came from shear and a
4.2. Formulation of the optimisation problem third one with both contributions taken into account. A
rectangular cross-section for the Upper beam was con-
The dependence of the penetration on selected design sidered and both a rectangular and ‘T’ shape cross-sections
variables is explicitly considered by writing pZp(s, a, x). for the Lower beam. The last column gives uðps0 ;a0 Þ for
The components of sZ(s1, s2, ., sn) are the n variables Upper beam dimensions of 3200!90!1600 mm3 and
related to cross-section dimensions of the beams and the bending force per unit length q of 180 N/mm. Fig. 6
distance between locking rods, which have lower and upper shows the penetration curves for each case in Table 1.
bounds, sK i and sC i , iZ1,., n, respectively. Variable a
defines a bending length, x is the position variable and both
retain their previous meaning.
The formulation of the dimensional optimisation pro-
blem is the following: for a given bending force q and
maximum bending length L, find the design s0 of a Press
Brake for which the maximum oscillation of the penetration
function is minimum. This can be expressed as: find the
(optimal) design s02D1, the bending length a02D2 and the
penetration pso ;ao such that,
uðps0 ;a0 Þ Z min max uðps;a Þ;
s2D1 a2D2
(16)
uðps;a Þ Z max pðs; a; xÞ K min pðs; a; xÞ
x2D0 x2D0
D1 Z s 2Rn =sK
i % si % si ; i Z 1; .; n ;
C
(17)
L
D2 Z a 2R = 0% a%
2
From the different expressions for the penetration (see,
for example, Eq. (15)) it can be concluded that the optimal
design s0 is not dependent on q and that the bending error is
proportional to it.
To evaluate the solutions of the minimax in Eq. (16) a
genetic algorithm [12,13,18] is used since it does not depend
on the analytical properties of the objective function.
Table 1
Optimal values for bending and shear contributions separately and together
0.0129 and 0.0253 mm for uðpv0 ;s0 ;a0 Þ—a gain in bending
accuracy of 7 and 18% (see Table 2).
Problem (16) was also solved for UB1/LB1 and
UB2/LB2 but setting s8ZdZL/2 and augmenting the
bounds on width for the Lower beam because there are no
side plates to accommodate. At the optimum (see Table 3)
all the bounds on dimensions were attained and the uðps0 ;a0 Þ
values calculated are an order of magnitude higher for this
conventional solution than for the corresponding sandwich
designs.
The introduction of an optimal initial deformation for
this case achieves a significant decrease in the bending error,
as shown in Table 4. In the case UB1/LB1 the bending error
decreases from 0.1092 to 0.0025 mm and from 0.2658 to
Fig. 9. Comparison between analytical (Timoshenko theory) and Finite
Element numerical results for UB2/LB2 using maximum bending length. 0.0064 mm in the case UB2/LB2. The improvement in
(a) Upper beam. (b) Lower beam. bending accuracy is radical for these conventional designs
because both beams deform with the same shape for all
Let S(v,x) be a cubic spline [19] with nodes (x1, xi,., bending lengths.
xm), m being the number of equally spaced points along half In the case of the sandwich solution for the Lower beam,
of the Upper beam. The components of the vector vZ different bending lengths originate very different defor-
(v1,vi,.,vm) are the initial deformation values at the points mations and there is no unique initial deformation that
xi and their bounds are given by: reduces appreciably the error for all lengths.
The correction introduced by an initial deformation is
D3 Z v 2Rn =vK i % vi % vi ; i Z 1; .; m
C
(18)
associated with a load value q. If the load changes and the
S(v,x) is added to penetration given by Eq. (9) in the initial deformation remains constant, the error will not
following way change proportionally to its value at q.
In order to introduce the optimal initial deflection for the
p ðxÞ Z ½wu ðxÞ K wl ðxÞ K d C Sðv; xÞ K SjxZa (19) actual load q, it seems advantageous to use an automatic
Table 3
Optimal dimensions for UB1/LB1 and UB2/LB2 with conventional location of the supports (units: mm)
Table 4
Optimal initial deformation for UB1/LB1 and UB2/LB2 with conventional location of the supports (units: mm)
[13] P.S.G. Coelho, L.O. Faria, J.M.B.B. Cardoso, Optimização dimen- [16] G.R. Cowper, The shear coefficient in Timoshenko’s beam theory,
sional, de forma e de uma deformada inicial para quinadoras, in: Proc. J. Appl. Mech. 33 (2) (1966) 335–340.
VIII Congresso de Mecânica Aplicada e Computacional, LNEC, [17] J.B. Pacheco, Utilização de Quinadoras e Guilhotinas, APTCP, Porto,
Lisboa, 31 May–2 June, 2004. 1992.
[14] C.L. Dym, I.H. Shames, Solid Mechanics—A Variational Approach, [18] D.E. Goldberg, Genetic Algorithms in Search, Optimisation, and
McGraw-Hill, Kogakusha, 1973 (pp. 174–194). Machine Learning, Addison-Wesley, Reading, MA, 1989.
[15] S.P. Timoshenko, Strength of Materials, vol. I, D. Van Nostrand [19] H. Pina, Métodos Numéricos, McGraw-Hill, Portugal, 1995 (pp.
Company, Inc., New York, 1970. 77–88).