Вы находитесь на странице: 1из 7

Proceedings of the ASME 2015 International Mechanical Engineering Congress and Exposition

IMECE2015
November 13-19, 2015, Houston, Texas

IMECE2015-53146

SHELL ELEMENT FORMULATION BASED FINITE ELEMENT MODELING, ANALYSIS


AND EXPERIMENTAL VALIDATION OF INCREMENTAL SHEET FORMING PROCESS

GOVIND N. SAHU1, SUMIT SAXENA1, PRASHANT K. JAIN1, J. J. ROY2, M. K. SAMAL2, PUNEET


TANDON1

1
PDPM Indian Institute of Information Technology, Design & Manufacturing Jabalpur, Jabalpur, 482005,
Madhya Pradesh, India
2
Bhabha Atomic Research Centre, Trombay, Mumbai, 400085, Maharashtra, India

ABSTRACT of ISF include thinning of the wall of the formed geometry,


This paper presents the effect of shell element formulations bending at the beginning and reverse bending at the base of
on the response parameters of incremental sheet metal forming formed component. This work adapts Finite Element Analysis
process. In this work, computational time, profile prediction and (FEA) to model the ISF process to improve the overall
thickness distribution are investigated by both finite element geometrical accuracy.
analysis and experimentally. The experimental results show that The degenerated continuum shell element formulation was
the thickness distribution is in good agreement with the results first introduced by Ahmad et al. [1] and it is common and
obtained with Belytschko-Tsay (BT) and Improved Flanagan- independent of any shell theory of finite element method. First
Belytschko (IFB) shell element formulations. These two shell time, Galbraith et al. [2] compared different shell element
element formulations do trade-off between computational time formulation of LSDYNA for the sheet metal forming process
and accuracy. For more accurate results, the BT shell element and found the strain distribution is more uniform in BT shell
formulation is better and for less computational time with good element than BT membrane element and suggested membrane
results, the IFB shell element is preferable. Finally, BT shell element cannot predict the through-thickness strain distribution.
element formulation has been chosen for FE Analysis of ISF Many researchers [3-7] developed and examined different
process in HyperWorks, since the results of thickness element formulations, which helped to improve the efficiency
distribution and profile prediction is in better agreement with and accuracy of FEA of the sheet metal forming process. The
the experimental results as well as the computational time is shell elements are frequently used in any sheet metal forming
less among the shell elements. process due to their robustness in elasto-plastic behavior. Due
to nonlinear stress distribution in sheet metal forming process,
INTRODUCTION the computational time is more in the case of FEA of ISF. To
Incremental sheet forming (ISF) is a flexible, rapid reduce the computational time, the reduced integration shell
manufacturing technique in which the freeform shapes can be elements (RISE) were mostly used [8]. The integration point on
formed without using any dedicated die. Numerically controlled the shell surface of RISE group of elements is only one, so the
continuous profiled tools are used, which induce localized stiffness of shell element is less in comparison to the fully
deformation of the sheet metal in incremental steps, and thus, integrated shell elements (FISE), due to which the elements
produces the desired geometry. The development of ISF deforms in the shape of hourglass. It is also called hourglass
stimulates small batch production and customized products. The effect or zero energy mode, which shows meaningless results.
ISF process offers many advantages, including better Many shell element formulations have been developed [11-13]
formability, reduced forming forces, and removal of dedicated to overcome the aforementioned phenomenon. In FEA of ISF,
dies compared to conventional forming process. The limitations the selection of shell element formulation is necessary to

1 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


identify which shell element type is more accurate and efficient
among the ones, covered as part of this study.
Although broad research on sheet metal forming by finite
element analysis has been conducted over the past decade, shell
element formulations of RADIOSS element library are not
properly investigated. In this work, six different shell element
formulations are categorized in two groups, namely, Reduced
Integrated Shell Elements (RISE) and Fully Integrated Shell
Elements (FISE). The RISE consists of Flanagan-Belytschko
(FB), Hallquist (H), Improved Flanagan-Belytschko (IFB),
Belytschko-Tsay (BT) and Zeng-Winkelmuller (ZW) shell (a) (b)
elements. The FISE group contains Batoz-Dhatt (BD) shell Fig. 2: Formed component (a) Frustum of cone (b) Mid cut
element. Aforementioned shell element formulations have been section for thickness measurement and profile prediction
chosen for investigation of ISF process due to their robustness
in elasto-plastic behavior, hourglass control and minimization of SHELL ELEMENT FORMUALTIONS
shear locking phenomenon as well as capability to analyze The degenerated shell element formulation theory [1] is
computational time and quality of results. Therefore, this work cost effective, versatile and has good applicability for thin and
investigates the role of four-noded shell element formulations to thick shells. In, FEA of sheet metal forming, some of the
perform accurate finite element modeling of the ISF process. In elements have no stiffness and deformed in a pattern that it
order to investigate the comparative advantages of each shell becomes anti-symmetric called hourglass mode or zero energy
element formulation, a frustum of a cone with 70 wall angle of mode, which shows meaningless results. There are two main
Al1050 S1B material is formed by ISF process and the hourglass modes happen in shell elements. First is a membrane
simulation results are compared with the experimental ones. mode (in-plane) and other one is W-mode (out-of-plane) [9].
Some of the researchers developed element formulations to
EXPERIMENTATION correct aforementioned phenomena. There are two methods to
For the validation of finite element model of incremental control the hourglass (1) Perturbation or penalty method:
sheet forming process, experiments are performed on OKUMA Deformation stability of an element is maintained by the
3-axis CNC vertical milling machine to make a frustum of cone. application of anti-hourglass forces to control the hourglass and
A circular fixture was used to clamp a circular sheet blank with it is used when the mesh quality is not so good, (2) Physical
a hardened stainless steel hemispherical end tool with straight stabilization method: it is used when the mesh quality is very
walls. The experimental setup is shown in fig. 1. Here , a blank good, to correct the internal energy by analytical way for
of 88 mm diameter and tool of 8 mm at 1500 mm/min feed rate hourglass control. In the standard 4-node shell element, six shell
with 0.5 mm increment depth (z) was used to fabricate the element formulations are available in RADIOSS version of 12.0
frustum of the cone with 70 mm major diameter (D) and 70 as listed in Tab. 1.
wall angle (). Hydraulic lubricating oil had been used to
minimize the friction between the tool and the sheet.
Aforementioned process parameters, based on literature, had Element types Short Integration point
been chosen for good surface finish. Circular grid pattern had form through
been sketched on the opposite side of the sheet blank to thickness
measure plastic strain. KISTLER force dynamometer had been Flanagan-Belytschko FB 1x7
used to measure the forming forces during experiments. Hallquist H 1x7
Improved Flanagan-Belytschko IFB 1x7
Belytschko-Tsay BT 1x7
Batoz-Dhatt BD 2x2x7
Zeng-Winkelmuller ZW 1x7
Tab. 1: Shell element formulations of RADIOSS

The integration point in the direction of through thickness may


be defined by the users for the above listed element type. In this
work, seven numbers of integration points in through thickness
direction were chosen and it can be increased up to 10 in
Fig. 1: Experimental setup for ISF process RADIOSS version of 12.0. The shell element formulations of

2 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


RADIOSS are categorized in two groups i.e. reduced integration (0.000551452 sec) was calculated by the ratio of total
and fully integrated shell element. simulation time to the number of steps. The mechanical
properties of Al1050-S1B aluminum alloy sheet were identified
FINITE ELEMENT MODELING by a uniaxial tensile test which are listed in Tab. 2. The material
The finite element modeling of incremental sheet forming properties of tool include density of 7895 kg/m3, Modulus of
process needs discretization of sheet blank and tool which is Elasticity 210 GPa and Poissons ratio 0.3.
shown in fig. 3. In ISF or any other sheet metal forming process Mesh model and tool path were same for FE analysis of the
the stress distribution is nonlinear due to the elasto-plastic frustum of cone, except shell element formulations. Process
nature of sheet, which leads to higher computational time. For parameters of ISF and finite element analysis procedure are
saving computational time the average element size of sheet shown in fig. 4 (a) and (b), where H is the height of frustum of
blank were kept 2 mm and for the forming tool 0.75 mm, for cone, z is incremental step depth, is wall angle, and r and R
proper contact and interface between the tool and sheet. The are minor and major radii of the frustum of cone respectively.
number of elements and nodes of sheet were 1668 and 1682 Tool path was generated from PowerMILL 9.0. It is desired to
respectively, while the number of elements and nodes of tool give the time steps in an explicit time integration manner
were 801 are 793 respectively. Nodes at outer periphery of sheet because of the incremental plastic deformation of sheet. The
were fixed and to make tool as a rigid body, the degrees-of- data of X v/s t; Y v/s t; and Z v/s t were exported as the desired
freedom of all tool nodes were constrained. Elasto-Plastic tool path in Hypermesh for the finite element modeling of
Johnson-Cook material model (M2_PLAS_JOHNS) had been incremental sheet forming process. Instead of uniform mesh, the
used to model the aluminum Al1050 sheet. The deformation of adaptive mesh has been chosen for finite element modeling of
tool is negligible as compared to the sheet. Therefore, elastic ISF process. In adaptive mesh, the mesh density at the critical
Hooks law (M1_ELAST) had been assigned to model the areas of the tool-sheet interface is higher than at that at non-
hemispherical end tool (hardened stainless steel) and for making critical areas of the tool-sheet interface. This is clearly evident
tool as a rigid body, the degrees of freedom of all the nodes of from fig. 3. At critical areas, fine meshes produces results with
the tool were restricted in all the directions. In addition, the tool high accuracy but at higher computational cost. Meanwhile, at
is assigned the correct mechanical properties. In the present non-critical areas of tool-sheet interface, a coarse mesh will take
work, the minimum gap for impact activation between the tool less computational time but may also produce less accurate
and the sheet were taken as 0.1 and coefficient of friction of the results. The fine and coarse meshes coupled together produces
lubricant used in the process was found to be at 0.3. Total results of acceptable quality and at acceptable computational
simulation time (2.05 sec) to form the component was cost. Besides, the smaller elements are much stiffer than the
calculated by the ratio of length of tool path trajectory larger elements. In uniform mesh, meshing with elements of only
(10245.97 mm) to the tool travel velocity (5000 mm/sec). one size is possible.

Parameters Symbols Values


Plasticity yield stress (MPa) ys 92.3
Plasticity hardening parameter (MPa) K 130.5
Strain hardening exponent n 0.1697
Density (kg/m2 ) 2705
Poisson's ratio v 0.3
Young Modulus (GPa) E 69
Coefficient of friction 0.33
Tab. 2: Mechanical properties Al 1050-O aluminium alloy sheet

ELASTO-PLASTIC MATERIAL MODEL BY JOHNSON-


COOK (PLASTICITY MODEL)
When the maximum stress developed during sheet metal
forming process is less than the yield stress, then the Johnson-
Cook material model behaves as linear elastic, but if more than
the yield stress then material behaves as nonlinear plastic. This
Fig. 3: Finite element model of tool and sheet for ISF material model is valid for truss, beam, brick and shell elements.
The Johnson-Cook material model describes the material
In order to reduce the total computational time, the tool behaviour under high-speed impact conditions. This model also
velocity had been increased to 5000 mm/sec. Time step

3 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


considers the effect of high temperature, large strains and high
strain rates.
The governing equation of stress distribution during plastic
deformation is given by [9]:
m
(a b pn ) 1 cln (1 T * ) (1)
0
where,
= Flow stress (Elastic + Plastic components)
p = Plastic strain (True strain)
a = Yield stress (a) (b)
b = Hardening modulus
n = Hardening exponent
c = Strain rate coefficient
= Strain rate
0
= Reference strain rate
m = Temperature exponent
T*= Homologous temperature (Dimensionless)
T 298
T* (2)
Tmelt 298
where,

T = Material temperature (K)


Tmelt = Melting temperature (K)
Assumed Troom = 298K [9], (c)
For the computation of temperature, the adiabatic conditions Fig. 4: (a) Process parameter of ISF for frustum of cone, (b)
have been assumed [9]: Spiral tool path for frustum of cone (c) Procedure for
E int finite element analysis of ISF process
T Ti (3)
Cp (Volume)
RESULTS AND DISCUSSION
where, In order to investigate the performances of different shell
element formulations for ISF process, some response parameters
Cp = Specific heat per unit of volume are analyzed with the experimental results like thickness
i = Initial temperature (K) variation, computational time and profile of component. Large
Eint = Internal energy difference has been found in the case of computational time
The internal energy per unit volume of the plastic deformation among reduced integration shell element (RISE) types and fully
process of thin structured can be written as [9]: integration shell element (FISE) type. Good thickness variations
t t are obtained in the RISE group as compared to the FISE group
2 2 and the biggest difference were seen in the case of FISE, when
Ei dz E 2dz the results are compared with the experimental results. However,
(4) while comparing the profile, the results obtained from RISE and
t t
FISE are approximately similar to the experimental results.
2 2
where,
COMPUTATIONAL TIME
The workstation Intel Xeon-2.4GHz-16GB RAM has been
Ei = Internal energy per unit volume (MLT-2)
used to perform the FE simulations of ISF process. The
E = Modulus of elasticity (N/mm2)
comparisons of computational time for different shell element
= Flow stress (N/mm2)
formulations are shown in Tab. 3. Processing time will increase
= Main component of strain tensor
with number of simulation simultaneously performed on a
t = Unitless quantity
single CPU.

4 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


Tab. 3 clearly shows that the BD (Batoz-Dhatt) shell THICKNESS VARIATION
element is more expensive, according to computational time as In any sheet metal forming process, the analysis of
it requires 34.35 CPU hours while IFB (Improved Flanagan- thickness distribution throughout the geometry is an important
Belytschko) and BT (Belytschko-Tsay) shell is quite cheap and task. Here, the contour plot of thickness variation of a frustum
fast and it requires 8.56 and 8.61 CPU hours respectively. The of cone formed by ISF process is shown with the help of fig. 6.
BD shell element comes under FISE group which is fully It has been observed that lowest minimum thickness is with BD
integrated with four gauss points (2x2) on shell surface and 7 shell element formulation and lowest maximum thickness is
integration points in through thickness direction (i.e. 2x2x7), with FB and H shell element formulations, but there is a need to
due to which the computational time is more as compared to observe the closeness of results with experimental value. The
RISE group (1x7 integration points through the thickness). One comparison of experimental and simulation results of thickness
of the reasons of higher computational time of FISE group may variation for the frustum of cone is shown in fig. 5.
be the output parameters, like forces and stresses that are
evaluated at every Gauss point (4 Gauss points) of shell element
and the results are presented after interpolation for an element.

Shell Element Total simulation Number of time


formulation time (h) steps
FB 8.76 3717
H 9.05 3717
IFB 8.56 3717
BT 8.61 3717
BD 34.35 3717
ZW 11.58 3717

Tab. 3: Comparison of computation time for different shell


element formulations
Fig. 5: Comparison of thickness variation with different shell
The computational time also depends upon the mesh density
element formulations and experimental results
of the sheet blank and forming tool. Aforementioned results are
obtained for the medium fine adaptive mesh with average
Due to necking, which leads to excessive thinning of
element size of sheet blank as 2 mm and that of forming tool as
formed wall thickness, the material gets accumulated at the
0.75 mm. In case of computational time, the BD shell element
bottom of the formed part and thickness at the bottom periphery
comes out as the most expensive and IFB shell element as the
of the cone increases. In fig. 6, when the fringe values ranges
least expensive among all shell element formulations considered
from 1 mm to 0.02 mm, it shows very large differences in the
here. However, it is required to check the accuracy of results in
colors, which is not easily seen. Hence, fringe ranges were
the context of other response parameters of ISF, like thickness
decided as per their minimum and maximum thickness results of
variation and profile of formed component.
FE model. Minimum value for fringes should be decided based
When simulations are performed on a single CPU at a time
on the minimum sheet thickness observed and the maximum
by using medium coarse adaptive meshing with 5 mm average
values were chosen based on the maximum sheet thickness
element size for the sheet blank with same integration points
observed in FEA simulation. Hence, fringe ranges are different
and 1 mm average element size for the forming tool. Then, result
for all shell element formulations. In fig. 6, it is clearly seen that
obtained by the BD shell element are even more expensive and
the maximum thinning was observed in the case of fully
it takes 15.25 CPU hours; while IFB shell elements takes the
integrated BD shell element and the rest of the shell elements
least i.e. 2.23 CPU hours. The computation efficiency of BT
like FB, H, IFB and ZW gives approximately similar results.
shell elements were affected and it takes 14.17 CPU hours since
When compared with the experimental results, the thickness
the simulation of FISE group was also performing together with
results of BT shell element are closer among all shell element
the BT shell element, which is very costly in terms of
formulations. In FEA of sheet metal forming, accurate
computational time. After comparison of results with the
prediction of thickness with the experimental results is quite
experimental results, BT shell element gives optimum results in
challenging. From fig. 5, it can be clearly observed that the
terms of profile prediction and thickness variation, which are in
prediction of thickness values for all the shell element
good agreement.
formulations did not exactly follow the experimental results.

5 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


(FB) (H)

(IFB) (BT)

(BD) (ZW)

Fig. 6: Contour plot of thickness distribution of different shell element formulation

However, the BT shell element gives more accurate results as The experimental and FEA results of profile prediction are
compared to the experimental result than the other shell element not matched properly at the bottom edge of the formed
formulations. In addition, the quality of results was improved component as the shell elements are slightly large in size, and
when the quadrilateral elements were used for Batoz Dhatt fully hence proper corners are not achieved. The results of profile
integrated shell element [9]. However, in the present work, prediction of the formed component are almost same for all the
combinations of quadrilateral and triangular mesh have been shell element formulations and the experimental results, but
used as a consequence of adaptive meshing. Since, the plasticity they do not match with the CAD geometry.
is computed over the integration points; this may lead to the bad
performance of BD shell element formulation.

PROFILE PREDICTION
The shape of the formed component is also considered as a
part of geometrical accuracy. From fig. 7, it is clearly seen that a
little bending was found at the clamping region at outer
periphery of the formed component. Bending occurs as there is
no support between outer periphery of the cone and clamped
region, which may be avoided by using backing plate. The wall
of the formed component, both experimentally and by FEA
simulation slightly deviates from the CAD geometry due to Fig. 7: Comparison of profile of frustum of cone among
springback and elastic properties of the material. Further, the different shell element formulations, CAD geometry and
desire depth is not achieved due to the pillow effect of the experimental results
material at the bottom of the formed component.

6 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo


CONCLUSIONS [5]. Henrard, C., Bouffioux, C., Eyckens, P., Sol, H., Duflou, J.
In this paper, the finite element analysis and experimental R., Van Houtte, P., & Habraken, A. M., 2011, Forming
analysis of ISF process have been investigated. The shell forces in single point incremental forming: prediction by
element formulations of RADIOSS element library have been finite element simulations, validation and sensitivity,
investigated for ISF process in terms of computational time, Computational mechanics, 47(5), pp. 573-590.
thickness distribution as well as profile prediction. For medium
fine adaptive meshing, the BT shell element is the most [6]. Cescotto S, Grober H., 1985, Calibration and application
desirable for the FE analysis of ISF process, as very good of an elastic viscoplastic constitutive equation for steels in
results have been found in terms of computational time and hot-rolling conditions, Engineering Computation, 2(2),
thickness distribution. For medium coarse adaptive meshing, the pp. 101106.
BT and IFB shell elements make a tradeoff between
computational cost and accuracy of results. If we need better [7]. Sena, J. I. V., 2011, "Finite element analysis of
results in lesser time, then the IFB shell element is more incrementally formed parts." International Journal of
desirable otherwise, the BT shell element has to be preferred for Mechatronics and Manufacturing Systems, 4(5), pp. 454-
the finite element analysis of ISF process. 469.
Present investigation was made primarily to study the trend
of thickness distribution, profile of component and [8]. De Sena, J. I. V., Guzmn, C. F., Duchne, L., Habraken, A.
computational time. The study was qualitative in nature and M., Behera, A. K., Duflou, J., de Sousa, R. A., 2015,
producing exact quantitative results was considered beyond the Simulation of a two-slope pyramid made by SPIF using an
scope of the present work. An investigation to produce highly adaptive remeshing method with solid-shell finite element.
accurate quantitative result would be made as part of future International Journal of Material Forming, pp. 1-12.
endeavors.
[9]. RADIOSS, 2012, Theoretical Manual, Version 12.0,
ACKNOWLEDGEMENT Block 110 Format.
Authors gratefully acknowledge the Board of Research in
Nuclear Sciences (BRNS), India for their support and funding [10]. LS-DYNA, 1998, Theoretical Manual. Livemore
for this research work. Software Technology Corporation, Livemore,.

REFERENCES [11]. Kosloff, D., & Frazier, G. A., 1978, Treatment of


[1]. Ahmad, Sohrabuddin, Bruce M. Irons, and O. C. hourglass patterns in low order finite element codes,
Zienkiewicz, 1970, "Analysis of thick and thin shell International journal for numerical and analytical methods
structures by curved finite elements." International Journal in geomechanics, 2(1), pp. 57-72.
for Numerical Methods in Engineering, 2(3), pp. 419-451.
[12]. Flanagan, D. P., & Belytschko, T., 1981, A uniform strain
[2]. Galbraith, P. C., and J. O. Hallquist, 1995, "Shell-element hexahedron and quadrilateral with orthogonal hourglass
formulations in LS-DYNA3D: their use in the modelling of control, International Journal for Numerical Methods in
sheet-metal forming." Journal of Materials Processing Engineering, 17(5), pp. 679-706.
Technology, 50, pp. 158-167.
[13]. Belytschko, T., & Tsay, C. S., 1983, A stabilization
[3]. Ambrogio, G., Costantino, I., De Napoli, L., Filice, L., procedure for the quadrilateral plate element with
Fratini, L., & Muzzupappa, M., 2004, Influence of some onepoint quadrature, International Journal for Numerical
relevant process parameters on the dimensional accuracy in Methods in Engineering, 19(3), pp. 405-419.
incremental forming: a numerical and experimental
investigation Journal of Materials Processing Technology,
153, pp. 501-507.

[4]. Firat, M., 2007, Computer aided analysis and design of


sheet metal forming processes: Part IThe finite element
modeling concepts, Materials & design, 28(4), pp. 1298-
1303.

7 Copyright 2015 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/86878/ on 05/18/2017 Terms of Use: http://www.asme.org/abo

Вам также может понравиться