Sunteți pe pagina 1din 11

Microforming: Experimental

Investigation of the Extrusion


Process for Micropins and its
Numerical Simulation Using
Jian Cao RKEM
Neil Krishnan
Microforming using a small machine (or so-called desktop machine) is an alternative new
Zhong Wang approach to those using full-size heavy equipment for manufacturing microparts. Micro-
parts are commonly defined as parts or structures with at least two dimensions in the
Hongsheng Lu submillimeter range, which are used extensively in electronics and micromechanical pro-
ducts. However, when scaling down a conventional forming process to microscale, the
Wing K. Liu influence of the so-called size effect needs to be considered. The individual microstructure
(size, shape, and orientation of grains) and the interfacial conditions show a significant
Anthony Swanson effect on the process characteristics. In this paper, the process of extrusion is investigated
to establish it as a viable process for microforming. A forming assembly is fabricated and
Department of Mechanical Engineering, used in conjunction with a loading substage to extrude micropins with a final diameter of
Northwestern University, 2145 Sheridan Road, 1 mm. The effect of grain size is investigated by using workpieces heat treated to produce
Evanston, IL 60208 grain sizes varying from 32 ␮m up to 211 ␮m. Two extrusion dies with different roughness
are used to study the effect of surface finish. While experiments lead to interesting ques-
tions and new discoveries, theoretical or numerical solutions are necessary tools for
process optimization. Here, knowing the limits of the current widely used numerical simu-
lation tools [i.e., the Finite Element Method (FEM)], a new method, the Reproducing
Kernel Element Method (RKEM), has recently been developed to address the limitations
of the FEM (for example, remeshing issue), while maintaining FEM’s advantages, e.g., the
polynomial reproducing property and function interpolation property. The new RKEM
method is used to simulate the microextrusion problem. Its results are compared with that
obtained from the FEM and the experiment result. Satisfactory results were obtained.
Future directions on the experimental and simulation work are addressed.
关DOI: 10.1115/1.1813468兴

1 Introduction large share of volume occupied by an individual grain. Thus, the


microforming process limitations are largely influenced by the
In the last decade, the emergence of miniaturization technolo-
workpiece dimensions, and this is commonly referred to as the
gies has revolutionized several fields of technology by leading to
size effect. Several researchers have tried to study the effects of
the development of innovative applications in industries such as
size on material behavior and surface interactions. Tiesler and
electronics, healthcare, automobiles, environmental monitoring,
Engel 关2兴 investigated the effect of miniaturization on bulk metal
etc. This trend toward miniaturization and higher integrated func-
forming through a series of double-cup extrusion experiments per-
tional density in devices has produced an increased demand for
formed on CuZn15 brass alloy. They used samples varying in
metallic microparts having small dimensions and close tolerances.
dimension from 0.5 mm to 4 mm, which were scaled in order to
One example of such microparts at the mesoscale 共defined as hav-
retain geometric similarity. The authors found that decreasing the
ing dimensions of the order of 100 s of micrometers to 10 mm兲 is
the IC connector pin. It is estimated that 24 billion pins are used sample dimensions increased the friction factor by 20 times. They
annually for IC carriers. Examples of these microparts, such as were also able to determine that an increase in the workpiece
connector pins, miniature screws, contact springs, and pins for IC surface roughness and larger 共150 ␮m versus 50 ␮m兲 coarse
sockets, are shown in Fig. 1. These parts have several functional grains in the sample caused the friction factor to increase.
differences and are made using a variety of fabrication processes. Raulea et al. 关3兴 investigated the effects related to the ratio of
A comprehensive review of the field of microforming can be aluminum grain size to sheet thickness in two experiments,
found in the work of Geiger et al. 关1兴. The authors present several uniaxial tensile tests in which the grain size was kept constant and
key aspects of microforming, such as the applications of such the sample thickness was varied, and bending experiments in
technology and products; interesting material behavior observed which the grain size was varied for constant specimen thickness.
at microscale; various processes involved in the fabrication of The results showed a decrease of yield stress with a decreasing
microscale parts; and the problems encountered with microscale number of grains over the thickness up to one grain over the
tooling, assembly, and handling. thickness, and beyond this 共i.e., for grain sizes larger than the
When a forming process is scaled down to microdimensions, thickness兲 the yield strength tends to increase with an increasing
the microstructure of the workpiece, the surface topology of the grain size. Similar results were also observed by Geiger et al. 关4兴
workpiece, and that of the tooling remain unchanged. The material who performed experimental investigations to study the size effect
of the workpiece cannot be considered as a continuum due to the seen in scaling down a forming process. The authors conducted a
series of tensile tests on cylindrical samples with a diameter of
Contributed by the Manufacturing Engineering Division for publication in the
0.2–2 mm and upsetting tests in which outer diameters changed
JOURNAL OF MANUFACTURING SCIENCE AND ENGINEERING. Manuscript received from 1 mm to 8.5 mm on geometrically similar samples of
March 29, 2004; Revised August 2, 2004. Associate Editor: J. Ni. CuZn15 brass alloy. The authors demonstrated that decreasing

642 Õ Vol. 126, NOVEMBER 2004 Copyright © 2004 by ASME Transactions of the ASME
mended the use of a kinematically admissible velocity field based
on a simple radial flow field combined with the asymptotic behav-
ior near a velocity-discontinuity surface to solve axisymmetric
direct and indirect extrusion problems. The shape of the dead zone
and the extrusion pressure were investigated for varying extrusion
ratios. The results for the upper bound on the extrusion pressure
were compared with other solutions based on radial flow only, and
it was shown that the asymptotic velocity field improved the pre-
diction of the extrusion pressure. Arentoft et al. 关14兴 investigated
the extrusion process with physical modeling and compared the
findings of their study with finite element predictions for different
reduction ratios and extrusion geometries. In order to perform the
physical modeling, the authors substituted the real metal to be
plastically formed with a soft model material, which behaved in a
Fig. 1 Examples of various metallic microparts used in manner analogous to the real metal. The commercial FEM code
industry DEFORM was used to run simulations of the extrusion process,
and a good comparison between the experiments and modeling
results was observed.
Numerical methods are also used to investigate the effect of
grain size caused an increase of flow stress. By using different
temperature on the final product of extrusion. The temperature
cross-sectional geometries in the tensile tests, the authors also
profile of an extruded part influences several parameters from sur-
demonstrated that an additional decrease in flow stress occurs due
face finish to mechanical properties. Sheppard 关15兴 presented a
to the increase in the share of surface grains in a sample. Com-
review of the various techniques used to calculate the temperature
pression tests on ring specimens were used to deduce that dry
friction showed a dependence on specimen size and average grain during extrusion. The author discussed several existing methods,
size. The authors also observed that the scattering of data in- such as analytical models, integral profile models, and finite ele-
creased when dimensions of samples were scaled down. ment models. Results are also presented comparing the integral
Cold microforming is limited by the dependence of the final profile method to the finite element method for an extrusion ratio
result on the grain size and the orientation of grains. This causes of 30:1, and these results were seen to be extremely close. Chanda
problems, such as inhomogenous deformation and large scatter, in et al. 关16兴 used FEM simulations to compare the extrusion pres-
results that limits product reproducibility. To address this issue, sure, thermal response, and stress state of iso-speed extrusion ver-
Engel et al. 关5兴 investigated the process of warm forming for the sus a stepwise ram speed change. Using a reduction ratio of 20:1,
manufacture of microparts. Warm forming is defined as forming the authors were able to demonstrate that a stepwise ram speed
performed at a temperature above that of room temperature where decrease enabled the temperature of the extrudate to reach a
hardening effects still occur and recrystallization does not take steady value, which corresponded to an almost constant exit tem-
place during forming. The authors performed upsetting, backward perature. This reduced the extrusion pressure and the tensile
can extrusion, and lateral extrusion on 0.5 mm diam samples of stresses acting on the extrudate, correspondingly reducing the
CuZn15 brass and X4CrNi18-10 steel from temperatures of 20°C chances of failure. Also, it was reasoned that a constant exit tem-
to 300°C. They were able to conclude that warm microforming perature for an extrudate would be advantageous in preventing
gives an improved material flow, a homogenized hardness distri- thermally induced phase changes, which are undesirable in certain
bution and a reduction of scatter in process parameters. Heating- metal alloys 共such as Al6061兲. Li et al. 关17兴 used three-
assisted microstamping was also investigated by Balendra and Qin dimensional 共3D兲 FEM simulations to study the temperature evo-
关6兴 who studied the feasibility of using localized heating to lution of 7075 aluminum alloy during direct extrusion. Al 7075 is
achieve higher aspect ratios and improved microstamping for one of the most-used, high-strength alloys for aircraft structural
high-strength materials. Using a high-power laser beam and elec- components and is considered to be extremely difficult to extrude.
trical heating, the desired temperature distribution was achieved in Using an extrusion ratio of 8:1, the authors performed simulations
both copper and steel-type work materials. The authors were able at various extrusion speeds to observe the evolution of extrusion
to conclude that localized heating enabled the stamping-force re- pressure and temperature distribution. The authors were able to
quirements to be reduced and correspondingly, the aspect ratios conclude that defects caused by incipient melting would occur
achievable with stamping were increased. even at a low ram speed of 1 mm/s, if the billet was not homog-
Saotome and Iwazaki 关7兴 investigated the applicability of the enized for 24 hours and subsequently fast-cooled. With such an
backward extrusion process to produce microparts for microelec- improvement in the metallurgical conditions of the billet, higher
tromechanical systems, such as microgears, below 50 ␮m module. ram speeds up to 8 mm/s could be used with no risk of melting.
Photochemically machinable glass was used to fabricate micro- All the above-mentioned analysis work on extrusions were per-
dies by photolithography and anisotropic etching. Al-78Zn super- formed at a scale level of 50⬃100 mm, much larger than the
plastic material and LaAlNi were used as the workpiece materials. micropins that were studied here.
These materials exhibited good microformability and deformed Tedious remeshing procedures are necessary in the conven-
under low applied stresses. The authors were successfully able to tional finite element method to deal with large deformation and
produce microgears of LiAlNi amorphous alloy with a 50 ␮m the motion of interfaces. Although the remeshing can be avoided
module and Al-78Zn superplastic alloy microgears, with a module in the meshfree method 关18兴, the enrichment along the interface to
of 20 ␮m and a pitch circle of 200 ␮m. obtain the material discontinuity in the numerical result is also
The ability to perform numerical simulations for the extrusion very complicated. The reproducing kernel element method
of micropins is important to better understand the microforming 共RKEM兲 关19–21兴, which has the built-in properties of FEM and
process and assist the design and decision-making processes. Con- meshfree methods, can effectively handle the forming process
ventional finite element methods are ubiquitous in current engi- with large deformation. It is easy to locally remesh the model in
neering analysis of problems, such as extrusion. Several methods the RKEM due to the natural conforming of the approximation. A
have been used to characterize the extrusion process, from ana- structured mesh required in the FEM is not necessary in RKEM.
lytical models 关8,9兴 to numerical methods, such as FEM 关10–12兴. The discontinuity in the shape function along the material inter-
In recent years, researchers have used various approaches to in- faces is accomplished by cutting off the global partition polyno-
vestigate the extrusion process. Alexandrov et al. 关13兴 recom- mial and the kernel function. No enrichment along the interface is

Journal of Manufacturing Science and Engineering NOVEMBER 2004, Vol. 126 Õ 643
Fig. 2 „a… Microstructure of brass sample after heat treatment
at 610°C for 1 h, grain sizeÄ87 ␮m; „b… microstructure of brass
sample after heat treatment at 700°C for 1 h, grain sizeÄ211 ␮m

needed. Moreover, the efficiency of the RKEM can be further


improved with using the stabilized conforming nodal integration Fig. 3 Schematic of extruded pin with dimensions for the two
cases: larger pin and smaller pin
关22兴.
In this paper, a novel microforming experiment setup is used to
characterize the effect of grain size and surface roughness on the
extrusion process for micropins. A forming assembly was fabri- Slip lines due to cold working can also be seen in the deformed
cated to conduct extrusion tests on microscale samples in order to grains. Figure 4共b兲 only has 10 grains across the diameter of the
characterize the process by measuring the force-displacement re- pin, and therefore, very few deformed grains can be clearly seen.
sponse. The forming assembly is mounted in a tensile substage, Slip lines are far more pronounced in the sample. Similarly, Figs.
which is capable of performing quasi-static tests in tension or 5共a兲 and 5共b兲, which again are geometrically similar and use the
compression. The tests are carried out using workpieces with dif- same material, show a very different deformation pattern. Only
ferent grain sizes and dies with different surface roughness. The four grains exist across the diameter of the pin in Fig. 5共b兲, and it
feasibility of using the RKEM to model the extrusion process is is extremely difficult to spot the individual deformed grains. In-
also investigated. A review of the RKEM is discussed in Section terestingly, if we compare Figs. 4共b兲 and 5共a兲 the deformation
4. Preliminary results for use of the RKEM to simulate micro- patterns are similar, but they have different geometry and different
forming are also presented and compared to conventional FEM grain sizes. The common point between these two cases is the
results. The paper ends with a brief section outlining our conclu- ratio of blank diameter to grain size. Both samples have 10 grains
sions and the direction of our future research in this field. across the diameter of the pins. These results show the relevance
of microstructure in modeling the microextrusion process.
Tensile tests were also performed to observe the effect of grain
2 Effect of Grain Size on Microforming
size and sample size on the flow stress of the material. Using
To study the impact of grain size on the extrusion of small pins, appropriate heat treatment, samples of different grain size were
two material samples of different grain size were created. Car- made. The grain sizes obtained were 32, 87, and 211 ␮m. Samples
tridge brass 共Cu/Zn: 70/30兲 rods were heat treated for 1 h at tem- of 32 ␮m grain size were produced by heat treating the brass at
peratures of 610°C and 700°C. This heat treatment was performed 550°C for 1 h. This produced grains having an average grain size
to remove the effect of cold working in the base material and to of 32 ␮m with a minimum size of 13 ␮m, a maximum size of 133
produce new grains of appropriate size. The heat treatment at ␮m, and a standard deviation of 34 ␮m. For each particular grain
610°C yielded samples having an average grain size of 87 ␮m size, three tensile test specimens of different dimensions were
with a minimum size of 34 ␮m, a maximum size of 264 ␮m, and
a standard deviation of 67 ␮m. The heat treatment at 700°C
yielded samples having an average grain size of 211 ␮m with a
minimum size of 45 ␮m, a maximum size of 432 ␮m, and a
standard deviation of 106 ␮m. The corresponding microstructures
are shown in Fig. 2. These microstructures were obtained by
sample polishing and etching using the NH4 OH/H2 O2 solution of
the appropriate concentration. The material samples prepared in
this manner were used to extrude pins of two sizes, a larger pin
with a base diameter of 2 mm and an extruded diameter of 1.2 mm
and a smaller pin with a base diameter of 0.8 mm and an extruded
diameter of 0.48 mm. The dimensions are chosen such that the
two pins are geometrically similar and the larger pin is scaled Fig. 4 „a… Deformed grains for the larger pin, initial grain
down by a factor of 2.5 to obtain the dimensions of the smaller sizeÄ87 ␮m; „b… deformed grains for the smaller pin, initial
pin. Figure 3 shows the pin geometry and the corresponding pin grain sizeÄ87 ␮m „scale barÄ50 ␮m…
dimensions. After the pins were extruded, the microstructures of
the deformed pins were investigated. The samples were polished
and etched, and the resulting grain structure was observed at a
position marked as location A in Fig. 3. The deformed microstruc-
tures for the smaller grain size are shown in Fig. 4, and those of
the larger grain size are shown in Fig. 5. In Fig. 4, the original
material is the same and the pins are geometrically similar. There-
fore, if we ignore the effect of grain size, the pins should have the
same deformation characteristics. However, such an assumption
cannot be made by comparing Figs. 4共a兲 and 4共b兲. In Fig. 4共a兲 Fig. 5 Deformed grains for the larger pin, initial grain size
there are 25 grains across the diameter of the pin. Thus, deforming Ä211 ␮m; deformed grains for the smaller pin, initial grain size
grains can be seen as one moves away from the exterior of the pin. Ä211 ␮m „scale barÄ50 ␮m…

644 Õ Vol. 126, NOVEMBER 2004 Transactions of the ASME


Fig. 8 Segmented die used for extrusion of micropins

Fig. 6 True stress versus true strain „211 ␮m grain size…


segmented to facilitate the removal of the pin after extrusion. The
dimensions of the new pin manufactured with this die are a
fabricated. The tensile specimens had diameters of 0.4, 0.8, and ⫽1.5 mm, b⫽1.0 mm, and ␪⫽30° 共as shown in Fig. 3兲. The seg-
1.32 mm. A sample tensile specimen is shown as an inset in Fig. mented die is then mounted onto a specially designed forming
6. The samples were tested using a Sintech universal testing ma- assembly shown in Fig. 9. The assembly consists of a ram
chine with a 2500 N capacity load cell. The displacement was mounted on a yoke that slides along linear bearings, which guide
measured using an extensometer mounted directly on the test the ram into the segmented die that is mounted and clamped in a
specimens. Two tests were performed for samples of diameter 0.8 die block. The guide rods for the yoke, the ram, and the seg-
mm and 1.32 mm, and one test for the sample with the diameter of mented die are all hardened, and their surfaces were ground to a
0.4 mm. The stress strain curves obtained from these tensile tests tolerance of ⫾0.01 mm. The die cavity in which the extrusion dies
are shown in Fig. 6 for the material having a grain size of 211 ␮m. are mounted was made using EDM. The forming assembly is then
The figures in the legend of the graph give the diameter of the test placed inside a loading substage 共Fig. 9兲, which is equipped with
specimen and the number of grains across the diameter. For a a load cell 共capacity 2000 lbs兲 and an LVDT to measure the ex-
given grain size, the stress-strain curves do not show the effect of trusion force and the corresponding ram displacement. Pins were
decreasing sample size even down to a diameter of 0.4 mm. The extruded from three material samples having grain sizes of 32, 87,
lack of sample size effect is also visible in the tensile test results and 211 ␮m, respectively. The initial workpieces were fabricated
for the material with grain sizes of 32 ␮m and 87 ␮m. As ex- in the same manner to ensure that they have the same surface
pected, if all the graphs are plotted together, the effect of grain roughness, which was measured to be 3.5 ␮m 共rms兲. Two dies
size on the flow stress can be clearly seen 共Fig. 7兲. The graphs can were used in the extrusion process, a rough die made by drilling
be distinctly grouped into three sets, each representing a particular and a smooth die made by drilling followed by polishing using a
grain size as marked on the figure. The trend of decreasing grain fine lapping compound to produce a better surface finish. The
size causing an increase in the flow stress can be clearly seen, but rough die has an route-mean-square surface roughness of 1.82 ␮m
the influence of the sample diameter on the measurement of flow in the 1.5 mm dia region and a rms surface roughness of 2.58 ␮m
stress is not observed here. This may be due to the fact that the in the 1 mm dia region, whereas the corresponding values for the
diameter of the tensile specimen is still at least twice that of the smooth die are 1.13 ␮m and 1.01 ␮m. For each combination of
grain size. Current work is ongoing to reduce the size of tensile die 共rough or smooth兲 and workpiece 共grain size 32, 87, and 211
specimens. ␮m兲, three extrusion tests were performed to check the repeatabil-
ity. After each test, the surface roughness of the die was measured
3 Force Displacement Response for the Microextru- again to observe changes in surface texture due to friction. All
sion Process surface roughness values were measured using a MicroXAM
surface-mapping microscope.
In order to measure the force-displacement response for the
extrusion of micropins, segmented dies were fabricated using con- 3.1 Extrusion Results Using the Smooth Die. Figures
ventional drilling and polishing as shown in Fig. 8. The die is 10–12 show the force-displacement response obtained during the
extrusion of micropins using the smooth die and workpieces with
grain sizes of 32, 87, and 211 ␮m, respectively. The results show
a fair degree of repeatability, and the trend is observed to be
similar in all cases. The average value of the maximum extrusion
force is 4.11 KN for the 32 ␮m samples, 3.67 KN for the 87 ␮m

Fig. 7 True stress versus true strain „all grain sizes… Fig. 9 Forming assembly and the tensile loading substage

Journal of Manufacturing Science and Engineering NOVEMBER 2004, Vol. 126 Õ 645
Fig. 10 Force-displacement response for extrusion of 32 ␮m
Fig. 13 Average and rms surface roughness after each test at
grain-size sample using smooth die
the 1.5 mm region of the smooth die

samples, and 3.64 KN for the 211 ␮m samples. As expected, the


force increases as the pin gets longer. The graphs also show the
extrusion force decreases with increasing grain size, although the
force-displacement response for unloading, which corresponds to
difference between the 87 and 211 ␮m results is very slight. The
the compliance of the forming assembly and the loading substage.
shape of the curve also matches well with the general response
The compliance appears to be the same, and, hence, no compli-
observed for extrusion. The initial portion of the curve, up to a
ance correction was performed. Figures 13 and 14 show the sur-
ram displacement of 0.8 mm, represents the expansion of the
face roughness of the smooth die in the 1.5 mm diam region and
workpiece to occupy the entire die cavity. From 0.8 to 1.5 mm
the 1 mm diam region, respectively, after each test is performed.
ram displacement, flow of material into the conical section of the
The graphs show error bars for the range of readings, both average
die starts taking place, and, therefore, the extrusion force increases
surface roughness and rms surface roughness are plotted, and the
rapidly. Beyond 1.5 mm, the 1 mm diam pin has already started to
grain size of each individual test is marked on the graphs. It is
be formed and length of the pin is increasing. Hence, the slope of
evident from these graphs that no appreciable change in surface
the curve is more gradual, mainly due to the fact that the friction
roughness occurs during testing.
3.2 Extrusion Results Using the Rough Die. Figures
15–17 show the force-displacement response obtained during the
extrusion of micropins using the rough die and workpieces with
grain sizes of 32, 87, and 211 ␮m, respectively. In this case, the
extrusion force for the first test is seen to be distinctly lower than
those of subsequent tests in all three graphs. This can be corre-
lated to the surface roughness readings. Figures 18 and 19 show
the surface roughness readings for the 1.5 mm region and the 1.0
mm region of the smooth die. It can be seen that the surface
roughness drops dramatically during the first set of tests and then
stabilizes to an approximately constant level. If the first set of
results is excluded from consideration, the results are similar to
those of the previous case and the values of the extrusion force are
approximately the same. This is in accordance with the observed
trends for surface roughness because the surface roughness de-
creases after the initial set of tests to a value similar to that of the
Fig. 11 Force-displacement response for extrusion of 87 ␮m previous case.
grain-size sample using smooth die

Fig. 12 Force-displacement response for extrusion of 211 ␮m Fig. 14 Average and rms surface roughness after each test at
grain-size sample using smooth die the 1.0 mm region of the smooth die

646 Õ Vol. 126, NOVEMBER 2004 Transactions of the ASME


Fig. 15 Force-displacement response for extrusion of 32 ␮m Fig. 18 Average and rms surface roughness after each test at
grain-size sample using rough die the 1.5 mm region of the rough die

Figure 20 shows an example of the deformed pin made during roughness values are lower than those of the corresponding dies,
extrusion. The final dimensions for all the pins made are listed in and they are fairly consistent across the grain sizes and die types
Table 1 along with mean and standard deviation 共SD兲 values on used in the experiments.
the basis of nine readings for each pin type. The dimensions of the
pins are slightly larger than the corresponding die dimensions pos- 4 Reproducing Kernel Element Method
sibly due to die compliance or the separation of the segmented Computational simulation has become a valuable tool to pro-
dies during forming. All SD values are around 0.01 mm. Table 2 vide valuable insights, such as material flow, die fill, and strain/
lists the surface roughness values for all pins measured using the stress distribution, in the design of forming process. This brings a
surface mapping microscope. The mean and standard deviation significant improvement in the die design and process control,
values are reported for six readings of each pin type. The surface which are traditionally based on costly time-consuming trial-and-
error experiments. Many FEM simulations have been carried out
to improve the knowledge of the forming process. Severe mesh
distortions usually occur in the simulation of the forming process;
therefore, a remeshing or rezoning procedure 关23兴 is necessary to
continue the analysis. The remeshing procedure is a formidable

Fig. 16 Force-displacement response for extrusion of 87 ␮m


grain-size sample using rough die

Fig. 19 Average and rms surface roughness after each test at


the 1.0 mm region of the rough die

Fig. 17 Force-displacement response for extrusion of 211 ␮m


grain-size sample using rough die Fig. 20 Sample of the deformed pins

Journal of Manufacturing Science and Engineering NOVEMBER 2004, Vol. 126 Õ 647
Table 1 Dimensions of the deformed pins

Grain size a (mm) b (mm)


Die 共␮m兲 Mean⫾SD Mean⫾SD
Smooth 32 1.57⫾0.009 1.05⫾0.009
Smooth 87 1.56⫾0.012 1.05⫾0.009
Smooth 211 1.57⫾0.012 1.06⫾0.008
Rough 32 1.55⫾0.010 1.08⫾0.006
Rough 87 1.56⫾0.012 1.08⫾0.011
Rough 211 1.55⫾0.010 1.07⫾0.013

task to undertake, especially in three dimensions, due to a struc-


tured mesh required in FEM. The meshfree method 关24 –27兴 can
easily handle large deformation without the need of using the Fig. 21 3D hexahedral sub-domain element
explicit mesh, however, an extra effort is needed to deal with the
boundary condition. A new class method, the Reproducing Kernel
Element Method 共RKEM兲 关19–21兴 has been recently developed.
This method takes the salient features of the FEM and meshfree 2. The smoothness of the shape function ⌿ I (x) is solely deter-
method. It can deal with large deformation and make the remesh- mined by that of the kernel function.
ing procedure much simpler. The RKEM method is illustrated 3. The shape functions ⌿ I (x) of RKEM have the Kronecker
briefly below. delta property at the associated node I, when the support size
Let ⍀傺R d be an open, bounded domain with a Lipschitz con- of the node is properly chosen, such that the influence do-
tinuous boundary ⌫⫽⳵⍀. The domain ⍀̄⫽⍀艛⌫ is decomposed main of the node only covers the elements that include the
node.
into a set of subdivisions 兵 ⍀ n 其 n⫽1
N
. On each subdomain ⍀ n , we
assume that for some integer I n ⭓1, there are the linearly indepen- A systematic procedure to construct the global partition poly-
In
dent global partition polynomial 兵 ␺ n,i 其 i⫽1 and the corresponding nomial is given in 关20,21兴. If a hexahedral element 共Fig. 21兲 is
In chosen as the subdomain, the global partition polynomial with
nodes 兵 xn,i 其 i⫽1 傺⍀ n , such that the following reproducing prop-
trilinear reproducing property can be constructed from Eq. 共1兲, as
erty of order k holds:
In

兺␺ ␥
n,i 共 x 兲 xn,i ⫽x

᭙ ␥ : 兩 ␥ 兩 ⭐k, ᭙x苸⍀̄ (1) 8
i⫽1
兺␺
i⫽1
n,i 共 x 兲 ⫽1
In the RKEM, the continuous global partition polynomial 8

¦
兺␺
In
兵 ␺ n,i 其 i⫽1 is patched over the whole domain; and the compactly
n,i 共 x 兲 x n,i ⫽x
reproducing kernel function is used to localize the global partition i⫽1
polynomial such that the required reproducing condition is satis- 8
fied. The following 共quasi兲interpolation operator on a continuous
function is proposed as
兺␺
i⫽1
n,i 共 x 兲 y n,i ⫽y

冋冕 册
8

兺␺
N In
n,i 共 x 兲 z n,i ⫽z
兩 v共 x兲 ⫽ 兺
n⫽1 ⍀n
K ␳ 共 y⫺x;x兲 dy 兺␺
i⫽1
n,i 共 x 兲v共 xn,i 兲 i⫽1
(3)
8
NP
兺 ␺ n,i 共 x兲 x n,i y n,i ⫽xy
⫽ 兺 ⌿ 共 x兲v
I⫽1
I I (2) i⫽1
8

where K ␳ (z;x) is a kernel function which is nonzero only when 兺␺


i⫽1
n,i 共 x 兲 x n,i z n,i ⫽xz
储z储⬍␳. The positive number ␳ represents the support size of the 8
kernel function with respect to its first argument. N P is the num-
ber of nodes in the domain, and ⌿ I (x) is the reproducing kernel 兺␺
i⫽1
n,i 共 x 兲 y n,i z n,i ⫽yz

element shape function of node I. As shown in 关19兴, there are 8


several distinguished properties of this new method.
兺␺
i⫽1
n,i 共 x 兲 x n,i y n,i z n,i ⫽xyz
1. The interpolation operator 兩 defined in Eq. 共2兲 can maintain
the reproducing property of global partition polynomials if
the operator owns the zeroth reproducing property. i.e.,

Table 2 RMS „ R q … and average „ R a … surface roughness of the extruded pins

Grain size R q @ 1.5 mm R a @ 1.5 mm R q @ 1.0 mm R a @ 1.0 mm


Die 共␮m兲 Mean⫾SD 共␮m兲 Mean⫾SD 共␮m兲 Mean⫾SD 共␮m兲 Mean⫾SD 共␮m兲
Smooth 32 0.82⫾0.10 0.56⫾0.06 0.50⫾0.08 0.37⫾0.07
Smooth 87 0.87⫾0.06 0.60⫾0.09 0.55⫾0.05 0.40⫾0.02
Smooth 211 0.78⫾0.05 0.57⫾0.05 0.54⫾0.07 0.35⫾0.04
Rough 32 0.80⫾0.03 0.56⫾0.04 0.48⫾0.05 0.37⫾0.05
Rough 87 0.87⫾0.04 0.53⫾0.03 0.51⫾0.02 0.40⫾0.03
Rough 211 0.86⫾0.04 0.58⫾0.05 0.55⫾0.11 0.40⫾0.06

648 Õ Vol. 126, NOVEMBER 2004 Transactions of the ASME


冤 冥冤 冥冤 冥
⫺T
1 x n,1 y n,1 z n,1 x n,1y n,1 x n,1z n,1 y n,1z n,1 x n,1y n,1z n,1
␺ n,1共 x兲 1 x n,2 y n,2 z n,2 x n,2y n,2 x n,2z n,2 y n,2z n,2 x n,2y n,2z n,2 1
␺ n,2共 x兲 x
1 x n,3 y n,3 z n,3 x n,3y n,3 x n,3z n,3 y n,3z n,3 x n,3y n,3z n,3
␺ n,3共 x兲 y
␺ n,4共 x兲 1 x n,4 y n,4 z n,4 x n,4y n,4 x n,4z n,4 y n,4z n,4 x n,4y n,4z n,4 z
⫽ (4)
␺ n,5共 x兲 1 x n,5 y n,5 z n,5 x n,5y n,5 x n,5z n,5 y n,5z n,5 x n,1y n,1z n,1 xy
␺ n,6共 x兲 1 x n,6 y n,6 z n,6 x n,6y n,6 x n,6z n,6 y n,6z n,6 x n,1y n,1z n,1 xz
␺ n,7共 x兲 yz
␺ n,8共 x兲 1 x n,7 y n,7 z n,7 x n,7y n,7 x n,7z n,7 y n,7z n,7 x n,1y n,1z n,1 xyz
1 x n,8 y n,8 z n,8 x n,8y n,8 x n,8z n,8 y n,8z n,8 x n,1y n,1z n,1

冤 冥
The global partition polynomial exists as long as the hexahedral b̃ I1 共 xL 兲 0 0
subdomain is well defined. As in the meshfree methods, the ap-
proximation of reproducing kernel element is naturally conform- 0 b̃ I2 共 xL 兲 0
ing. The mesh used to construct the reproducing kernel element 0 0 b̃ I3 共 xL 兲
interpolant is not necessary to be a structured mesh required in B̃I 共 xL 兲 ⫽ (10)
FEM. Therefore, it is suitable for the locally remeshing which is b̃ I2 共 xL 兲 b̃ I1 共 xL 兲 0
difficult to accomplish in conventional finite element methods.
0 b̃ I3 共 xL 兲 b̃ I2 共 xL 兲
4.1 Stabilized Nodal Integration in RKEM. A strain-
smoothing technique was developed by Chen et al. 关22兴 as a regu- b̃ I3 共 xL 兲 0 b̃ I1 共 xL 兲
larization for material instabilities in strain localization problems.
The modifying strain smoothing at nodal points was proposed by
Chen et al. 关22兴 in the stabilized nodal integration as
b̃ Ii 共 xL 兲 ⫽
1
AL 冕 ⌫L
⌿I 共 x兲 n i 共 x兲 d⌫ (11)

As shown in 关22兴, the following integration constraints are sat-


˜␧ ihj 共 xL 兲 ⫽ 冕

␧ ihj 共 x兲 ⌽ 共 x;x⫺xL 兲 d⍀
isfied when the smoothed strain is employed into the Galerkin
approximation with nodal integration.


NP


共 u i,h j ⫹u hj,i 兲
2
⌽ 共 x;x⫺xL 兲 d⍀ (5) 兺 B̃ 共 x 兲 A ⫽0
L⫽1
I L L for all interior nodes

兵 I:supp共 ⌿ I 兲 艚⌫⫽0 其 (12)


where ␧ ihj is the strain obtained from displacement approximation, NP N Ph
and ⌽ is a distribution function. In this work, a piecewise constant
distribution function is chosen, 兺 B̃ 共 x 兲 A ⫽ 兺 N 共 x 兲 w
L⫽1
I L L
L⫽1
I L L for boundary nodes

再 1/A L , x苸⍀ L 兵 I:supp共 ⌿ I 兲 艚⌫⫽0 其 (13)


⌽ 共 x;x⫺xL 兲 ⫽ (6) where N Ph is the number of integration points on the natural
0, x苸⍀ L
boundary and w L is the weight of boundary integration.
Therefore, the accuracy and convergence for nodal integration
where ⍀ L is the representative domain of node L obtained from, is ensured. The performance of stabilized nodal integration in lin-
for example, the Voronoi diagram in Fig. 22 and A L is the volume ear and nonlinear problems can be found in 关22兴. Although the
of the representative domain ⍀ L . Applying the divergence theo- strain smoothing technique can stabilize spurious modes in most
rem, Eq. 共5兲 can be expressed as cases with nodal integration, the hourglassing appears in some
special cases. For example, in the extrusion simulation as shown
˜␧ ihj 共 xL 兲 ⫽
1
2A L 冕 ⍀L
共 u i,h j ⫹u hj,i 兲 d⍀⫽
1
2A L 冕 ⌫L
共 u ih n j ⫹u hj n i 兲 d⌫
in Fig. 23, the hourglass mode appears around the corner due to
the contact constraint and then spreads to the whole domain as the
deformation goes further. To avoid these kinds of spurious modes
(7)

where ⌫ L is the boundary of the representative domain ⍀ L and n i


is the surface normal. Note in the previous strain-smoothing tech-
nique, the smoothed strain at the nodal point is exact for the
constant strain field. Plugging the reproducing kernel element
shape function into Eq. 共7兲 yields

˜␧h 共 xL 兲 ⫽ 兺
I苸G L
B̃I 共 xL 兲 uI (8)

where G L is a group of nodes whose shape function influences


node L. For 3D, the matrices ˜␧ h , B̃I (xL ), and uI take the form

˜␧h 共 xL 兲 ⫽ 关 ˜␧ h11 ,␧
˜ h22 ,␧
˜ h33,2␧
˜ h12,2␧
˜ h23,2␧
˜ h31兴 T uI ⫽ 关 u 1I ,u 2I ,u 3I 兴 T
(9) Fig. 22 Representative domain of node L

Journal of Manufacturing Science and Engineering NOVEMBER 2004, Vol. 126 Õ 649
Fig. 23 Deformations of extrusion in 2D. The development of
the hourglass mode can be clearly seen.

in such special cases, the representative domain of node L is fur-


ther decomposed into several subdivisions as shown in Fig. 24,
and the center of each subdivision is used as an integration point. Fig. 24 Subdivision and its center in the representative do-
The strain smoothing technique is applied to obtain the smoothed main of node L
strain at each integration point as follows:

˜␧ ihj 共 xLk 兲 ⫽
1
2A Lk
冕k
⍀L
共 u i,h j ⫹u hj,i 兲 d⍀⫽
1
2A Lk

⌫L
k
共 u ih n j ⫹u hj n i 兲 d⌫
effect of microstructure was ignored. The results show a good
(14) match in terms of the ability of RKEM to accurately capture the
deformation characteristics. Figure 26 shows a plot of the force
where xLk is the center of the subdivision ⍀ Lk of the representative
displacement response obtained from the two numerical simula-
domain of node L. ⌫ Lk is the boundary of the subdivision ⍀ Lk , and tions in comparison to the actual experimental result. The force-
A Lk is the volume of the subdivision ⍀ Lk . Similarly, the integration displacement curves predicted by the RKEM compare well with
constraints are satisfied when the smoothed strain is employed both the FEM results and the experiment results.
into the Galerkin approximation with multiple integration points
are used for the representative domain of each node.
NP NL
5 Discussions and Conclusion
兺 兺 B̃ 共 x 兲 A ⫽0
L⫽1 k⫽1
I
k
L
k
L for all interior nodes
In this paper, we have explored the feasibility of using a me-
soscale machine to extrude as a viable manufacturing process to
兵 I:supp共 ⌿ I 兲 艚⌫⫽0 其 (15) make micropins of the order of 1 mm. Extrusion is a fast process
NP NL N Ph that is ideally suited for mass production; it is a well-established
兺 兺 B̃ 共 x 兲 A ⫽ 兺 N 共 x 兲 w
L⫽1 k⫽1
I
k
L
k
L
L⫽1
I L L for boundary nodes process at the conventional macroscale. It is also a bulk-
deformation process that produces less waste as compared to other
manufacturing processes, such as machining, which is subtractive
兵 I:supp共 ⌿ I 兲 艚⌫⫽0 其 (16) in nature. Recent work published in the literature has shown that
where N L is the number of subdivisions in the representative due to size effects, the knowledge of conventional extrusion can-
domain. not be extended to the microextrusion process. This research en-
deavor aims to 共a兲 characterize this size effect and the consequent
4.2 Simulation Results Using RKEM. In this section, pre- impacts it has upon the parameters of extrusion at microscale; and
liminary results for numerical simulation of the microextrusion 共b兲 develop numerical methods to simulate the material disconti-
process are discussed. The RKEM was used to simulate the ex- nuity across the grains. The tensile tests conducted on samples of
trusion of a micropin for the 211 ␮m grain-size sample. As a differing grain size and sample dimensions distinctly show a
comparison, an FEM simulation using a commercial Finite Ele- variation in the flow stress curve due to inherent microstructures
ment Method code ABAQUS/Explicit was used. In both simula- of the samples. However, no apparent effect of sample dimension
tions about 15,000 nodes were used to generate the desired mesh was observed. The effect of surface grains could not be seen for
using eight-node brick elements 共C3D8R in ABAQUS/Explicit兲. A this particular material 共CuZn30 brass兲 at the tested sample di-
coarser mesh can be used if adaptive meshing is invoked in the mensions and grain sizes.
simulation. However, for the purpose of this comparison, no adap- Extrusion tests performed on the samples of different grain size
tive mesh was used in FEM or RKEM. In both simulations, the showed that the process is dependent on the surface texture of the
parameters for the material models are selected based on the re- die. However, the die surface tends to smooth out in subsequent
sults of the tensile tests presented in Section 2 for the 211 ␮m testing and then settles down to a value, which is approximately
grain-size material. The material model characteristics are given constant. The phenomenon observed for the copper alloy exam-
by a Young’s modulus E⫽5429 MPa, a Poisson’s ratio of v ined here is not necessarily true for other material. Force-
⫽0.34, a von-Mises yield surface, and isotropic hardening. The displacement curves for the extrusion process were seen to be
plastic behavior of the material is modeled using the power law fairly repeatable with the same trend observed in all samples
␴ (␧
¯ )⫽K(␧ ¯ ) n , where the strength coefficient K⫽379 MPa and the tested. The trend also matches well with the results seen at con-
strain hardening exponent n⫽0.49 共the corresponding values for ventional extrusion dimensions. The dimensions of the pins were
the 87 ␮m grain-size material are K⫽385 MPa and n⫽0.45 and observed to be slightly larger than those of the dies, probably due
for the 32 ␮m grain size material, the values are K⫽393 MPa and to compliance of the dies and the mounting or to separation of the
n⫽0.40). Both models use a Coulomb friction model with a co- segmented dies during extrusion. The standard deviation of all
efficient of friction ␮⫽0.1. In addition, the normalized support dimensions was seen to be around 0.01 mm. This would be im-
size for the RKEM method is taken as 0.8. Figure 25 shows a plot portant to establish the process capability of the extrusion setup
for the equivalent plastic strain obtained from the FEM simulation and subsequently determine the tolerances of the pins made in this
and RKEM. The material was treated as a continuum, and the manner. Bulk deformation processes usually do not produce high

650 Õ Vol. 126, NOVEMBER 2004 Transactions of the ASME


Fig. 27 The influence domains of nodes near the material
interface

tion 4兲, this new method is ideal for the numerical simulation of
the extrusion process as the need for remeshing the model is far
less stringent than that of conventional FEMs.
In future work, the possibility of extruding pins smaller than 1
mm is being explored. Preliminary results indicate that surface
interactions play a big part in the extrusion of pins of smaller
dimensions. The extrusion force is significantly higher than ex-
Fig. 25 Comparison of equivalent plastic strain using „a… pected. Using the analytical model in 关9兴, the force required to
ABAQUS Explicit and „b… the RKEM extrude a billet of diameter 0.8 mm to form a pin of diameter 0.53
mm was estimated to be 1.37 kN. However, tests done using the
extrusion method discussed in this paper have revealed that the
force required to form a pin of such small size is in excess of 2
tolerances as in other processes, such as machining, and this kN. This is pretty close to the buckling load of the steel ram,
would restrict the applications for which such technology could be which was calculated as 2.2 kN. We think that the higher force
used. seen in the microforming is due to the higher friction force com-
In order to perform numerical investigations of the microform- pared to the conventional extrusion process. The RKEM approach
ing process, the use of the Reproducing Kernel Element Method is also being modified to account for individual grains and their
共RKEM兲 was recommended. Results using the RKEM indicate behavior during extrusion. To consider the effect of microstructure
that the method compares well to conventional FEM approaches in microforming, the principle of virtual work should include the
in its ability to capture the deformation behavior of microextru- grain boundary energy, energy of driving forces due to the curva-
sion. The force-displacement response predicted by the RKEM ture of the interface, elastic strain energy and the work of external
also compares well with experiments and the conventional FEM. forces 关18兴. This requires the numerical method to effectively cap-
The key technical contribution of this paper lies in the novel ture the local behavior of grain boundary migration. It is a very
experiment setup used and the special properties of the RKEM.
By using a forming assembly in conjunction with a loading stage,
the authors were able to investigate the force-displacement re-
sponse for microextrusion with material of different grain sizes.
Although many researchers have conducted research in extrusion,
to the best of our knowledge, very few extrusion experiments
have been performed at this scale of the order of 1 mm. In order
to simulate the extrusion process, tedious remeshing is required
due to the large deformations and motion of the interfaces. Be-
cause of the inherent advantages of the RKEM 共discussed in Sec-

Fig. 28 „a… Discretization of material interface in 2-D „b… Shape


function of node I „c… The x-derivative of shape function of
Fig. 26 Comparison of the force-displacement response for node I on the side of material B „d… The x-derivative of shape
ABAQUS Explicit, the RKEM, and experimental results function of node I on the side of material A

Journal of Manufacturing Science and Engineering NOVEMBER 2004, Vol. 126 Õ 651
complex problem due to the fact that the interface moves, and, 关5兴 Engel, U., Egerer, E., and Geiger, M., 2003, ‘‘Production of Microparts by
Cold and Warm Forging,’’ Proc. of 1st International Conference on Micro and
therefore, tedious remeshing procedures are needed in the conven- Nano Technology, Copenhagen, Denmark, Nov., pp. 69–72.
tional finite element method. The RKEM, can effectively handle 关6兴 Balendra, R., and Qin, Y., 2004, ‘‘Research Dedicated to the Development of
this problem. The discontinuity along the interface can be recov- Advanced Metal Forming Technologies,’’ J. Mater. Process. Technol., 145, pp.
ered by truncating the influence domain of the node. Thus the 144 –152.
influence domain of the node only covers the domain that has the 关7兴 Saotome, Y., and Iwazaki, H., 2001, ‘‘Superplastic Backward Microextrusion
of Microparts for Micro-Electro-Mechanical Systems,’’ J. Mater. Process.
same material as at the node point. The cut global partition poly- Technol., 119, pp. 307–311.
nomial and kernel function can automatically create a discontinu- 关8兴 Hosford, W. F., and Caddell, R. M., 1983, Metal Forming: Mechanics and
ity in the derivative of the shape function along the material in- Metallurgy, 2nd Edition, Prentice-Hall, Engelwood Cliffs, NJ, pp. 181–187.
terface 关28兴. As shown in Fig. 27, the material interface is defined 关9兴 Groover, M. P., 1996, Fundamentals of Modern Manufacturing: Materials,
Processes and Systems, Prentice Hall, Englewood Cliffs, NJ, pp. 479– 482.
through the common nodes belonging to the adjacent materials. 关10兴 Bhattacharyya, D., Richards, P. J., and Somashekar, A. A., 1992, ‘‘Modeling of
The shape function and its derivatives in two dimensions are plot- Metal-Extrusion Using the Phoenics Package,’’ J. Mater. Process. Technol., 35,
ted in Fig. 28. The displacement continuity for the common nodes pp. 93–111.
on the material interface is satisfied due to the interpolation prop- 关11兴 Huh, H., and Lee, C. H., 1993, ‘‘Eulerian Finite-Element Modeling of the
Extrusion Process for Work-Hardening Materials With the Extended Concept
erty of reproducing kernel element shape functions. Moreover, it of Limit Analysis,’’ J. Mater. Process. Technol., 38, pp. 51– 61.
is easy to locally remesh the model in the RKEM because an 关12兴 Reinikainen, T., Welo, T., Korhonen, A. S., and Kivivuori, S., 1994, ‘‘Com-
unstructured mesh can be used due to the natural conforming of parison of Two Commercial FEM Codes in Cold-Extrusion Simulation,’’ J.
the approximation. The RKEM provides a potential tool to simu- Mater. Process. Technol., 42, pp. 137–146.
关13兴 Alexandrov, S., Mishuris, G., Miszuris, W., and Sliwa, R. E., 2001, ‘‘On the
late the microforming process while considering the effect of Dead Zone Formation and Limit Analysis in Axially Symmetric Extrusion,’’
microstructure. Int. J. Mech. Sci., 43, pp. 367–379.
关14兴 Arentoft, M., Gronostajski, Z., Niechajowicz, A., and Wanheim, T., 2000,
‘‘Physical and Mathematical Modeling of Extrusion Processes,’’ J. Mater. Pro-
Acknowledgment cess. Technol., 106, pp. 2–7.
关15兴 Sheppard, T., 1999, ‘‘Temperature Changes Occurring During Extrusion of
The support from the NSF Grant No. DMI-0115079 is greatly Metals: Comparison of Bulk, Numerical, and Integral Profile Predictions With
appreciated. Experimental Data,’’ Mater. Sci. Technol., 15, pp. 459– 463.
关16兴 Chanda, T., Zhou, J., and Duszczyk, J., 2001, ‘‘A Comparative Study on Iso-
Nomenclature Speed Extrusion and Isothermal Extrusion of 6061 Al Alloy Using 3D FEM
Simulation,’’ J. Mater. Process. Technol., 114, pp. 145–153.
A L ⫽ volume of the representative domain of node L 关17兴 Li, L., Zhou, J., and Duszczyk, J., 2004, ‘‘Prediction of Temperature Evolution
G L ⫽ group of nodes whose shape function influences node During the Extrusion of 7075 Aluminum Alloy at Various Ram Speeds by
Means of 3D FEM Simulation,’’ J. Mater. Process. Technol., 145, pp. 360–
L 370.
兩 v ⫽ reproducing kernel element interpolation operator on 关18兴 Chen, J. S., and Lu, H., 2002, ‘‘A Double Grid Method for Modeling of
a continuous function Microstructure Evolution,’’ Proc. of Fifth World Congress on Computational
K ␳ ⫽ kernal function with support size ␳ Mechanics, Vienna, Austria, July.
关19兴 Liu, W. K., Han, W., Lu, H., Li, S., and Cao, J., 2004, ‘‘Reproducing Kernel
N L ⫽ number of subdivisions in a representative domain of Element Method: Part I. Theoretical Formulation,’’ Comput. Methods Appl.
node L Mech. Eng. 共in press兲.
N P ⫽ number of nodes in the domain 关20兴 Li, S., Lu, H., Han, W., Liu, W. K., and Simkins, D. C., 2004, ‘‘Reproducing
w L ⫽ integration weight Kernel Element, Part II. Global Conforming Im/Cn Hierarchy,’’ Comput.
Methods Appl. Mech. Eng. 共in press兲.
␧ ihj ⫽ strain obtained from displacement approximation 关21兴 Lu, H., Li, S., Simkins, D. C., Liu, W. K., and Cao, J., 2004, ‘‘Reproducing
˜␧ ihj ⫽ strain obtained after smoothing Kernel Element Method Part III. Generalized Enrichment and Applications,’’
Comput. Methods Appl. Mech. Eng. 共in press兲.
⌽ ⫽ distribution function 关22兴 Chen, J. S., Wu, C. T., Yoon, S., and You, Y., 2001, ‘‘A Stabilized Conforming
␳ ⫽ support size Nodal Integration for Galerkin Meshfree Methods,’’ Int. J. Numer. Methods
␺ n,i ⫽ global partitioning polynomial Eng., 50, pp. 435– 466.
关23兴 Park, J. S., and Hwang, S. M., 1991, ‘‘Automatic Remeshing in Finite Element
⌿ I ⫽ reproducing kernel element shape function of node I Simulation of Metal Forming Processes by Guide Grid Method,’’ J. Mater.
Process. Technol., 27, pp. 73– 89.
关24兴 Belytschko, T., Lu, Y. Y., and Gu, L., 1994, ‘‘Element-Free Galerkin Meth-
References ods,’’ Int. J. Numer. Methods Eng., 37, pp. 229–256.
关1兴 Geiger, M., Kleiner, M., Eckstein, R., Tiesler, N., and Engel, U., 2001, ‘‘Mi- 关25兴 Liu, W. K., Jun, S., and Zhang, Y. F., 1995, ‘‘Reproducing Kernel Particle
croforming,’’ CIRP Ann., 50共2兲, pp. 445– 462. Methods,’’ Int. J. Numer. Methods Fluids, 20, pp. 1081–1106.
关2兴 Tiesler, N., and Engel, U., 2000, ‘‘Microforming—Effects of Minaturization,’’ 关26兴 Liu, W. K., and Jun, S., 1998, ‘‘Multiple-Scale Reproducing Kernel Particle
Metal Forming 2000, Balkema, Rotterdam, pp. 355–360. Methods for Large Deformation Problems,’’ Int. J. Numer. Methods Eng.,
关3兴 Raulea, L. V., Govaert, L. E., and Baaijens, F. P. T., 1999, ‘‘Grain and Speci- 41共7兲, pp. 1339–1362.
men Size Effects in Processing Metal Sheets,’’ Advanced Technology of Plas- 关27兴 Chen, J. S., Pan, C., Wu, C. T., and Liu, W. K., 1996, ‘‘Reproducing Kernel
ticity, Proc. of 6th ICTP, Sept. 19–24, 2, pp. 939–944. Particle Methods for Large Deformation Analysis of Nonlinear Structures,’’
关4兴 Geiger, M., Messner, A., Engel, U., Kals, R., and Vollersten, F., 1995, ‘‘Design Comput. Methods Appl. Mech. Eng., 139, pp. 195–227.
of Microforming Processes—Fundamentals, Material Data and Friction Be- 关28兴 Lu, H., Liu, W. K., Chen, J. S., and Cao, J., 2003, ‘‘Consistent Smoothing
havior,’’ Proc. of 9th International Cold Forging Congress, Solihull, UK, May, Technique and Treatment of Material Discontinuity in the Reproducing Kernel
pp. 155–163. Element Method,’’ submitted to the Int. J. Numer. Methods Eng.

652 Õ Vol. 126, NOVEMBER 2004 Transactions of the ASME

S-ar putea să vă placă și