SlideShare une entreprise Scribd logo
1  sur  10
Télécharger pour lire hors ligne
Journal of Materials Processing Technology 190 (2007) 189–198




                                Simulation of springback variation in forming
                                      of advanced high strength steels
                                                       Peng Chen b , Muammer Koc a,∗
                                                                               ¸
                                 a   NSF I/UCR Center for Precision Forming (CPF) and Department of Mechanical Engineering,
                                                 Virginia Commonwealth University (VCU), Richmond, VA, USA
                                       b Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA

                                               Received in revised form 23 February 2007; accepted 27 February 2007



Abstract
   Variations in the mechanical and dimensional properties of the incoming material, lubrication and other forming process parameters are the
main causes of springback variation. Variation of springback prevents the applicability of the springback prediction and compensation techniques.
Hence, it leads to amplified variations and problems during assembly of the stamped components, in turn, resulting in quality issues. To predict the
variation of springback and to improve the robustness of the forming process, variation simulation analysis could be adopted in the early design
stage. Design of experiment (DOE) and finite element analysis (FEA) approach was used for the variation simulation and analysis of the springback
for advanced high strength steel (AHSS) parts. To avoid the issues caused by the deterministic FEA simulation, random number generation was
used to introduce uncertainties in DOE. This approach was, then, applied to investigate the effects of variations in material, blank holder force and
friction on the springback variation for an open-channel shaped part made of dual phase (DP) steel. This approach provides a rapid and accurate
understanding of the influence of the random process variations on the springback variation of the formed part using FEA techniques eliminating
the need for lengthy and costly physical experiments.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Springback variation; Variation simulation; Design of experiment; Finite element analysis; Computer experiment; Advanced high strength steel




1. Introduction                                                                    due to challenges in terms of formability, springback, joining
                                                                                   and die life issues.
    During the past two decades, the automotive industry has                          Springback is an elastically driven change in shape and form
been in a quest to achieve low-mass vehicles due to government                     of a part upon unloading after a part is formed. The concerns
regulations and consumer demands for light, fuel-efficient and                      about springback and quality control grow among auto-makers
low-emission vehicles. In addition, requirements for improved                      as the use of high strength–low weight materials increases [1].
safety, performance, and comfort features continue putting                         Parts made of AHSS demonstrate more springback than parts
pressure on the automotive manufacturers especially in an envi-                    made of mild steel do. Moreover, the experience with forming
ronment where there are ever-increasing global competition                         of AHSS materials is limited compared to mild steel. Variation
and continuous cost reduction demands. As an alternative to                        of springback prevents the applicability and use of springback
aluminum and magnesium alloys, advanced/ultra high strength                        prediction and compensation techniques. Hence, it leads to
steels (AHSS) have been proven to be a proper choice of mate-                      amplified variations and problems during assembly of compo-
rial for various body and structural automotive parts meeting the                  nents, and in turn, results in quality issues. Thus, understanding,
required low-mass, affordable cost and increased performance                       accurate characterization, prediction, control and reduction of
requirements. However, wide application of high strength steel                     springback and its variation have become very crucial in terms
in many potential auto body and structural parts is still limited                  of decreasing development times and reducing scrap rate in mass
                                                                                   production to achieve cost effective fabrication of AHSS parts.
                                                                                      Numerous studies during the last 40 years have attempted to
 ∗   Corresponding author. Tel.: +1 804 827 7029.                                  determine the controlling factors in springback and find ways to
     E-mail address: mkoc@vcu.edu (M. Koc). ¸                                      reduce it. An early study by Baba and Tozawa [2] focused on

0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.02.046
190                              P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                               c

the effect of stretching a sheet by a tensile force, during or after       simulation, which is especially designed for the assembly pro-
bending, in minimizing springback. Other studies investigated              cess (welding) simulation based on the assumption of linear
the role of process variables on springback. Zhang and Lee [3]             elasticity.
showed the influence of blank holder force, elastic modulus,                    Generally speaking, Monte-Carlo simulation is effective for
strain hardening exponent, blank thickness and yield strength              any kind of non-linear process. However, since large number of
on the magnitude of the final springback strain in a part. Geng             FEA simulations is required (more than 100 for reliable results),
and Wagoner [4] studied the effects of plastic anisotropy and its          it is not practical and too costly to apply it for a complicated
evolution in springback. They developed a constitutive equation            case, such as metal forming. Therefore, some researchers [22]
for 6022-T4 aluminum alloy using a new anisotropic hardening               applied DOE techniques in their FE simulation to effectively
model and proved that Barlat’s yield function is more accurate             reduce the amount of simulations. However, since the FEA sim-
than other yield functions in their case.                                  ulation is deterministic (no random error), replicate observations
    Some researchers evaluated FE simulation procedures in                 from running the simulation with the same inputs will be identi-
terms of their springback prediction accuracy. Mattiasson et               cal. Despite some similarities to physical experiments, the lack
al. [5], Wagoner et al. [6], Li et al. [7] and Lee and Yang [8]            of random error makes FEA simulations different from physi-
found that FEM simulations of springback are much more sen-                cal experiments. In the absence of independent random errors,
sitive to numerical tolerances than forming simulations are. Li            the rationale for least-squares fitting of a response surface is not
et al. [9] investigated the effects of element type on the spring-         clear [23]. The usual measures of uncertainty derived from least-
back simulation. Yuen [10] and Tang [11] found that different              squares residuals have no obvious statistical meaning [23,24].
unloading scheme will affect the accuracy of the springback                According to Welch et al. [25], in the presence of systematic
prediction. Similarly, Focellese et al. [12] and Narasimhan and            error rather than random error, statistical testing is inappropri-
Lovell [13] pointed out that different integration scheme will             ate. For deterministic FEA simulation, some statistics, including
also influence the result of springback simulation. In 1999,                F-statistics, have no statistical meaning since they assume the
Park et al. [14] and Valente and Traversa [15] attempted to                observations include an error term which has mean of zero and
link dynamic explicit simulations of forming operations to static          a non-zero standard deviation. Consequently, the use of step-
implicit simulations of springback. It was proved that this tech-          wise regression for polynomial model fitting is not appropriate
nique is very effective for the springback simulation. Li et               since it utilizes F-statistic value when adding/removing model
al. [16] explored a variety of issues in the springback simu-              parameters [24].
lations. They concluded that (1) typical forming simulations                   In an effort to solve the above problem, McKay et al. [26]
are acceptably accurate with 5–9 through-thickness integration             introduced Latin hypercube sampling which ensures that each
points for shell/beam type elements, whereas springback anal-              of the input variables has all portions of its range presented.
ysis within 1% numerical error requires up to 51 points, and               Sacks et al. [23,27] proposed the design and analysis of com-
more typically 15–25 points, depending on R/t, sheet tension               puter (DACE) method to model the deterministic output as the
and friction coefficient. (2) More contact nodes are necessary              realization of a stochastic process, thereby providing a statis-
for accurate springback simulations than for forming simulation,           tical basis for designing experiments for efficient prediction.
approximately one node per 5◦ of turn angle versus 10◦ recom-              Kleijnen [28] suggested incorporating substantial random error
mended for forming. (3) Three-dimensional shell and non-linear             through random number generators. Therefore, it is natural to
solid elements are preferred for springback prediction even for            design and analyze such stochastic simulation experiments using
large w/t ratios because of the presence of persistent anticlas-           standard techniques for physical experiments. Some researchers
tic curvature. For R/t > 5.6, shell elements are preferred since           (e.g., Giunta et al. [29,30] and Venter et al. [31]), have also
solid elements are too computation-intensive. For R/t < 5.6, non-          employed metamodeling techniques such as RSM for modeling
linear 3D solid elements are required for accurate springback              deterministic computer experiments which contain numerical
prediction.                                                                noise. This numerical noise is used as a surrogate for random
    Most of the research efforts on springback focused on the              error, thus allowing the standard least-squares approach to be
accurate prediction and compensation of springback. The issue              applied. However, the assumption of equating numerical noise
of springback variation was seldom concerned. Moreover, none               to random error is questionable.
of the studies in the area of springback prediction touched                    In this work, random number generator was used to ensure
upon the variation simulation of springback. As lead times                 the correctness of the regression model. Assuming a Gaussian
are shortened and materials of high strength–low weight are                distribution, the uncertainty was introduced by random number
used in manufacturing, a fundamental understanding of the                  generation for controlled factors at different level and uncon-
springback variation has become essential for accurate and                 trolled variables according to their mean and range. By using
rapid design of tooling and processes in the early design stage.           this method, the effects of variations in material (mechanical
As far as the variation simulation is concerned, most avail-               properties) and process (blank holder force and friction) on
able references are related to the assembly processes [17–21].             the springback variation were investigated for an open-channel
In 1997, Liu and Hu [17] summarized two variation simula-                  shaped part made of DP steel. The variations in stamping process
tion approaches for sheet metal assembly. They are: (1) direct             are introduced in Section 2 of the paper. In Section 3, measure-
Monte-Carlo simulation, which is popular and good for all kinds            ments of springback are defined. Section 4 is the validation of the
of random process simulation, and (2) mechanistic variation                FE model using existing experimental results in the literature.
P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                       c                                                                                               191




                                                         Fig. 1. Springback variation and its sources.


In the following section, variation analysis of an open-channel                   • Batch-to-batch variation represents the variability among the
part made of advanced high strength steel is presented. The last                    individual batches, which is mostly caused by material vari-
section is discussions and conclusions.                                             ation from batch to batch and the variation introduced by
                                                                                    tooling setup.
2. Variations in stamping process
                                                                                  3. Measurement of springback for open-channel
   In this study, the objective is to understand and accurately                   drawing
predict the variation of springback in an open-channel drawing
considering the variations of material and process as shown in                        A schematic view of die, punch, blank and their dimensions
Fig. 1.                                                                           for open-channel drawing, which is used in the analyses for
   Total variation of springback in the stamping process has                      this study, is shown in Fig. 3. Fig. 4 shows the formed part after
several components. Generally, different variation components                     springback. Three measurements, namely the springback of wall
can be attributed to different sources [32]. The following are the                opening angle (β1 ), the springback of flange angle (β2 ) and side-
major categories of variation source (Fig. 2):                                    wall curl radius (ρ) shown in Fig. 5, were used to characterize the
                                                                                  total springback considering only the cross-sectional shapes of
                                                                                  formed parts obtained before and after the removal of tools. The
• Part-to-part variation is also referred to as system-level vari-
                                                                                  springback in the direction orthogonal to the cross-section, such
  ation or inherent variation. It is the amount of variation that
                                                                                  as twisting, was not considered since it is negligible is this case.
  can be expected across consecutive parts produced by the pro-
                                                                                  As there is no clear distinction to separate a cross-section curve
  cess during a given run. It is caused by the random variation
                                                                                  for individual measurement of springback angles and sidewall
  of all the uncontrolled (controllable and uncontrollable) pro-
                                                                                  curl, two assumptions deduced from the sample observations
  cess variables. In the variation simulation of this paper, blank
  thickness was considered as the uncontrolled variable.
• Within batch variation is usually due to the variations of the
  controlled variables such as BHF, material property and fric-
  tion.




                                                                                  Fig. 3. A schematic view of tools and dimensions for open-channel drawing
         Fig. 2. Source of variation in a typical forming process.                [32].
192                                 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                  c

                                                                                   to construct a circular arc is used. Eq. (1) lists all the equations
                                                                                   needed for the calculation of the β1 , β2 and ρ.

                                                                                                        ox · A0 B0
                                                                                   θ1 = arccos
                                                                                    0
                                                                                                      |ox| · A0 B0
                                                                                                        ox · A0 B0
                                                                                   θ2 = arccos
                                                                                    0
                                                                                                      |ox| · A0 B0

                Fig. 4. Open-channel parts after drawing.                                               ox · AB
                                                                                   θ1 = arccos
                                                                                                      |ox| · |AB|
                                                                                                       AB · ED
                                                                                   θ2 = arccos
                                                                                                      |ED| · |AB|                                  (1)
                                                                                   β1 = θ1 − θ1
                                                                                              0

                                                                                   β2 = θ2 − θ2
                                                                                         0

                                                                                                    xB + yB − xA − yA − ((yA − yB )/
                                                                                                     2    2    2    2

                                                                                                    (yC − yB ))(xC + yC − xB − yB )
                                                                                                                 2    2    2     2
                                                                                   xO =
                                                                                            2 xB − xA + (xC − xB )((yA − yB )/(yC − yB ))
                                                                                            xA + yA − xB − yB + 2xO (xB − xA )
                                                                                             2    2    2    2
                                                                                   yO =
                                                                                                       2(yA − yB )
                                                                                   ρ=        (xA − xO )2 + (yA − yO )2

                                                                                   4. Finite element modeling and validation for the
                   Fig. 5. Illustration of springbacks.
                                                                                   open-channel drawing of AHSS

are introduced for the springback measurement. Firstly, it is                         The simulation work for this study is based on the exper-
assumed that wall opening angle, flange closing angles and side-                    imental results of Lee et al. [33]. Information about the
wall curl vary independently. Secondly, the sidewall curl could                    geometry and dimensions of the tooling and blank are pre-
be approximated by a piece of circular arc.                                        sented in Fig. 3. The initial dimension of the blank sheet was
    Fig. 5 also shows the measurements placements (A–E). Two                       300 mm (length) × 35 mm (width). Forming was carried out on
measurements were conducted before springback, namely the x                        a 150 tonnes double action hydraulic press with a punch speed
and y coordinates of A and B, which is denoted as A0 and B0                        of 1 mm/s, and the total punch stroke was 70 mm. Blank holder
in this work. They are used to compute the wall angle (θ1 ) and0                   force (BHF) was 2.5 kN. The blank material used was DP Steel
flange angle (θ2  0 ) before springback. After springback, another                  with the material properties presented in Fig. 6 based on the
five measurements were placed on A–E, which were used in the                        tensile tests by Lee et al. [34].
calculation of the wall angle (θ 1 ), flange angle (θ 2 ) and sidewall                 Considering the geometric symmetry of the process, only
curl radius (ρ) after springback. To estimate the sidewall curl                    half of the blank was simulated. The material was modeled as
radius, a curve fitting technique that employs three points (A–C)                   an elastic–plastic material with isotropic elasticity, using the




                                                          Fig. 6. Material properties of DP steel [33].
P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                        c                                                                                                 193

Table 1
Different FEA procedure used in simulation
                                          Case

                                          1             2            3           4            5              6              7          8              9

Element type                              Solid         Solid        Solid       Solid        Shell          Shell          Shell      Shell          Shell
Contact                                   Soft          Soft         Soft        Hard         Soft           Soft           Hard       Hard           Soft
Forming analysis (dynamic)                Implicit      Implicit     Implicit    Implicit     Implicit       Implicit       Explicit   Explicit       Explicit
Springback analysis (static)              Implicit      Implicit     Implicit    Implicit     Implicit       Implicit       Implicit   Implicit       Implicit
Through-thickness element number          5             9            21          9            9              21             5          15             9
  or integration point



Hill anisotropic yield criterion for the plasticity. The coefficients            in Table 1) of different element type, different contact condi-
of Hill yield criterion (R11 = 1.0, R22 = 1.01951, R33 = 1.00219,               tion, different through-thickness element number and different
R12 = 0.992318, R13 = 1.0, R23 = 1.0) were computed from the                    analysis type were tried in the simulation. The term, soft con-
r-values as presented in Fig. 6. The friction coefficient between                tact, denotes exponential pressure-overclosure definition for the
tools and the sheet blank was assumed to be constant and                        normal behavior between contacting surfaces.
0.1. To determine the appropriate element type, contact con-                       The comparison of different simulation procedure for the
ditions, through-thickness element number and analysis type                     prediction of wall opening angle (θ 1 ), flange angle (θ 2 ) and side-
for simulation using ABAQUS, nine combinations (as tabulated                    wall curl radius (ρ) is shown in Figs. 7–9. It was found that the
                                                                                sidewall curl is very sensitive to the contact condition used in
                                                                                simulation. Since the soft contact tends to soften the contact-
                                                                                ing surface, it actually depresses the sidewall curl, which is not
                                                                                true for advanced high strength steel. Among these combina-
                                                                                tions, case 4 (hard contact), case 7 (hard contact) and case 8
                                                                                (hard contact) show a good match with the experiment results
                                                                                in all three springback measurements. Hence, hard contact is

                                                                                Table 2
                                                                                Original experiment design
                                                                                Run order           BHF (kN)            Friction           Material
Fig. 7. Effects of different FEA procedure on the prediction of wall opening
angle.                                                                           1                  13.75               0.15               1.1
                                                                                 2                  13.75               0.1                1
                                                                                 3                  13.75               0.15               0.9
                                                                                 4                   2.5                0.15               1
                                                                                 5                  25                  0.1                1.1
                                                                                 6                   2.5                0.1                1.1
                                                                                 7                  25                  0.15               1
                                                                                 8                   2.5                0.15               1
                                                                                 9                   2.5                0.1                1.1
                                                                                10                  13.75               0.05               0.9
                                                                                11                  25                  0.05               1
                                                                                12                   2.5                0.1                0.9
                                                                                13                   2.5                0.05               1
                                                                                14                  13.75               0.1                1
Fig. 8. Effects of different FEA procedure on the prediction of flange closing   15                  13.75               0.1                1
angle.                                                                          16                  13.75               0.05               0.9
                                                                                17                   2.5                0.1                0.9
                                                                                18                  13.75               0.1                1
                                                                                19                  25                  0.05               1
                                                                                20                  13.75               0.15               0.9
                                                                                21                  25                  0.1                0.9
                                                                                22                  13.75               0.05               1.1
                                                                                23                  13.75               0.1                1
                                                                                24                  25                  0.15               1
                                                                                25                  13.75               0.15               1.1
                                                                                26                  25                  0.1                0.9
                                                                                27                  13.75               0.05               1.1
                                                                                28                  13.75               0.1                1
                                                                                29                   2.5                0.05               1
Fig. 9. Effects of different FEA procedure on the prediction of sidewall curl   30                  25                  0.1                1.1
radius.
194                                        P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                         c

Table 3                                                                              5. Variation simulation of springback and results
Assumed statistics of variables
                            Mean                Range                     S.D.          In this study, we only considered the “part-to-part” and
Uncontrolled factor
                                                                                     “within batch” variations. The variation simulation and anal-
  Part thickness                 1.2 mm           1.18–1.22 mm            0.066667   ysis of the springback of DP steel part are described step by step
                                                                                     as follows.
Controlled factor
  BHF level-1                    2.5 kN            2.4–2.6                0.033333
  BHF level-2                   13.75 kN         13.65–13.85              0.033333   Step 1 (Design of experiment). BHF, material property and fric-
  BHF level-3                   25 kN             24.9–25.1               0.033333   tion were chosen as design factors. Box–Behnken RSM design
  Friction level-1               0.05             0.04–0.06               0.003333   with 2-replicate and 6-center-point was used for this 3-factor and
  Friction level-2               0.1              0.09–0.11               0.003333   3-level experiment design. The levels of the material property
  Friction level-3               0.15             0.14–0.16               0.003333   are considered as 110, 100 and 90% of the stress–strain curve
  Material level-1           90%                    88–92%                0.00667    in Fig. 6, which indicates the strength of the material. Table 2
  Material level-2          100%                    98–102%               0.00667    shows the original experiment table.
  Material level-3          110%                   108–112%               0.00667
                                                                                     Steps 2 and 3Random number generation of controlled and
                                                                                     uncontrolled variablesIt is assumed that most random processes
preferred. It can be seen that element type and forming anal-                        conform to a Gaussian distribution. Moreover, irrespective of
ysis type do not affect the accuracy of springback prediction                        the parent distribution of the population, the distribution of the
much. Therefore, to reduce the computation time, hard con-                           average of random samples taken from the population tends to
tact, shell element, explicit (dynamic) for forming and implicit                     be normal as the sample size increases (Central Limit Theo-
(static) for springback were used in further simulations. Dif-                       rem). Therefore, once we know the mean and standard deviation
ferent through-thickness integration points (5, 9, 15, and 21) of                    (S.D.) of a random process, we can generate a random number
shell elements were also tried in the simulation, which showed no                    according to its Gaussian distribution. According to the statistics
much influence on the prediction of springback. Therefore, nine                       chosen in Table 3, the original experiment table was random-
through-thickness integration points were used in the further                        ized as shown in Table 4. Fig. 10 is an illustration of the number
simulations.                                                                         randomization.

Table 4
Randomized (random number generation) experiment table and simulation results
Run                 BHF (kN)                Friction             Material            Part thickness           β1 (◦ )        β2 (◦ )           ρ (mm)

 1                  13.649243               0.1405               1.0819              11.8017                  16.1242        12.3581           191.5870
 2                  13.755335               0.0904               0.9814              11.9484                  16.5589        12.1747           169.7480
 3                  13.721143               0.1534               0.8811              11.9619                  12.5694         9.4674           301.4050
 4                   2.3992                 0.1512               0.9933              11.9606                  18.4420        11.0999           133.4730
 5                  24.9033                 0.0970               1.0923              11.9473                  19.8823        13.4837           137.7005
 6                   2.5053                 0.1005               1.0913              12.0388                  17.8932        11.0234           138.2284
 7                  25.0027                 0.1466               1.0063              12.0067                  17.0396        11.4995           158.0187
 8                   2.4711                 0.1521               0.9910              12.0383                  17.5935        10.8937           140.5959
 9                   2.5291                 0.1016               1.0999              12.0638                  15.0679        11.1937           140.6841
10                  13.779106               0.0404               0.8937              11.9781                  14.1995        11.0920           156.6561
11                  24.9970                 0.0536               1.0160              11.9351                  17.7311        12.9760           151.3312
12                   2.5072                 0.0958               0.8941              12.0725                   9.8603        10.0025           178.0388
13                   2.4983                 0.0422               0.9933              11.8973                  14.9086         9.9908           159.4871
14                  13.757157               0.0995               1.0031              12.0039                  16.4520        12.2300           169.0799
15                  13.748318               0.0959               1.0025              12.0272                  16.6234        12.2783           166.0132
16                  13.737183               0.0555               0.9081              11.9032                  15.3990        11.3933           166.4367
17                   2.4872                 0.0970               0.8889              12.0405                  14.7347         9.3532           168.8445
18                  13.791958               0.1035               1.0077              11.9982                  16.3857        12.2430           171.5795
19                  24.9737                 0.0527               0.9922              11.9569                  17.2571        12.6146           156.6909
20                  13.780872               0.1515               0.9056              12.0172                  12.6997         9.7055           298.5300
21                  24.9978                 0.0977               0.8982              12.0322                  15.9928        11.2144           172.0237
22                  13.772126               0.0484               1.1079              11.9815                  18.5080        13.6613           122.9733
23                  13.718735               0.0964               0.9919              11.8797                  16.5546        12.2819           169.2909
24                  24.9718                 0.1483               1.0078              11.9887                  17.0759        11.5550           158.1165
25                  13.785849               0.1407               1.1077              12.0855                  17.0869        13.9510           178.8833
26                  25.0469                 0.1037               0.9091              11.8921                  16.4418        11.4263           169.8623
27                  13.768493               0.0428               1.0955              12.0642                  18.7483        14.9717           125.4564
28                  13.751131               0.0960               0.9936              12.0129                  16.4303        12.1832           169.6901
29                   2.542                  0.0480               1.0014              11.9466                  15.2362        10.1161           157.6620
30                  24.9880                 0.1004               1.0918              12.0695                  18.7641        14.3375           139.4873
P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                        c                                                                                                 195

                                                                                Table 7
                                                                                Recommended variable level for the minimum variance of β1 , β2 and ρ according
                                                                                to Monte-Carlo simulation
                                                                                                           Material              Friction             BHF

                                                                                Min[Var(β1 )]              –                     High                 Middle
                                                                                Min[Var(β2 )]              –                     High                 Low
                                                                                Min[Var(ρ)]                Middle                Middle               –



                                                                                equations. For instance, the coefficient of the XMaterial is +1.5867
                                                                                in Eq. (2), therefore, the bigger the material strength, the larger
                                                                                the springback. Table 5 tabulates the optimal variable level for
                                                                                the minimum of each springback. A more detailed indication
Fig. 10. Schematic diagram of the variable randomizations (random number        of the relationship between the springbacks and the factors is
generation).                                                                    shown in Fig. 11. As in the parameter’s range studied in this
                                                                                work, springback increases with BHF and friction, which agrees
Table 5                                                                         with the experimental observations [35]. Papeleux and Pon-
Recommended variable level for the minimum of β1 , β2 and maximum of ρ          thot [35] reported that springback increases with small BHF,
                           Material              Friction              BHF      but decreases as the BHF increases for large force values.
Minimum β1                 Low                   Low                   Low
                                                                                This phenomenon can be explained by the fact that with low
Minimum β2                 Low                   Low                   Low      BHF, the punch induces mostly bending stresses in the mate-
Maximum ρ                  Low                   Low                   –        rial, but as the blankholder holds the blank more severely, the
                                                                                stresses included by the punching phase become mostly tensile
                                                                                stresses.
Step 4 (Simulation). Simulations were run according to Table 4
and the simulation results are shown in Table 4 as well.                        Step 6 (Variation sensitivity analysis). Three methods were
                                                                                used to analyze the effects of the factors on the variation of
Step 5 (Regression analysis). Eqs. (2)–(4) are the regression                   the springback. Finally, it was found that the springback varia-
models of β1 , β2 , and ρ as functions of BHF, material and fric-               tion magnitude is too small in this case and not distinguishable
tion. The variables in these equations are coded (−1, 0, 1) factors             from the system noise.
in the DOE. To investigate whether the factors’ effect on each
springback is significant, analysis of variance (ANOVA) was
used. Factors with a P-value larger than 0.05 were considered as                5.1. Monte-Carlo simulation
insignificant and ignored in the regression model. For example,
                                                                                    Monte-Carlo simulation was applied to Eqs. (2)–(4). Accord-
the main effect of friction on β1 is negligible:
                                                                                ing to the parameter levels used in the DOE, it is assumed
β1 = 16.5008 + 1.5867XMaterial + 0.7287XBHF                                     that all factors have equal variance in the Monte-Carlo simula-
                                                                                tion, i.e., Var(XMaterial ) = Var(XBHF ) = Var(XFriction ) = 0.32 , with
       −0.8454XBHF XFriction                                           (2)      a zero mean value for each factor (coded factors). Monte-Carlo
                                                                                simulation was run 100 times for each situation (a specific factor
β2 = 12.2319 + 1.3329XMaterial + 0.9646XBHF                                     at a specific level). The corresponding variance of the springback
                                                                                was recorded in Table 6. Table 7 summarizes the optimal variable
       −0.3929XFriction − 0.7297XBHF − 0.5529XBHF XFriction
                                 2
                                                                                level for the minimum variance of each springback according to
                                                                       (3)      Table 6.

r = 169.234 − 8.384XMaterial + 18.482XMaterial XFriction               (4)      5.2. Sensitivity analysis

  The effect of each factor on each springback could be deter-                     According to Eqs. (2)–(4), the variance of β1 , β2 and ρ are
mined by the sign of the corresponding coefficient in the above                  expressed as Eqs. (5)–(7) via linearized sensitivity analysis.

Table 6
Springback variation
                Material                                         Friction                                           BHF

                Low              Middle          High            Low              Middle         High               Low             Middle           High

Var(β1 )         0.0667          0.0667           0.0667          0.5349          0.324           0.2563             0.3501          0.2571           0.3207
Var(β2 )         0.1146          0.1146           0.1146          0.4798          0.3276          0.2367             0.1866          0.1926           0.2656
Var(ρ)          37.419           0               37.419          73.712           7.1786         10.414             11.813          11.813           11.813
196                             P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                              c




                                             Fig. 11. Response surface plots of (a) β1 , (b) β2 and (c) ρ.

                                                                              Var(r) = [−27.3 − 19.35XFriction + 27.194XMaterial ]2
Var(β1 ) = [1.8861 − 1.6648XMaterial ]2 Var(XMaterial )
                                                                                          ×Var(XMaterial ) + [−54.444XBHF ]2 Var(XBHF )
           +[1.028 − 0.8454XFriction ]2 Var(XBHF )
                                                                                          +[22.745 − 19.35XMaterial + 19.82XFriction ]2
           +[0.8454XBHF ]2 Var(XFriction )                          (5)
                                                                                          ×Var(XFriction )                                  (7)
Var(β2 ) = [1.3329 + 0.2899XBHF ]2 Var(XMaterial )
                                                                                 The variance of the response is determined by the variance
           +[0.9646 − 1.4594XBHF + 0.2899XMaterial ]2                         of each factor and the sensitivity coefficient (the quantity in the
                                                                              square parentheses). To minimize the variance of the springback,
           ×Var(XBHF ) + [−0.3929 − 0.5528XBHF ]2
                                                                              the most efficient way is to minimize the sensitivity coeffi-
           ×Var(XFriction )                                         (6)       cients in the equation. Table 8 tabulates the optimal variable
P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                               c                                                                                                  197

Table 8                                                                                (Tables 7 and 8) actually do not have any meaning because
Recommended variable level for the minimum variance of β1 , β2 and ρ according         the springback variations are totally random and uncontrollable
to sensitivity analysis
                                                                                       in this case. In other words, conclusions from both methods
                             Material                  Friction            BHF         are neither correct nor wrong. Since the system-level noises
Min[Var(β1 )]                High                      High                Middle      were introduced by random number generation (Table 3) in our
Min[Var(β2 )]                High                      Middle              High        computer experiment, we can solve the problem by reducing
Min[Var(ρ)]                  High                      Low                 Middle      the standard deviations used in the random number generation.
                                                                                       However, this kind of adjustment would not be easy in reality,
                                                                                       since the tuning of the system-level noise is usually impossible
Table 9
Extracted data (β1 ) used in MINITAB for Taguchi analysis                              in most cases.
BHF (kN)          Friction            Material          S.D. (β1 )      Mean (β1 )
                                                                                       6. Conclusions
13.75             0.15                1.1               0.68073         16.6056
13.75             0.1                 1                 0.09154         16.5008           The effects of BHF, material and friction on springback
13.75             0.15                0.9               0.09214         12.6346        and springback variation of DP steel channel have been ana-
 2.5              0.15                1                 0.59998         18.0178
25                0.1                 1.1               0.79069         19.3232
                                                                                       lyzed parametrically using the FEA and DOE with random
 2.5              0.1                 1.1               1.99779         16.4806        number generation (computer experiment). On the basis of the
25                0.15                1                 0.02567         17.0578        quantitative and qualitative analysis made herein, the following
13.75             0.05                0.9               0.84817         14.7993        conclusions could be drawn.
25                0.05                1                 0.33517         17.4941           The sidewall curl is very sensitive to the contact condition in
 2.5              0.1                 0.9               3.44672         12.2975
 2.5              0.05                1                 0.23165         15.0724
                                                                                       the simulation; hard contact is preferred for high strength steel.
25                0.1                 0.9               0.31749         16.2173           Springback variation in this case is not distinguishable from
13.75             0.05                1.1               0.16992         18.6282        the system-level noise. Therefore, it is uncontrollable in this
                                                                                       case. In order to reduce springback variation, the standard devi-
                                                                                       ations used for variable randomization has to be decreased;
level for the minimum springback variation suggested by Eqs.                           virtually, it means that a system-level adjustment of the press
(5)–(7), which does not agree with Table 7. This discrepancy                           has to be performed to reduce the part-to-part variation of the
was explained by the third method.                                                     equipment. On the other hand, if the springback variation is large
                                                                                       and uncontrollable, then the springback compensation technique
5.3. Taguchi approach                                                                  has to be chosen with it in mind.
                                                                                          A methodology for the variation simulation of springback
   Taguchi analysis was used to analyze the springback vari-                           was developed, which provides a rapid understanding of the
ation. MINITAB, a statistical software, was used to analyze                            influence of the random process variations on the springback
the existing experiment results (Table 4). MINITAB can auto-                           variation of the formed part using FEA techniques eliminating
matically extract data (standard deviation and mean) from the                          the need for lengthy and costly physical experiments.
available experimental observations. For example, Table 9 is the
extracted data of β1 used in MINITAB for Taguchi analysis. In                          References
MINITAB, the main effects of each design factor on the standard
deviations of the response are obtained via regression analysis,                        [1] W.D. Carden, L.M. Geng, D.K. Matlock, R.H. Wagoner, Measurement of
and the significance of these effects were tested via analysis                               springback, Int. J. Mech. Sci. 44 (2002) (2002) 79–101.
                                                                                        [2] A. Baba, Y. Tozawa, Effects of tensile force in stretch-forming process on
of variation (ANOVA) and F-tests. P-values (P) were used to                                 the springback, Bull. JSME 7 (1964) 835–843.
determine which of the effects in the model are statistically sig-                      [3] Z.T. Zhang, D. Lee, Effects of process variables and material properties
nificant, which are compared with a -level of 0.05. As shown                                 on the springback behavior of 2D-draw bending parts, in: Automotive
in Table 10, none of the effects are significant, which indicates                            Stamping Technology, SAE, 1995, pp. 11–18.
that the springback variation in this case is not distinguishable                       [4] L.M. Geng, R.H. Wagoner, Role of plastic anisotropy and its evolution on
                                                                                            springback, Int. J. Mech. Sci. 44 (1) (2002) 123–148.
from the system-level noise. In other words, the springback                             [5] K. Mattiasson, A. Strange, P. Thilderkvist, A. Samuelsson, Simulation of
variation is not controllable in this case. Therefore, the conclu-                          springback in sheet metal forming, in: 5th International Conference on
sions based on Monte-Carlo simulation and sensitivity analysis                              Numerical Methods in Industrial Forming Process, New York, 1995, pp.
                                                                                            115–124.
                                                                                        [6] R.H. Wagoner, W.P. WDCarden, D.K. Carden, Matlock, Springback after
Table 10
                                                                                            drawing and bending of metal sheets, vol. 1, in: Proceedings of the IPMM
F-tests for the standard deviation of each springback
                                                                                            ’97—Intelligent Processing and Manufacturing of Materials, 1997, pp.
                         P-value                                                            1–10.
                                                                                        [7] K.P. Li, L.M. Geng, R.H. Wagoner, Simulation of springback with the
                         S.D. (β1 )                 S.D. (β2 )            S.D. (ρ)          draw/bend test, IPMM ’99, IEEE, Vancouver, BC, Canada, 1999, ISBN
Material                 0.283                      0.088                 0.853             0-7803-5489-3, p. 1.
BHF                      0.098                      0.648                 0.747         [8] S.W. Lee, D.Y. Yang, An assessment of numerical parameters influenc-
Friction                 0.635                      0.495                 0.542             ing springback in explicit finite element analysis of sheet metal forming
                                                                                            process, J. Mater. Process. Technol. 80–81 (1998) 60–67.
198                                     P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198
                                                      c

 [9] K.P. Li, L. Geng, R.H. Wagoner, Simulation of springback: choice of ele-       [24] T.W. Simpson, J.D. Peplinski, P.N. Koch, J.K. Allen, On the use of statistics
     ment Advanced Technology of Plasticity, vol. III, Springer, Berlin, 1999,           in design and the implications for deterministic computer experiments, in:
     pp. 2091–2098.                                                                      Proceedings of ASME DETC’97, 1977.
[10] W.Y.D. Yuen, Springback in the stretch–bending of sheet metal with non-        [25] W.J. Welch, W.K. Yu, S.M. Kang, J. Sacks, Computer experiments for
     uniform deformation, J. Mater. Process. Technol. 22 (1990) 1–20.                    quality control by parameter design, J. Qual. Technol. 22 (1) (1990) 15–22.
[11] S.C. Tang, Analysis of springback in sheet forming operation Advanced          [26] M.D. McKay, W.J. Conover, R.J. Beckman, A comparison of three methods
     Technology of Plasticity, vol. 1, Springer, Berlin, 1987, pp. 193–197.              for selecting values of input variables in the analysis of output from a
[12] L. Focellese, F. Fratini, M.F. Gabrielli, The evaluation of springback in 3D        computer code, Technometrics 21 (1979) 239–245.
     stamping and coining processes, J. Mater. Process. Technol. 80–81 (1998)       [27] J. Sacks, S.B. Schiller, W.J. Welch, Designs for computer experiments,
     108–112.                                                                            Technometrics 31 (1989) 41–47.
[13] N. Narasimhan, M. Lovell, Predicting springback in sheet metal forming:        [28] J.P.C. Kleijnen, Statistical Tools for Simulation Practitioners, Statistics
     an explicit to implicit sequential solution procedure, Finite Elements Anal.        Textbooks and Monographs, vol. 76, M. Dekker, New York, 1987.
     Des. 33 (1999) 29–42.                                                          [29] A.A. Giunta, J.M. Dudley, R. Narducci, B. Grossman, R.T. Haftka,
[14] D.W. Park, J.J. Kang, J.P. Hong, Springback simulation by combined                  W.H. Mason, L.T. Watson, Noisy aerodynamic response and smooth
     method of explicit and implicit FEM, in: Proceedings of NUMISHEET’99,               approximations in high-speed civil transport design, vol. 2, in: 5th
     1999, pp. 35–40.                                                                    AIAA/USA/NASA/ISSMO Symposium on Multidisciplinary Analysis and
[15] M. Valente, D. Traversa, Springback calculation of sheet metal parts after          Optimisation, 1994, pp. 1117–1128.
     trimming and flanging, in: Proceedings of NUMISHEET, 1999, pp. 59–64.           [30] A.A. Giunta, V. Balabanov, D. Haim, B. Grossman, W.H. Mason, L.T.
[16] K.P. Li, W.P. Carden, R.H. Wagoner, Simulation of springback, Int. J. Mech.         Watson, Wing design for a high-speed civil transport using a design of
     Sci. 44 (2002) 103–122.                                                             experiments methology, vol. 1, in: 6th AIAA/USA/NASA/ISSMO Sym-
[17] S.C. Liu, S.J. Hu, Variation simulation for deformable sheet metal assem-           posium on Multidisciplinary Analysis and Optimisation, 1996, pp. 168–
     blies using finite element methods, J. Manuf. Sci. Eng. Trans. ASME 119              183.
     (3) (1997) 368–374.                                                            [31] G. Venter, R.T. Haftka, J.H. Starnes, Construction of response surfaces for
[18] S.D. Button, Determinant assembled stowage bins—a case study, Polym.                design optimization applications, vol. 1, in: 6th AIAA/USA/NASA/ISSMO
     Compos. 20 (1) (1999) 86–97.                                                        Symposium on Multidisciplinary Analysis and Optimisation, 1996, pp.
[19] S.C. Liu, S.J. Hu, An offset finite-element model and its applications in            548–564.
     predicting sheet-metal assembly variation, Int. J. Machine Tools Manuf.        [32] K.D. Majeske, P.C. Hammett, Identifying sources of variation in sheet metal
     35 (11) (1995) 1545–1557.                                                           stamping., Int. J. Flexible Manuf. Syst. 15 (2003) 5–18.
[20] F.M. Swanstrom, T. Hawke, Design for manufacturing and assembly: a             [33] M.G. Lee, D.Y. Kim, C.M. Kim, M.L. Wenner, K.S. Chung, Spring-back
     case study in cost reduction for composite wing tip structures, SAMPE J.            evaluation of automotive sheets based on isotropic–kinematic hardening
     36 (3) (2000) 9–16.                                                                 laws and non-quadratic anisotropic yield functions. Part III. Applications,
[21] R.J. Eggert, Design variation simulation of thick-walled cylinders, J. Mech.        Int. J. Plasticity 21 (5) (2004) 915–953.
     Des. 117 (2) (1995) 221–228.                                                   [34] M.G. Lee, D.Y. Kim, C.M. Kim, M.L. Wenner, R.H. Wagoner, K.S. Chung,
[22] S.D. Kini. An approach to integrating numerical and response surface mod-           Spring-back evaluation of automotive sheets based on isotropic–kinematic
     els for robust design of production systems, Ph.D. Thesis, The Ohio State           hardening laws and non-quadratic anisotropic yield functions. Part II. Char-
     University, 2004.                                                                   acterization of material properties, Int. J. Plasticity 21 (5) (2004) 883–914.
[23] J. Sacks, W.J. Welch, T.J. Mitchell, H.P. Wynn, Design and analysis of         [35] L. Papeleux, J.P. Ponthot, Finite element simulation of springback in sheet
     computer experiments, Stat. Sci. 4 (4) (1989) 409–435.                              metal forming, J. Mater. Process. Technol. 125–126 (2002) 785–791.

Contenu connexe

Tendances

OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
IAEME Publication
 
Ijciet 10 01_032
Ijciet 10 01_032Ijciet 10 01_032
Ijciet 10 01_032
IAEME Publication
 

Tendances (20)

IRJET- Evaluation of Ductility Demand in a Multi Storey Building having Symme...
IRJET- Evaluation of Ductility Demand in a Multi Storey Building having Symme...IRJET- Evaluation of Ductility Demand in a Multi Storey Building having Symme...
IRJET- Evaluation of Ductility Demand in a Multi Storey Building having Symme...
 
Performance of Flat Slab Structure Using Pushover Analysis
Performance of Flat Slab Structure Using Pushover AnalysisPerformance of Flat Slab Structure Using Pushover Analysis
Performance of Flat Slab Structure Using Pushover Analysis
 
IRJET-A Comparative Study of RC Column and Composite Column with Flat Slab Sy...
IRJET-A Comparative Study of RC Column and Composite Column with Flat Slab Sy...IRJET-A Comparative Study of RC Column and Composite Column with Flat Slab Sy...
IRJET-A Comparative Study of RC Column and Composite Column with Flat Slab Sy...
 
Performance based plastic design method for steel concentric braced
Performance based plastic design method for steel concentric bracedPerformance based plastic design method for steel concentric braced
Performance based plastic design method for steel concentric braced
 
Iisrt akshata ht
Iisrt akshata htIisrt akshata ht
Iisrt akshata ht
 
Soil sheet pile interaction part ii numerical analysis and simulation
Soil sheet pile interaction  part ii numerical analysis and simulationSoil sheet pile interaction  part ii numerical analysis and simulation
Soil sheet pile interaction part ii numerical analysis and simulation
 
OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
OPTIMUM DESIGN OF SEMI-GRAVITY RETAINING WALL SUBJECTED TO STATIC AND SEISMIC...
 
Geometrically Variations of Steel Frame Structures: P-Delta Analysis
Geometrically Variations of Steel Frame Structures: P-Delta AnalysisGeometrically Variations of Steel Frame Structures: P-Delta Analysis
Geometrically Variations of Steel Frame Structures: P-Delta Analysis
 
IRJET- Seismic Retrofitting
IRJET- Seismic RetrofittingIRJET- Seismic Retrofitting
IRJET- Seismic Retrofitting
 
IRJET- A Performance Study of High Raise Building with Flat Slab System u...
IRJET-  	  A Performance Study of High Raise Building with Flat Slab System u...IRJET-  	  A Performance Study of High Raise Building with Flat Slab System u...
IRJET- A Performance Study of High Raise Building with Flat Slab System u...
 
IRJET- Behaviour of Prefabricated Steel Reinforced Concrete Column with C...
IRJET-  	  Behaviour of Prefabricated Steel Reinforced Concrete Column with C...IRJET-  	  Behaviour of Prefabricated Steel Reinforced Concrete Column with C...
IRJET- Behaviour of Prefabricated Steel Reinforced Concrete Column with C...
 
Performance evaluation on thin whitetopping
Performance evaluation on thin whitetoppingPerformance evaluation on thin whitetopping
Performance evaluation on thin whitetopping
 
Ijciet 10 01_032
Ijciet 10 01_032Ijciet 10 01_032
Ijciet 10 01_032
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
 
A Parametric Study of Reinforced Concrete Slabs-on-Grade in Industrial Buildings
A Parametric Study of Reinforced Concrete Slabs-on-Grade in Industrial BuildingsA Parametric Study of Reinforced Concrete Slabs-on-Grade in Industrial Buildings
A Parametric Study of Reinforced Concrete Slabs-on-Grade in Industrial Buildings
 
IRJET - Analysis of Flat Slab Structural System in Different Earthquake Zones...
IRJET - Analysis of Flat Slab Structural System in Different Earthquake Zones...IRJET - Analysis of Flat Slab Structural System in Different Earthquake Zones...
IRJET - Analysis of Flat Slab Structural System in Different Earthquake Zones...
 
IRJET-Cyclic Response of Perforated Beam in Steel Column Joints
IRJET-Cyclic Response of Perforated Beam in Steel Column JointsIRJET-Cyclic Response of Perforated Beam in Steel Column Joints
IRJET-Cyclic Response of Perforated Beam in Steel Column Joints
 
IRJET- Buckling Behavior of Cold Formed Steel Column Under Axial Loading
IRJET- Buckling Behavior of Cold Formed Steel Column Under Axial LoadingIRJET- Buckling Behavior of Cold Formed Steel Column Under Axial Loading
IRJET- Buckling Behavior of Cold Formed Steel Column Under Axial Loading
 
Influence of Aspect Ratio & Plan Configurations on Seismic Performance of Mul...
Influence of Aspect Ratio & Plan Configurations on Seismic Performance of Mul...Influence of Aspect Ratio & Plan Configurations on Seismic Performance of Mul...
Influence of Aspect Ratio & Plan Configurations on Seismic Performance of Mul...
 
IRJET- Pushover Analysis on Reinforced Concrete Building using ETABS
IRJET- Pushover Analysis on Reinforced Concrete Building using ETABSIRJET- Pushover Analysis on Reinforced Concrete Building using ETABS
IRJET- Pushover Analysis on Reinforced Concrete Building using ETABS
 

Similaire à Simulation of springback variation in forming of advanced high strength steels

Experimental investigation of springback variation in forming of high strengt...
Experimental investigation of springback variation in forming of high strengt...Experimental investigation of springback variation in forming of high strengt...
Experimental investigation of springback variation in forming of high strengt...
Peng Chen
 
Parameter identification of rockfall protection
Parameter identification of rockfall protectionParameter identification of rockfall protection
Parameter identification of rockfall protection
Juan Escallón
 
Resistane of concrte slab due to shear effect
Resistane of concrte slab due to shear effectResistane of concrte slab due to shear effect
Resistane of concrte slab due to shear effect
luqman geotechnic
 
Ijciet 10 01_020
Ijciet 10 01_020Ijciet 10 01_020
Ijciet 10 01_020
IAEME Publication
 

Similaire à Simulation of springback variation in forming of advanced high strength steels (20)

IOSR Journal of Engineering (IOSR-JEN) Volume 4 Issue 9 Version 3
IOSR Journal of Engineering (IOSR-JEN) Volume 4 Issue 9 Version 3IOSR Journal of Engineering (IOSR-JEN) Volume 4 Issue 9 Version 3
IOSR Journal of Engineering (IOSR-JEN) Volume 4 Issue 9 Version 3
 
Theinfluencesoftheresidualformingdataonthequasi staticaxialcrashresponceofato...
Theinfluencesoftheresidualformingdataonthequasi staticaxialcrashresponceofato...Theinfluencesoftheresidualformingdataonthequasi staticaxialcrashresponceofato...
Theinfluencesoftheresidualformingdataonthequasi staticaxialcrashresponceofato...
 
Evaluation of Reduction in Compressive Strength of Singly Symmetric CFS Membe...
Evaluation of Reduction in Compressive Strength of Singly Symmetric CFS Membe...Evaluation of Reduction in Compressive Strength of Singly Symmetric CFS Membe...
Evaluation of Reduction in Compressive Strength of Singly Symmetric CFS Membe...
 
IJET-Waste Water Treatment Unit using Activated Charcoal
IJET-Waste Water Treatment Unit using Activated CharcoalIJET-Waste Water Treatment Unit using Activated Charcoal
IJET-Waste Water Treatment Unit using Activated Charcoal
 
Study on Steel Beam Column Joint using Different Connections – State of Art
Study on Steel Beam Column Joint using Different Connections – State of ArtStudy on Steel Beam Column Joint using Different Connections – State of Art
Study on Steel Beam Column Joint using Different Connections – State of Art
 
Experimental investigation of springback variation in forming of high strengt...
Experimental investigation of springback variation in forming of high strengt...Experimental investigation of springback variation in forming of high strengt...
Experimental investigation of springback variation in forming of high strengt...
 
A Review on Finite Element Analysis of Automobile roof header Manufactured By...
A Review on Finite Element Analysis of Automobile roof header Manufactured By...A Review on Finite Element Analysis of Automobile roof header Manufactured By...
A Review on Finite Element Analysis of Automobile roof header Manufactured By...
 
Design Modification of Failure Mode Effect Analysis of Vibrating Feeder used ...
Design Modification of Failure Mode Effect Analysis of Vibrating Feeder used ...Design Modification of Failure Mode Effect Analysis of Vibrating Feeder used ...
Design Modification of Failure Mode Effect Analysis of Vibrating Feeder used ...
 
STRENGTHENING OF PRECAST BEAM-COLUMN JOINT USING STEEL ENCASEMENT
STRENGTHENING OF PRECAST BEAM-COLUMN JOINT USING STEEL ENCASEMENTSTRENGTHENING OF PRECAST BEAM-COLUMN JOINT USING STEEL ENCASEMENT
STRENGTHENING OF PRECAST BEAM-COLUMN JOINT USING STEEL ENCASEMENT
 
FEA Based Validation of Weld Joint Used In Chassis of Light Commercial Vehicl...
FEA Based Validation of Weld Joint Used In Chassis of Light Commercial Vehicl...FEA Based Validation of Weld Joint Used In Chassis of Light Commercial Vehicl...
FEA Based Validation of Weld Joint Used In Chassis of Light Commercial Vehicl...
 
1 i18 ijsrms0206178-v2-i6-prasad diwanji-30nov15
1 i18 ijsrms0206178-v2-i6-prasad diwanji-30nov151 i18 ijsrms0206178-v2-i6-prasad diwanji-30nov15
1 i18 ijsrms0206178-v2-i6-prasad diwanji-30nov15
 
IRJET- Comparison of Reliability of Circular and Square CFST Columns usin...
IRJET-  	  Comparison of Reliability of Circular and Square CFST Columns usin...IRJET-  	  Comparison of Reliability of Circular and Square CFST Columns usin...
IRJET- Comparison of Reliability of Circular and Square CFST Columns usin...
 
Parameter identification of rockfall protection
Parameter identification of rockfall protectionParameter identification of rockfall protection
Parameter identification of rockfall protection
 
Utilization of steel in construction of high performance structures: A Review
Utilization of steel in construction of high performance structures: A ReviewUtilization of steel in construction of high performance structures: A Review
Utilization of steel in construction of high performance structures: A Review
 
Resistane of concrte slab due to shear effect
Resistane of concrte slab due to shear effectResistane of concrte slab due to shear effect
Resistane of concrte slab due to shear effect
 
Parametric study on reinforced concrete beam
Parametric study on reinforced concrete beamParametric study on reinforced concrete beam
Parametric study on reinforced concrete beam
 
A REVIEW ON OPTIMIZATION OF RESISTANCE SPOT WELDING OF ALUMINUM COMPONENTS US...
A REVIEW ON OPTIMIZATION OF RESISTANCE SPOT WELDING OF ALUMINUM COMPONENTS US...A REVIEW ON OPTIMIZATION OF RESISTANCE SPOT WELDING OF ALUMINUM COMPONENTS US...
A REVIEW ON OPTIMIZATION OF RESISTANCE SPOT WELDING OF ALUMINUM COMPONENTS US...
 
REVIEW STATIC AND DYNAMIC ANALYSIS OF A LAMINATED COMPOSITE BEAM
REVIEW STATIC AND DYNAMIC ANALYSIS OF A LAMINATED COMPOSITE BEAMREVIEW STATIC AND DYNAMIC ANALYSIS OF A LAMINATED COMPOSITE BEAM
REVIEW STATIC AND DYNAMIC ANALYSIS OF A LAMINATED COMPOSITE BEAM
 
Ijciet 10 01_020
Ijciet 10 01_020Ijciet 10 01_020
Ijciet 10 01_020
 
Comparative Analysis of Composite Materials based on Stress and Vibration by ...
Comparative Analysis of Composite Materials based on Stress and Vibration by ...Comparative Analysis of Composite Materials based on Stress and Vibration by ...
Comparative Analysis of Composite Materials based on Stress and Vibration by ...
 

Dernier

1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdf
QucHHunhnh
 
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Krashi Coaching
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
ciinovamais
 
Beyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global ImpactBeyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global Impact
PECB
 
1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdf
QucHHunhnh
 

Dernier (20)

INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptxINDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
 
1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdf
 
Arihant handbook biology for class 11 .pdf
Arihant handbook biology for class 11 .pdfArihant handbook biology for class 11 .pdf
Arihant handbook biology for class 11 .pdf
 
Disha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfDisha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdf
 
Student login on Anyboli platform.helpin
Student login on Anyboli platform.helpinStudent login on Anyboli platform.helpin
Student login on Anyboli platform.helpin
 
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
 
The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13
 
Unit-IV- Pharma. Marketing Channels.pptx
Unit-IV- Pharma. Marketing Channels.pptxUnit-IV- Pharma. Marketing Channels.pptx
Unit-IV- Pharma. Marketing Channels.pptx
 
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
 
Paris 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activityParis 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activity
 
Activity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdfActivity 01 - Artificial Culture (1).pdf
Activity 01 - Artificial Culture (1).pdf
 
Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111
 
General AI for Medical Educators April 2024
General AI for Medical Educators April 2024General AI for Medical Educators April 2024
General AI for Medical Educators April 2024
 
Beyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global ImpactBeyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global Impact
 
APM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across SectorsAPM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across Sectors
 
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
 
Key note speaker Neum_Admir Softic_ENG.pdf
Key note speaker Neum_Admir Softic_ENG.pdfKey note speaker Neum_Admir Softic_ENG.pdf
Key note speaker Neum_Admir Softic_ENG.pdf
 
Introduction to Nonprofit Accounting: The Basics
Introduction to Nonprofit Accounting: The BasicsIntroduction to Nonprofit Accounting: The Basics
Introduction to Nonprofit Accounting: The Basics
 
1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdf
 
9548086042 for call girls in Indira Nagar with room service
9548086042  for call girls in Indira Nagar  with room service9548086042  for call girls in Indira Nagar  with room service
9548086042 for call girls in Indira Nagar with room service
 

Simulation of springback variation in forming of advanced high strength steels

  • 1. Journal of Materials Processing Technology 190 (2007) 189–198 Simulation of springback variation in forming of advanced high strength steels Peng Chen b , Muammer Koc a,∗ ¸ a NSF I/UCR Center for Precision Forming (CPF) and Department of Mechanical Engineering, Virginia Commonwealth University (VCU), Richmond, VA, USA b Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA Received in revised form 23 February 2007; accepted 27 February 2007 Abstract Variations in the mechanical and dimensional properties of the incoming material, lubrication and other forming process parameters are the main causes of springback variation. Variation of springback prevents the applicability of the springback prediction and compensation techniques. Hence, it leads to amplified variations and problems during assembly of the stamped components, in turn, resulting in quality issues. To predict the variation of springback and to improve the robustness of the forming process, variation simulation analysis could be adopted in the early design stage. Design of experiment (DOE) and finite element analysis (FEA) approach was used for the variation simulation and analysis of the springback for advanced high strength steel (AHSS) parts. To avoid the issues caused by the deterministic FEA simulation, random number generation was used to introduce uncertainties in DOE. This approach was, then, applied to investigate the effects of variations in material, blank holder force and friction on the springback variation for an open-channel shaped part made of dual phase (DP) steel. This approach provides a rapid and accurate understanding of the influence of the random process variations on the springback variation of the formed part using FEA techniques eliminating the need for lengthy and costly physical experiments. © 2007 Elsevier B.V. All rights reserved. Keywords: Springback variation; Variation simulation; Design of experiment; Finite element analysis; Computer experiment; Advanced high strength steel 1. Introduction due to challenges in terms of formability, springback, joining and die life issues. During the past two decades, the automotive industry has Springback is an elastically driven change in shape and form been in a quest to achieve low-mass vehicles due to government of a part upon unloading after a part is formed. The concerns regulations and consumer demands for light, fuel-efficient and about springback and quality control grow among auto-makers low-emission vehicles. In addition, requirements for improved as the use of high strength–low weight materials increases [1]. safety, performance, and comfort features continue putting Parts made of AHSS demonstrate more springback than parts pressure on the automotive manufacturers especially in an envi- made of mild steel do. Moreover, the experience with forming ronment where there are ever-increasing global competition of AHSS materials is limited compared to mild steel. Variation and continuous cost reduction demands. As an alternative to of springback prevents the applicability and use of springback aluminum and magnesium alloys, advanced/ultra high strength prediction and compensation techniques. Hence, it leads to steels (AHSS) have been proven to be a proper choice of mate- amplified variations and problems during assembly of compo- rial for various body and structural automotive parts meeting the nents, and in turn, results in quality issues. Thus, understanding, required low-mass, affordable cost and increased performance accurate characterization, prediction, control and reduction of requirements. However, wide application of high strength steel springback and its variation have become very crucial in terms in many potential auto body and structural parts is still limited of decreasing development times and reducing scrap rate in mass production to achieve cost effective fabrication of AHSS parts. Numerous studies during the last 40 years have attempted to ∗ Corresponding author. Tel.: +1 804 827 7029. determine the controlling factors in springback and find ways to E-mail address: mkoc@vcu.edu (M. Koc). ¸ reduce it. An early study by Baba and Tozawa [2] focused on 0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2007.02.046
  • 2. 190 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c the effect of stretching a sheet by a tensile force, during or after simulation, which is especially designed for the assembly pro- bending, in minimizing springback. Other studies investigated cess (welding) simulation based on the assumption of linear the role of process variables on springback. Zhang and Lee [3] elasticity. showed the influence of blank holder force, elastic modulus, Generally speaking, Monte-Carlo simulation is effective for strain hardening exponent, blank thickness and yield strength any kind of non-linear process. However, since large number of on the magnitude of the final springback strain in a part. Geng FEA simulations is required (more than 100 for reliable results), and Wagoner [4] studied the effects of plastic anisotropy and its it is not practical and too costly to apply it for a complicated evolution in springback. They developed a constitutive equation case, such as metal forming. Therefore, some researchers [22] for 6022-T4 aluminum alloy using a new anisotropic hardening applied DOE techniques in their FE simulation to effectively model and proved that Barlat’s yield function is more accurate reduce the amount of simulations. However, since the FEA sim- than other yield functions in their case. ulation is deterministic (no random error), replicate observations Some researchers evaluated FE simulation procedures in from running the simulation with the same inputs will be identi- terms of their springback prediction accuracy. Mattiasson et cal. Despite some similarities to physical experiments, the lack al. [5], Wagoner et al. [6], Li et al. [7] and Lee and Yang [8] of random error makes FEA simulations different from physi- found that FEM simulations of springback are much more sen- cal experiments. In the absence of independent random errors, sitive to numerical tolerances than forming simulations are. Li the rationale for least-squares fitting of a response surface is not et al. [9] investigated the effects of element type on the spring- clear [23]. The usual measures of uncertainty derived from least- back simulation. Yuen [10] and Tang [11] found that different squares residuals have no obvious statistical meaning [23,24]. unloading scheme will affect the accuracy of the springback According to Welch et al. [25], in the presence of systematic prediction. Similarly, Focellese et al. [12] and Narasimhan and error rather than random error, statistical testing is inappropri- Lovell [13] pointed out that different integration scheme will ate. For deterministic FEA simulation, some statistics, including also influence the result of springback simulation. In 1999, F-statistics, have no statistical meaning since they assume the Park et al. [14] and Valente and Traversa [15] attempted to observations include an error term which has mean of zero and link dynamic explicit simulations of forming operations to static a non-zero standard deviation. Consequently, the use of step- implicit simulations of springback. It was proved that this tech- wise regression for polynomial model fitting is not appropriate nique is very effective for the springback simulation. Li et since it utilizes F-statistic value when adding/removing model al. [16] explored a variety of issues in the springback simu- parameters [24]. lations. They concluded that (1) typical forming simulations In an effort to solve the above problem, McKay et al. [26] are acceptably accurate with 5–9 through-thickness integration introduced Latin hypercube sampling which ensures that each points for shell/beam type elements, whereas springback anal- of the input variables has all portions of its range presented. ysis within 1% numerical error requires up to 51 points, and Sacks et al. [23,27] proposed the design and analysis of com- more typically 15–25 points, depending on R/t, sheet tension puter (DACE) method to model the deterministic output as the and friction coefficient. (2) More contact nodes are necessary realization of a stochastic process, thereby providing a statis- for accurate springback simulations than for forming simulation, tical basis for designing experiments for efficient prediction. approximately one node per 5◦ of turn angle versus 10◦ recom- Kleijnen [28] suggested incorporating substantial random error mended for forming. (3) Three-dimensional shell and non-linear through random number generators. Therefore, it is natural to solid elements are preferred for springback prediction even for design and analyze such stochastic simulation experiments using large w/t ratios because of the presence of persistent anticlas- standard techniques for physical experiments. Some researchers tic curvature. For R/t > 5.6, shell elements are preferred since (e.g., Giunta et al. [29,30] and Venter et al. [31]), have also solid elements are too computation-intensive. For R/t < 5.6, non- employed metamodeling techniques such as RSM for modeling linear 3D solid elements are required for accurate springback deterministic computer experiments which contain numerical prediction. noise. This numerical noise is used as a surrogate for random Most of the research efforts on springback focused on the error, thus allowing the standard least-squares approach to be accurate prediction and compensation of springback. The issue applied. However, the assumption of equating numerical noise of springback variation was seldom concerned. Moreover, none to random error is questionable. of the studies in the area of springback prediction touched In this work, random number generator was used to ensure upon the variation simulation of springback. As lead times the correctness of the regression model. Assuming a Gaussian are shortened and materials of high strength–low weight are distribution, the uncertainty was introduced by random number used in manufacturing, a fundamental understanding of the generation for controlled factors at different level and uncon- springback variation has become essential for accurate and trolled variables according to their mean and range. By using rapid design of tooling and processes in the early design stage. this method, the effects of variations in material (mechanical As far as the variation simulation is concerned, most avail- properties) and process (blank holder force and friction) on able references are related to the assembly processes [17–21]. the springback variation were investigated for an open-channel In 1997, Liu and Hu [17] summarized two variation simula- shaped part made of DP steel. The variations in stamping process tion approaches for sheet metal assembly. They are: (1) direct are introduced in Section 2 of the paper. In Section 3, measure- Monte-Carlo simulation, which is popular and good for all kinds ments of springback are defined. Section 4 is the validation of the of random process simulation, and (2) mechanistic variation FE model using existing experimental results in the literature.
  • 3. P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c 191 Fig. 1. Springback variation and its sources. In the following section, variation analysis of an open-channel • Batch-to-batch variation represents the variability among the part made of advanced high strength steel is presented. The last individual batches, which is mostly caused by material vari- section is discussions and conclusions. ation from batch to batch and the variation introduced by tooling setup. 2. Variations in stamping process 3. Measurement of springback for open-channel In this study, the objective is to understand and accurately drawing predict the variation of springback in an open-channel drawing considering the variations of material and process as shown in A schematic view of die, punch, blank and their dimensions Fig. 1. for open-channel drawing, which is used in the analyses for Total variation of springback in the stamping process has this study, is shown in Fig. 3. Fig. 4 shows the formed part after several components. Generally, different variation components springback. Three measurements, namely the springback of wall can be attributed to different sources [32]. The following are the opening angle (β1 ), the springback of flange angle (β2 ) and side- major categories of variation source (Fig. 2): wall curl radius (ρ) shown in Fig. 5, were used to characterize the total springback considering only the cross-sectional shapes of formed parts obtained before and after the removal of tools. The • Part-to-part variation is also referred to as system-level vari- springback in the direction orthogonal to the cross-section, such ation or inherent variation. It is the amount of variation that as twisting, was not considered since it is negligible is this case. can be expected across consecutive parts produced by the pro- As there is no clear distinction to separate a cross-section curve cess during a given run. It is caused by the random variation for individual measurement of springback angles and sidewall of all the uncontrolled (controllable and uncontrollable) pro- curl, two assumptions deduced from the sample observations cess variables. In the variation simulation of this paper, blank thickness was considered as the uncontrolled variable. • Within batch variation is usually due to the variations of the controlled variables such as BHF, material property and fric- tion. Fig. 3. A schematic view of tools and dimensions for open-channel drawing Fig. 2. Source of variation in a typical forming process. [32].
  • 4. 192 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c to construct a circular arc is used. Eq. (1) lists all the equations needed for the calculation of the β1 , β2 and ρ. ox · A0 B0 θ1 = arccos 0 |ox| · A0 B0 ox · A0 B0 θ2 = arccos 0 |ox| · A0 B0 Fig. 4. Open-channel parts after drawing. ox · AB θ1 = arccos |ox| · |AB| AB · ED θ2 = arccos |ED| · |AB| (1) β1 = θ1 − θ1 0 β2 = θ2 − θ2 0 xB + yB − xA − yA − ((yA − yB )/ 2 2 2 2 (yC − yB ))(xC + yC − xB − yB ) 2 2 2 2 xO = 2 xB − xA + (xC − xB )((yA − yB )/(yC − yB )) xA + yA − xB − yB + 2xO (xB − xA ) 2 2 2 2 yO = 2(yA − yB ) ρ= (xA − xO )2 + (yA − yO )2 4. Finite element modeling and validation for the Fig. 5. Illustration of springbacks. open-channel drawing of AHSS are introduced for the springback measurement. Firstly, it is The simulation work for this study is based on the exper- assumed that wall opening angle, flange closing angles and side- imental results of Lee et al. [33]. Information about the wall curl vary independently. Secondly, the sidewall curl could geometry and dimensions of the tooling and blank are pre- be approximated by a piece of circular arc. sented in Fig. 3. The initial dimension of the blank sheet was Fig. 5 also shows the measurements placements (A–E). Two 300 mm (length) × 35 mm (width). Forming was carried out on measurements were conducted before springback, namely the x a 150 tonnes double action hydraulic press with a punch speed and y coordinates of A and B, which is denoted as A0 and B0 of 1 mm/s, and the total punch stroke was 70 mm. Blank holder in this work. They are used to compute the wall angle (θ1 ) and0 force (BHF) was 2.5 kN. The blank material used was DP Steel flange angle (θ2 0 ) before springback. After springback, another with the material properties presented in Fig. 6 based on the five measurements were placed on A–E, which were used in the tensile tests by Lee et al. [34]. calculation of the wall angle (θ 1 ), flange angle (θ 2 ) and sidewall Considering the geometric symmetry of the process, only curl radius (ρ) after springback. To estimate the sidewall curl half of the blank was simulated. The material was modeled as radius, a curve fitting technique that employs three points (A–C) an elastic–plastic material with isotropic elasticity, using the Fig. 6. Material properties of DP steel [33].
  • 5. P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c 193 Table 1 Different FEA procedure used in simulation Case 1 2 3 4 5 6 7 8 9 Element type Solid Solid Solid Solid Shell Shell Shell Shell Shell Contact Soft Soft Soft Hard Soft Soft Hard Hard Soft Forming analysis (dynamic) Implicit Implicit Implicit Implicit Implicit Implicit Explicit Explicit Explicit Springback analysis (static) Implicit Implicit Implicit Implicit Implicit Implicit Implicit Implicit Implicit Through-thickness element number 5 9 21 9 9 21 5 15 9 or integration point Hill anisotropic yield criterion for the plasticity. The coefficients in Table 1) of different element type, different contact condi- of Hill yield criterion (R11 = 1.0, R22 = 1.01951, R33 = 1.00219, tion, different through-thickness element number and different R12 = 0.992318, R13 = 1.0, R23 = 1.0) were computed from the analysis type were tried in the simulation. The term, soft con- r-values as presented in Fig. 6. The friction coefficient between tact, denotes exponential pressure-overclosure definition for the tools and the sheet blank was assumed to be constant and normal behavior between contacting surfaces. 0.1. To determine the appropriate element type, contact con- The comparison of different simulation procedure for the ditions, through-thickness element number and analysis type prediction of wall opening angle (θ 1 ), flange angle (θ 2 ) and side- for simulation using ABAQUS, nine combinations (as tabulated wall curl radius (ρ) is shown in Figs. 7–9. It was found that the sidewall curl is very sensitive to the contact condition used in simulation. Since the soft contact tends to soften the contact- ing surface, it actually depresses the sidewall curl, which is not true for advanced high strength steel. Among these combina- tions, case 4 (hard contact), case 7 (hard contact) and case 8 (hard contact) show a good match with the experiment results in all three springback measurements. Hence, hard contact is Table 2 Original experiment design Run order BHF (kN) Friction Material Fig. 7. Effects of different FEA procedure on the prediction of wall opening angle. 1 13.75 0.15 1.1 2 13.75 0.1 1 3 13.75 0.15 0.9 4 2.5 0.15 1 5 25 0.1 1.1 6 2.5 0.1 1.1 7 25 0.15 1 8 2.5 0.15 1 9 2.5 0.1 1.1 10 13.75 0.05 0.9 11 25 0.05 1 12 2.5 0.1 0.9 13 2.5 0.05 1 14 13.75 0.1 1 Fig. 8. Effects of different FEA procedure on the prediction of flange closing 15 13.75 0.1 1 angle. 16 13.75 0.05 0.9 17 2.5 0.1 0.9 18 13.75 0.1 1 19 25 0.05 1 20 13.75 0.15 0.9 21 25 0.1 0.9 22 13.75 0.05 1.1 23 13.75 0.1 1 24 25 0.15 1 25 13.75 0.15 1.1 26 25 0.1 0.9 27 13.75 0.05 1.1 28 13.75 0.1 1 29 2.5 0.05 1 Fig. 9. Effects of different FEA procedure on the prediction of sidewall curl 30 25 0.1 1.1 radius.
  • 6. 194 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c Table 3 5. Variation simulation of springback and results Assumed statistics of variables Mean Range S.D. In this study, we only considered the “part-to-part” and Uncontrolled factor “within batch” variations. The variation simulation and anal- Part thickness 1.2 mm 1.18–1.22 mm 0.066667 ysis of the springback of DP steel part are described step by step as follows. Controlled factor BHF level-1 2.5 kN 2.4–2.6 0.033333 BHF level-2 13.75 kN 13.65–13.85 0.033333 Step 1 (Design of experiment). BHF, material property and fric- BHF level-3 25 kN 24.9–25.1 0.033333 tion were chosen as design factors. Box–Behnken RSM design Friction level-1 0.05 0.04–0.06 0.003333 with 2-replicate and 6-center-point was used for this 3-factor and Friction level-2 0.1 0.09–0.11 0.003333 3-level experiment design. The levels of the material property Friction level-3 0.15 0.14–0.16 0.003333 are considered as 110, 100 and 90% of the stress–strain curve Material level-1 90% 88–92% 0.00667 in Fig. 6, which indicates the strength of the material. Table 2 Material level-2 100% 98–102% 0.00667 shows the original experiment table. Material level-3 110% 108–112% 0.00667 Steps 2 and 3Random number generation of controlled and uncontrolled variablesIt is assumed that most random processes preferred. It can be seen that element type and forming anal- conform to a Gaussian distribution. Moreover, irrespective of ysis type do not affect the accuracy of springback prediction the parent distribution of the population, the distribution of the much. Therefore, to reduce the computation time, hard con- average of random samples taken from the population tends to tact, shell element, explicit (dynamic) for forming and implicit be normal as the sample size increases (Central Limit Theo- (static) for springback were used in further simulations. Dif- rem). Therefore, once we know the mean and standard deviation ferent through-thickness integration points (5, 9, 15, and 21) of (S.D.) of a random process, we can generate a random number shell elements were also tried in the simulation, which showed no according to its Gaussian distribution. According to the statistics much influence on the prediction of springback. Therefore, nine chosen in Table 3, the original experiment table was random- through-thickness integration points were used in the further ized as shown in Table 4. Fig. 10 is an illustration of the number simulations. randomization. Table 4 Randomized (random number generation) experiment table and simulation results Run BHF (kN) Friction Material Part thickness β1 (◦ ) β2 (◦ ) ρ (mm) 1 13.649243 0.1405 1.0819 11.8017 16.1242 12.3581 191.5870 2 13.755335 0.0904 0.9814 11.9484 16.5589 12.1747 169.7480 3 13.721143 0.1534 0.8811 11.9619 12.5694 9.4674 301.4050 4 2.3992 0.1512 0.9933 11.9606 18.4420 11.0999 133.4730 5 24.9033 0.0970 1.0923 11.9473 19.8823 13.4837 137.7005 6 2.5053 0.1005 1.0913 12.0388 17.8932 11.0234 138.2284 7 25.0027 0.1466 1.0063 12.0067 17.0396 11.4995 158.0187 8 2.4711 0.1521 0.9910 12.0383 17.5935 10.8937 140.5959 9 2.5291 0.1016 1.0999 12.0638 15.0679 11.1937 140.6841 10 13.779106 0.0404 0.8937 11.9781 14.1995 11.0920 156.6561 11 24.9970 0.0536 1.0160 11.9351 17.7311 12.9760 151.3312 12 2.5072 0.0958 0.8941 12.0725 9.8603 10.0025 178.0388 13 2.4983 0.0422 0.9933 11.8973 14.9086 9.9908 159.4871 14 13.757157 0.0995 1.0031 12.0039 16.4520 12.2300 169.0799 15 13.748318 0.0959 1.0025 12.0272 16.6234 12.2783 166.0132 16 13.737183 0.0555 0.9081 11.9032 15.3990 11.3933 166.4367 17 2.4872 0.0970 0.8889 12.0405 14.7347 9.3532 168.8445 18 13.791958 0.1035 1.0077 11.9982 16.3857 12.2430 171.5795 19 24.9737 0.0527 0.9922 11.9569 17.2571 12.6146 156.6909 20 13.780872 0.1515 0.9056 12.0172 12.6997 9.7055 298.5300 21 24.9978 0.0977 0.8982 12.0322 15.9928 11.2144 172.0237 22 13.772126 0.0484 1.1079 11.9815 18.5080 13.6613 122.9733 23 13.718735 0.0964 0.9919 11.8797 16.5546 12.2819 169.2909 24 24.9718 0.1483 1.0078 11.9887 17.0759 11.5550 158.1165 25 13.785849 0.1407 1.1077 12.0855 17.0869 13.9510 178.8833 26 25.0469 0.1037 0.9091 11.8921 16.4418 11.4263 169.8623 27 13.768493 0.0428 1.0955 12.0642 18.7483 14.9717 125.4564 28 13.751131 0.0960 0.9936 12.0129 16.4303 12.1832 169.6901 29 2.542 0.0480 1.0014 11.9466 15.2362 10.1161 157.6620 30 24.9880 0.1004 1.0918 12.0695 18.7641 14.3375 139.4873
  • 7. P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c 195 Table 7 Recommended variable level for the minimum variance of β1 , β2 and ρ according to Monte-Carlo simulation Material Friction BHF Min[Var(β1 )] – High Middle Min[Var(β2 )] – High Low Min[Var(ρ)] Middle Middle – equations. For instance, the coefficient of the XMaterial is +1.5867 in Eq. (2), therefore, the bigger the material strength, the larger the springback. Table 5 tabulates the optimal variable level for the minimum of each springback. A more detailed indication Fig. 10. Schematic diagram of the variable randomizations (random number of the relationship between the springbacks and the factors is generation). shown in Fig. 11. As in the parameter’s range studied in this work, springback increases with BHF and friction, which agrees Table 5 with the experimental observations [35]. Papeleux and Pon- Recommended variable level for the minimum of β1 , β2 and maximum of ρ thot [35] reported that springback increases with small BHF, Material Friction BHF but decreases as the BHF increases for large force values. Minimum β1 Low Low Low This phenomenon can be explained by the fact that with low Minimum β2 Low Low Low BHF, the punch induces mostly bending stresses in the mate- Maximum ρ Low Low – rial, but as the blankholder holds the blank more severely, the stresses included by the punching phase become mostly tensile stresses. Step 4 (Simulation). Simulations were run according to Table 4 and the simulation results are shown in Table 4 as well. Step 6 (Variation sensitivity analysis). Three methods were used to analyze the effects of the factors on the variation of Step 5 (Regression analysis). Eqs. (2)–(4) are the regression the springback. Finally, it was found that the springback varia- models of β1 , β2 , and ρ as functions of BHF, material and fric- tion magnitude is too small in this case and not distinguishable tion. The variables in these equations are coded (−1, 0, 1) factors from the system noise. in the DOE. To investigate whether the factors’ effect on each springback is significant, analysis of variance (ANOVA) was used. Factors with a P-value larger than 0.05 were considered as 5.1. Monte-Carlo simulation insignificant and ignored in the regression model. For example, Monte-Carlo simulation was applied to Eqs. (2)–(4). Accord- the main effect of friction on β1 is negligible: ing to the parameter levels used in the DOE, it is assumed β1 = 16.5008 + 1.5867XMaterial + 0.7287XBHF that all factors have equal variance in the Monte-Carlo simula- tion, i.e., Var(XMaterial ) = Var(XBHF ) = Var(XFriction ) = 0.32 , with −0.8454XBHF XFriction (2) a zero mean value for each factor (coded factors). Monte-Carlo simulation was run 100 times for each situation (a specific factor β2 = 12.2319 + 1.3329XMaterial + 0.9646XBHF at a specific level). The corresponding variance of the springback was recorded in Table 6. Table 7 summarizes the optimal variable −0.3929XFriction − 0.7297XBHF − 0.5529XBHF XFriction 2 level for the minimum variance of each springback according to (3) Table 6. r = 169.234 − 8.384XMaterial + 18.482XMaterial XFriction (4) 5.2. Sensitivity analysis The effect of each factor on each springback could be deter- According to Eqs. (2)–(4), the variance of β1 , β2 and ρ are mined by the sign of the corresponding coefficient in the above expressed as Eqs. (5)–(7) via linearized sensitivity analysis. Table 6 Springback variation Material Friction BHF Low Middle High Low Middle High Low Middle High Var(β1 ) 0.0667 0.0667 0.0667 0.5349 0.324 0.2563 0.3501 0.2571 0.3207 Var(β2 ) 0.1146 0.1146 0.1146 0.4798 0.3276 0.2367 0.1866 0.1926 0.2656 Var(ρ) 37.419 0 37.419 73.712 7.1786 10.414 11.813 11.813 11.813
  • 8. 196 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c Fig. 11. Response surface plots of (a) β1 , (b) β2 and (c) ρ. Var(r) = [−27.3 − 19.35XFriction + 27.194XMaterial ]2 Var(β1 ) = [1.8861 − 1.6648XMaterial ]2 Var(XMaterial ) ×Var(XMaterial ) + [−54.444XBHF ]2 Var(XBHF ) +[1.028 − 0.8454XFriction ]2 Var(XBHF ) +[22.745 − 19.35XMaterial + 19.82XFriction ]2 +[0.8454XBHF ]2 Var(XFriction ) (5) ×Var(XFriction ) (7) Var(β2 ) = [1.3329 + 0.2899XBHF ]2 Var(XMaterial ) The variance of the response is determined by the variance +[0.9646 − 1.4594XBHF + 0.2899XMaterial ]2 of each factor and the sensitivity coefficient (the quantity in the square parentheses). To minimize the variance of the springback, ×Var(XBHF ) + [−0.3929 − 0.5528XBHF ]2 the most efficient way is to minimize the sensitivity coeffi- ×Var(XFriction ) (6) cients in the equation. Table 8 tabulates the optimal variable
  • 9. P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c 197 Table 8 (Tables 7 and 8) actually do not have any meaning because Recommended variable level for the minimum variance of β1 , β2 and ρ according the springback variations are totally random and uncontrollable to sensitivity analysis in this case. In other words, conclusions from both methods Material Friction BHF are neither correct nor wrong. Since the system-level noises Min[Var(β1 )] High High Middle were introduced by random number generation (Table 3) in our Min[Var(β2 )] High Middle High computer experiment, we can solve the problem by reducing Min[Var(ρ)] High Low Middle the standard deviations used in the random number generation. However, this kind of adjustment would not be easy in reality, since the tuning of the system-level noise is usually impossible Table 9 Extracted data (β1 ) used in MINITAB for Taguchi analysis in most cases. BHF (kN) Friction Material S.D. (β1 ) Mean (β1 ) 6. Conclusions 13.75 0.15 1.1 0.68073 16.6056 13.75 0.1 1 0.09154 16.5008 The effects of BHF, material and friction on springback 13.75 0.15 0.9 0.09214 12.6346 and springback variation of DP steel channel have been ana- 2.5 0.15 1 0.59998 18.0178 25 0.1 1.1 0.79069 19.3232 lyzed parametrically using the FEA and DOE with random 2.5 0.1 1.1 1.99779 16.4806 number generation (computer experiment). On the basis of the 25 0.15 1 0.02567 17.0578 quantitative and qualitative analysis made herein, the following 13.75 0.05 0.9 0.84817 14.7993 conclusions could be drawn. 25 0.05 1 0.33517 17.4941 The sidewall curl is very sensitive to the contact condition in 2.5 0.1 0.9 3.44672 12.2975 2.5 0.05 1 0.23165 15.0724 the simulation; hard contact is preferred for high strength steel. 25 0.1 0.9 0.31749 16.2173 Springback variation in this case is not distinguishable from 13.75 0.05 1.1 0.16992 18.6282 the system-level noise. Therefore, it is uncontrollable in this case. In order to reduce springback variation, the standard devi- ations used for variable randomization has to be decreased; level for the minimum springback variation suggested by Eqs. virtually, it means that a system-level adjustment of the press (5)–(7), which does not agree with Table 7. This discrepancy has to be performed to reduce the part-to-part variation of the was explained by the third method. equipment. On the other hand, if the springback variation is large and uncontrollable, then the springback compensation technique 5.3. Taguchi approach has to be chosen with it in mind. A methodology for the variation simulation of springback Taguchi analysis was used to analyze the springback vari- was developed, which provides a rapid understanding of the ation. MINITAB, a statistical software, was used to analyze influence of the random process variations on the springback the existing experiment results (Table 4). MINITAB can auto- variation of the formed part using FEA techniques eliminating matically extract data (standard deviation and mean) from the the need for lengthy and costly physical experiments. available experimental observations. For example, Table 9 is the extracted data of β1 used in MINITAB for Taguchi analysis. In References MINITAB, the main effects of each design factor on the standard deviations of the response are obtained via regression analysis, [1] W.D. Carden, L.M. Geng, D.K. Matlock, R.H. Wagoner, Measurement of and the significance of these effects were tested via analysis springback, Int. J. Mech. Sci. 44 (2002) (2002) 79–101. [2] A. Baba, Y. Tozawa, Effects of tensile force in stretch-forming process on of variation (ANOVA) and F-tests. P-values (P) were used to the springback, Bull. JSME 7 (1964) 835–843. determine which of the effects in the model are statistically sig- [3] Z.T. Zhang, D. Lee, Effects of process variables and material properties nificant, which are compared with a -level of 0.05. As shown on the springback behavior of 2D-draw bending parts, in: Automotive in Table 10, none of the effects are significant, which indicates Stamping Technology, SAE, 1995, pp. 11–18. that the springback variation in this case is not distinguishable [4] L.M. Geng, R.H. Wagoner, Role of plastic anisotropy and its evolution on springback, Int. J. Mech. Sci. 44 (1) (2002) 123–148. from the system-level noise. In other words, the springback [5] K. Mattiasson, A. Strange, P. Thilderkvist, A. Samuelsson, Simulation of variation is not controllable in this case. Therefore, the conclu- springback in sheet metal forming, in: 5th International Conference on sions based on Monte-Carlo simulation and sensitivity analysis Numerical Methods in Industrial Forming Process, New York, 1995, pp. 115–124. [6] R.H. Wagoner, W.P. WDCarden, D.K. Carden, Matlock, Springback after Table 10 drawing and bending of metal sheets, vol. 1, in: Proceedings of the IPMM F-tests for the standard deviation of each springback ’97—Intelligent Processing and Manufacturing of Materials, 1997, pp. P-value 1–10. [7] K.P. Li, L.M. Geng, R.H. Wagoner, Simulation of springback with the S.D. (β1 ) S.D. (β2 ) S.D. (ρ) draw/bend test, IPMM ’99, IEEE, Vancouver, BC, Canada, 1999, ISBN Material 0.283 0.088 0.853 0-7803-5489-3, p. 1. BHF 0.098 0.648 0.747 [8] S.W. Lee, D.Y. Yang, An assessment of numerical parameters influenc- Friction 0.635 0.495 0.542 ing springback in explicit finite element analysis of sheet metal forming process, J. Mater. Process. Technol. 80–81 (1998) 60–67.
  • 10. 198 P. Chen, M. Ko¸ / Journal of Materials Processing Technology 190 (2007) 189–198 c [9] K.P. Li, L. Geng, R.H. Wagoner, Simulation of springback: choice of ele- [24] T.W. Simpson, J.D. Peplinski, P.N. Koch, J.K. Allen, On the use of statistics ment Advanced Technology of Plasticity, vol. III, Springer, Berlin, 1999, in design and the implications for deterministic computer experiments, in: pp. 2091–2098. Proceedings of ASME DETC’97, 1977. [10] W.Y.D. Yuen, Springback in the stretch–bending of sheet metal with non- [25] W.J. Welch, W.K. Yu, S.M. Kang, J. Sacks, Computer experiments for uniform deformation, J. Mater. Process. Technol. 22 (1990) 1–20. quality control by parameter design, J. Qual. Technol. 22 (1) (1990) 15–22. [11] S.C. Tang, Analysis of springback in sheet forming operation Advanced [26] M.D. McKay, W.J. Conover, R.J. Beckman, A comparison of three methods Technology of Plasticity, vol. 1, Springer, Berlin, 1987, pp. 193–197. for selecting values of input variables in the analysis of output from a [12] L. Focellese, F. Fratini, M.F. Gabrielli, The evaluation of springback in 3D computer code, Technometrics 21 (1979) 239–245. stamping and coining processes, J. Mater. Process. Technol. 80–81 (1998) [27] J. Sacks, S.B. Schiller, W.J. Welch, Designs for computer experiments, 108–112. Technometrics 31 (1989) 41–47. [13] N. Narasimhan, M. Lovell, Predicting springback in sheet metal forming: [28] J.P.C. Kleijnen, Statistical Tools for Simulation Practitioners, Statistics an explicit to implicit sequential solution procedure, Finite Elements Anal. Textbooks and Monographs, vol. 76, M. Dekker, New York, 1987. Des. 33 (1999) 29–42. [29] A.A. Giunta, J.M. Dudley, R. Narducci, B. Grossman, R.T. Haftka, [14] D.W. Park, J.J. Kang, J.P. Hong, Springback simulation by combined W.H. Mason, L.T. Watson, Noisy aerodynamic response and smooth method of explicit and implicit FEM, in: Proceedings of NUMISHEET’99, approximations in high-speed civil transport design, vol. 2, in: 5th 1999, pp. 35–40. AIAA/USA/NASA/ISSMO Symposium on Multidisciplinary Analysis and [15] M. Valente, D. Traversa, Springback calculation of sheet metal parts after Optimisation, 1994, pp. 1117–1128. trimming and flanging, in: Proceedings of NUMISHEET, 1999, pp. 59–64. [30] A.A. Giunta, V. Balabanov, D. Haim, B. Grossman, W.H. Mason, L.T. [16] K.P. Li, W.P. Carden, R.H. Wagoner, Simulation of springback, Int. J. Mech. Watson, Wing design for a high-speed civil transport using a design of Sci. 44 (2002) 103–122. experiments methology, vol. 1, in: 6th AIAA/USA/NASA/ISSMO Sym- [17] S.C. Liu, S.J. Hu, Variation simulation for deformable sheet metal assem- posium on Multidisciplinary Analysis and Optimisation, 1996, pp. 168– blies using finite element methods, J. Manuf. Sci. Eng. Trans. ASME 119 183. (3) (1997) 368–374. [31] G. Venter, R.T. Haftka, J.H. Starnes, Construction of response surfaces for [18] S.D. Button, Determinant assembled stowage bins—a case study, Polym. design optimization applications, vol. 1, in: 6th AIAA/USA/NASA/ISSMO Compos. 20 (1) (1999) 86–97. Symposium on Multidisciplinary Analysis and Optimisation, 1996, pp. [19] S.C. Liu, S.J. Hu, An offset finite-element model and its applications in 548–564. predicting sheet-metal assembly variation, Int. J. Machine Tools Manuf. [32] K.D. Majeske, P.C. Hammett, Identifying sources of variation in sheet metal 35 (11) (1995) 1545–1557. stamping., Int. J. Flexible Manuf. Syst. 15 (2003) 5–18. [20] F.M. Swanstrom, T. Hawke, Design for manufacturing and assembly: a [33] M.G. Lee, D.Y. Kim, C.M. Kim, M.L. Wenner, K.S. Chung, Spring-back case study in cost reduction for composite wing tip structures, SAMPE J. evaluation of automotive sheets based on isotropic–kinematic hardening 36 (3) (2000) 9–16. laws and non-quadratic anisotropic yield functions. Part III. Applications, [21] R.J. Eggert, Design variation simulation of thick-walled cylinders, J. Mech. Int. J. Plasticity 21 (5) (2004) 915–953. Des. 117 (2) (1995) 221–228. [34] M.G. Lee, D.Y. Kim, C.M. Kim, M.L. Wenner, R.H. Wagoner, K.S. Chung, [22] S.D. Kini. An approach to integrating numerical and response surface mod- Spring-back evaluation of automotive sheets based on isotropic–kinematic els for robust design of production systems, Ph.D. Thesis, The Ohio State hardening laws and non-quadratic anisotropic yield functions. Part II. Char- University, 2004. acterization of material properties, Int. J. Plasticity 21 (5) (2004) 883–914. [23] J. Sacks, W.J. Welch, T.J. Mitchell, H.P. Wynn, Design and analysis of [35] L. Papeleux, J.P. Ponthot, Finite element simulation of springback in sheet computer experiments, Stat. Sci. 4 (4) (1989) 409–435. metal forming, J. Mater. Process. Technol. 125–126 (2002) 785–791.