The International Journal of Advanced Manufacturing Technology (2020) 111:371–382

https://doi.org/10.1007/s00170-020-06117-9

ORIGINAL ARTICLE

Thermo-fluid simulation using particle method based

on hand-pouring motion in casting process
Hitoshi Tokunaga 1 & Yuichi Motoyama 1 & Kazuyo Iwamoto 1 & Toshimitsu Okane 1

Received: 6 February 2020 / Accepted: 20 September 2020 / Published online: 28 September 2020
# Springer-Verlag London Ltd., part of Springer Nature 2020

Abstract
Casting simulators are widely used to evaluate casting designs. However, most simulators using conventional methods such as
FDM or FEM entail difficulty in evaluating casting designs sufficiently based on ladle motions or an operator’s pouring motion
because they do not readily accommodate dynamically changing wall boundaries and do not easily accommodate the ladle
motion as input. For this study, the authors extended the thermo-fluid simulation based on particle method, an earlier proposed
method, so that the measured ladle motion can be input. The authors measured an operator’s ladle motion using marker tracking
technique for AR and executed thermo-fluid simulations using the proposed simulation method. Comparisons of the experiments
and the simulation results clarified that the proposed method can accurately predict flow behaviors of the molten metal resulting
from the ladle motion. Comparisons of the proposed simulation and the simulation with a conventional simulation using the
inflow condition of the constant flow rate clarified the importance of executing the simulation based on the ladle motion for high-
accuracy prediction of molten metal behavior.

Keywords Aluminum alloy casting . Augmented reality (AR) . Pouring motion . Smoothed particle hydrodynamics (SPH) .
Thermo-fluid simulation

1 Introduction accommodate the ladle motion, i.e., the operator’s pouring

Many casting processes used by human operators include
hand pouring of molten metal with ladles. Differences of mo-
tions of various operators and differences of hand motions
used each time engender differences of molten metal flow
behavior, and consequently influence the quality of castings.
To produce higher-quality castings, evaluating operator
pouring motion is important.
In recent years, many studies have simulated molten met-
al flow, heat transfer, solidification, and so on [1], based on
which casting simulators have been developed. They are
used widely to evaluate casting designs. Nevertheless, most
use finite element method (FEM) or finite difference meth-
od (FDM). Such conventional simulators do not readily
* Hitoshi Tokunaga
tokunaga.h@aist.go.jp
1
Advanced Manufacturing Research Institute, National Institute of
Advanced Industrial Science and Technology (AIST), 1-2-1 Namiki,
Tsukuba, Ibaraki 305-8564, Japan
motion, as input. Therefore, most casting designers usually
derive inflow conditions of constant flow rate and constant
cross-sectional area from the amount of molten metal, the
pouring time, or the pressure head and execute casting sim-
ulations under these conditions. Such simulators have diffi-
culty predicting casting defects resulting from the opera-
tor’s pouring motion. Therefore, the simulators are difficult
to use as pouring motion evaluation tools to assess casting
quality.
Particle methods are simulation methods able to use
time-varying wall boundaries more easily than other con-
ventional methods [2, 3]. In particle methods, a fluid is
represented by a finite number of calculation elements
(particles). The overall fluid behavior is predicted by cal-
culating interaction among the particles. The smoothed
particle hydrodynamics (SPH) method [2] and moving par-
ticle semi-implicit (MPS) method [3] are well known as
particle methods that are applicable to casting process anal-
yses [4–10] because they can express various alloy states
by assigning each particle its own physical property
values, and because they can readily accommodate moving
wall boundaries such as the moving ladle shape. Cleary

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

372
formulated interparticle heat conduction with different
thermal conductivity and formulated the phase transforma-
tion of particles based on enthalpy [4]. Based on that for-
mulation, Cleary proposed a simulation method to predict
molten metal flow and solidification using SPH method. In
the method, a solid was modeled as a highly viscous fluid.
Kazama et al. proposed a method of stably simulating flow
and solidification based on Cleary’s method by introducing
a Riemann solver [5]. They then applied the method to the
simulation of a semi-solid casting. Ichimiya et al. proposed
a method predicting flow, solidification, and shrinkage
using the SPH method [6]. Their method introduced the
simplified marker and cell (SMAC) method, which can
calculate the uncompressed condition with high accuracy.
Their method calculated the solidified layer behavior using
discrete element method (DEM) [11], which was expanded
to address shrinkage occurring along with temperature
change. Hirata et al. proposed a heat transfer and solidifi-
cation analysis method [7] using the MPS method [3] and
proposed an integrated simulation method of flow and so-
lidification shrinkage formation [8]. Tokunaga et al. pro-
posed a method for accelerating SPH thermo-fluid simula-
tion using graphics processing units (GPUs) and proposed
a method for designing runners interactively during casting
flow simulation [9]. Tokunaga et al. also developed a
method to simulate the formation and flow of cold flakes
in high-pressure die casting by coupling flow analysis with
mechanical analysis [10].
These particle simulation methods are expected to be
able to accommodate the operator’s ladle motion as a
time-varying wall boundary. However, few such methods
exist. Watanuki et al. developed a hand-pouring training
system for casting processes [12]. Their system detected an
operator’s ladle motion using marker tracking for augment-
ed reality (AR), and evaluated the detected ladle motion by
the molten metal flow behavior that was predicted using a
particle method. Although their system is useful for hand-
pouring training, the system for evaluating pouring opera-
tions during the casting process was not evaluated suffi-
ciently as a simulator in the following terms. They did not
discuss details of the pouring motion influence on the flow
behavior of the molten metal. They did not consider heat
transfer. Moreover, they used only water, not molten met-
al, for experimentation and simulation. They examined on-
ly the rotational component of the ladle motion.
This paper describes extension of an earlier proposed
method [9, 10] to a simulation method based on casting
operation by the operator. Specifically, the method uses
the ladle motion as input consisting of translation and rota-
tion. Then we demonstrate the usefulness of this method.
The authors developed a method using marker tracking
technique for AR to measure the ladle motion, thereby
reflecting the operator’s pouring motion [13, 14]. First, a
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int J Adv Manuf Technol (2020) 111:371–382
pouring experiment is executed using a mold with quartz
glass installed so that the molten metal behavior is visible.
The experiment using the measurement technique described
above also allows measurement of the ladle motion.
Second, this paper describes a thermo-fluid simulation ex-
ecuted using the measured ladle motion as input. Finally, to
demonstrate the usefulness of the proposed method for the
molten metal flow behavior, this paper describes a compar-
ison of pouring experiment results and results of simulation
of the following points: (1) contour shape of the molten
metal and (2) the molten metal tip arrival position in the
mold at each time. To demonstrate the importance of the
simulation based on the operator’s ladle motion, this paper
also presents comparison of results obtained using the pro-
posed method and results of the simulation under the inflow
condition of the constant flow rate.
2 Simulation technique, experimental
condition, and simulation condition
2.1 SPH method for thermo-fluid simulation
The SPH method, a fully Lagrangian, mesh-free, and particle-
based technique, is used for this study. Using this method, the
governing equations are discretized using a finite set of dis-
crete values defined at interpolation points. Each point,
representing a particle, is assigned physical properties such
as density, velocity, or enthalpy. By updating the property
values of every particle using discretized equations, one can
calculate the overall behavior of the fluid.
To model the casting flow, the following governing equa-
tions are used.
Dρ
¼ −ρ∇  v ð1Þ
Dt
Dv 1 1
¼ − ∇p þ ∇  τ þ f S þ f ð2Þ
Dt ρ ρ
DH DT 1
¼ Cp ¼ ∇  ðλ∇T Þ ð3Þ
Dt Dt ρ
In those equations, ρ signifies density, v denotes velocity, p
represents pressure, τ is the deviatoric stress tensor, fS stands
for the force attributable to the interfacial tension such as
surface tension, f stands for external force, H represents en-
thalpy, Cp represents the specific heat, T denotes temperature,
and λ stands for thermal conductivity. Equations (1), (2), and
(3) respectively represent the continuity equation, the momen-
tum equation, and the energy equation. In Eq. (3), latent heat
during solidification is incorporated into the enthalpy.
To calculate these equations, this study introduces their
discretization forms proposed by Cleary [4]: this study uses
a combination of the discretization form of the continuity
Int J Adv Manuf Technol (2020) 111:371–382 373

equation (Eq. (4)), the momentum equation (Eq. (5)), and the variation is less than 1% and that the flow is nearly incom-
energy equation (Eq. (6)), as shown below. pressible [4]. The mass of each particle is calculated by mul-
tiplying the density with the volume calculated from the par-
dρi

¼ ∑ m j vi −v j  ∇i W ij ð4Þ ticle size set at the start of the simulation. It is maintained until
dt j the end of the simulation.
" !
 #
dvi pi p j ξ 4μi μ j vi −v j  rij
¼ − ∑ mj þ 2 − ∇i W ij 2.2 Input of ladle motion to particle method
dt j ρ2i ρj ρi ρ j μi þ μ j r2ij þ η2
simulation
þ f Si þ f i
The SPH simulation, which employs the ladle motion as input
ð5Þ
measured during casting operations, is described next.
dH i 2m j 2λi λ j
 rij First, the method proposed herein measures the ladle mo-
¼∑ T i −T j 2  ∇i W ij ð6Þ
dt ρ ρ λ
j i j i þ λ j rij þ η2 tion (time series of position and orientation), which highlights
the operator’s pouring operation. The method employs a ladle
Therein, ρi, mi, vi, pi, μi, λi, Hi, and Ti respectively denote motion measurement technique using marker tracking for AR
the density, mass, velocity, pressure, dynamic viscosity, ther- [13, 14]. As presented in Fig. 1, with AR markers placed on
mal conductivity, enthalpy, and temperature of the ith particle. the mold and ladle, the pouring operation, which includes the
rij represents the relative position of the ith particle with re- markers, is recorded using a video camera. The relative posi-
spect to the jth particle. rij denotes |rij|. fSi stands for the force tion and orientation of the ladle with respect to the mold is
attributable to the interfacial tension such as surface tension. fi calculated using AR marker tracking.
stands for external forces including the gravity force. ξ is the As presented in Fig. 2a, from time series data of the ladle
viscous scaling factor. In addition, η is a small parameter used motion, the method calculates the position and orientation x(t)
to smooth out the singularity. Wij is the kernel function cen- of the ladle at time t. The measured ladle motion is discrete
tered on the ith particle position. This study uses the quintic data that includes a time component. Therefore, the position
function of Wendland [15] as the kernel function to correct and orientation x(t) are obtained by interpolating them.
tensile instability. For this study, surface tension fSi is calcu- Velocity and angular velocity can also be calculated from
lated using the surface tension model for the MPS method the position and orientation.
based on interparticle potential developed by Kondo et al. [16] Next, as presented in Fig. 2b, the method moves the shape,
Usually, heat transfer between mold and molten metal is which is a wall boundary, of the ladle according to x(t).
calculated using the heat transfer coefficient. For the present Furthermore, as presented in Fig. 2c, under the updated wall
study, heat transfer between the mold and molten metal is boundaries, the method calculates the molten metal behavior
calculated using Eq. (7). Relevant details are presented in at t using SPH method.
the Appendix 1.
dH i 2m j λi λ j hrv
 rij 2.3 Experiment conditions
¼∑
 T i −T j 2  ∇i W ij
dt j ρi ρ j λi λ j þ αij λi hr ij þ 1−αij λ j hr ij rij þ η2
Figures 3 and 4 present the experimental settings. Figure 3
ð7Þ
shows the mold and ladle used for this pouring experiment.
In that equation, h denotes the heat transfer coefficient, rv The ladle comprises a crucible and a handle. The crucible had
represents the distance measured vertically to the mold surface 98 mm inner diameter, 146 mm depth, and a mold with
from particle j (mold particle) representing the mold to the 200 mm cavity length and 33 mm width. As depicted in
particle i (molten metal particle) representing the molten met- Fig. 3, the mold had quartz glass on its front face to enable
al, and αij (0 ≤ αij ≤ 1) is a parameter expressing the position molten metal flow observations. The crucible and mold had
of the mold surface with respect to the position of the mold AR markers on them to measure the ladle motion.
particle j and the molten metal particle i. Three video cameras placed as portrayed in Fig. 4 recorded
The equation of state, defined as Eq. (8), is used to calculate the pouring operation with the mold and ladle. The diagonally
the pressure from the value of density [2, 4]. forward camera shot a video for tracking AR markers. The
  γ  cameras placed on the front side and upper side recorded the
100ρ0 v2max ρ molten metal flow behavior. The mold and ladle had coordi-
p¼ −1 ð8Þ
γ ρ0 nate systems with coordinates defined as presented in Fig. 5
(details are presented in the next section).
Therein, ρ0 stands for the reference density; vmax represents The experiment used 330 g molten aluminum alloy ISO
the maximum flow velocity. The exponent γ = 7 is used. The Al-Si7Mg (JIS AC4CH). The initial temperature of the molten
combined use of Eqs. (4), (5), and (8) ensures that the density alloy was 720 °C. Its pouring temperature was 706 °C. The

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

374 Int J Adv Manuf Technol (2020) 111:371–382

Fig. 1 Computing ladle motion Video camera Movie Ladle motion Ladle motion
from recorded motion of AR
markers attached to the ladle and Ladle
mold
detection (t0, x0),…,(ti, xi),(ti+1, xi+1),…
0.3
x y z

Displacement, m
AR markers AR marker 0.2

tracking 0.1
0
Sprue -0.1
-0.2
-0.3
0 2 4 6 8
Mold Time, s

mold material was ISO 185/JL/250 (JIS FC250). Its initial
temperature was 120 °C. The mold was coated with boron
nitride as a mold release agent.
Table 1 presents the chemical composition of the molten
metal used for experiments. Pouring experiments were con-
ducted twice. The time interval of the pouring operation was
about 4 s. The time interval from the start to the end of the
molten metal outflow from the crucible was about 1.534 s in
the first experiment. It was about 1.634 s in the second exper-
iment. These time intervals were obtained from videos taken
during the pouring experiments.
2.4 Simulation condition
Figure 5 shows 3D shape models used for casting simulations
in this paper. These shapes are the same as the mold and ladle
shapes used for the pouring experiments. The simulations
used these shape models and the ladle motion calculated from
data measured during pouring experiments. The simulations
in this paper used the interpolation process with the derived
ladle motions by simply linearly interpolating the measured
data. Figure 5 presents the coordinate systems set to the shape
models of the mold and ladle. The coordinate system of the
ladle has its origin at the center of the upper plane of the ladle.
The roll angle, pitch angle, and yaw angle respectively signify
rotation about the x-axis, y-axis, and z-axis. Coordinates x, y,
and z represent the relative position of the origin of the coor-
dinate system of the ladle with respect to the mold. The sim-
ulation used 31,598 fluid particles and 275,542 mold and ladle
particles. The mold particles filled the inside of the mold wall.
The simulation calculated heat conduction in the molten metal
and the mold and calculated heat transfer only between the
mold and the molten metal. The simulation used 1225 W/m2K
as the heat transfer coefficient, as calculated based on results
of preliminary tests using a mold coated with boron nitride.
For comparison, simulations were also done with inflow con-
ditions of constant velocity value vp determined using Eq. (9)
according to the pouring time in the experiment.
Vp
vp ¼ ð9Þ
Ap Δt p
In that equation, Vp represents the poured molten metal
volume, Ap stands for the cross-section area of the molten
metal on the upper plane of the mold, and Δtp denotes the
time interval of the molten metal outflow from the crucible.

Interpolated motion x(t)

Displacement, m

0.3
Ladle motion Interpolation 0.2
x y z

(t0, x0),…,(ti, xi),(ti+1, xi+1),… (a) 0.1

0
0.3
Displacement, m

-0.1
x y z
0.2 -0.2
0.1 Time t Moving -0.3
0 0 2 t 4 6 8

-0.1 Ladle shape shape (b) Time, s

-0.2
-0.3
SPH method
0 2 t 4 6 8
Time, s Ladle shape at t
z xi+1, ti+1 Analysis Simulated
Interpolated (c) behavior of
ladle motion x(t) x
Ladle molten metal
xi, ti
GPU
xi-1, ti-1 y
Fig. 2 Casting simulation from the measured ladle motion

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Int J Adv Manuf Technol (2020) 111:371–382 375

Fig. 3 Mold and ladle with

Quartz glass AR marker AR marker
attached AR markers

Crucible

Mold with attached AR marker Ladle with attached AR marker

Figure 6 presents physical properties of the molten metal
used for simulations. In that figure, TL and TS respectively
represent liquidus and solidus temperatures. The physical
properties of the solid–liquid coexistence area in Fig. 6 were
calculated using material property calculation software
(JMatPro Ver.10.1; Sente Software Ltd.) from the chemical
composition of the molten metal (Al-Si7Mg) used in the
experiment presented in Table 1. The physical properties
of liquid and solid area were set as constant. Mutations
across the solidus line in Fig. 6 are caused by the sudden
crystallization of Al+Si+Mg2Si ternary eutectic. The value
changes through the TS and the 575 °C are caused by the
crystallization of Al+Si binary eutectic. The simulations
described below predicted that the temperature of the mol-
ten metal did not fall below liquidus until completion of
filling. This paper therefore does not discuss Fig. 6 further.
Video camera䠄upper side)
This study applied 0.878 N/m as the surface tension coef-
ficient of the aluminum alloy. This value was the calculation
result obtained from JMatPro.
3 Results and discussion
3.1 Measurement of ladle motion
Under the experimental conditions described above, pouring
experiments were executed twice. Figure 7 presents time se-
ries data (x, y, z, roll, pitch, and yaw) of the ladle motion
measured in these experiments. Figure 8 shows these data
superimposed for comparison and shows that the translational
motions in the first and second experiments were almost iden-
tical, except for the initial position. For the pitch angle and
yaw angle, the motions in the first and second experiments
were almost identical. For the roll angle, the motion of the first
experiment was faster than the motion of the second experi-
ment. Figure 7 shows that noise occurred around 2.5 s in the
first motion data and at around 1.2 s in the second motion data.

Video camera for z

tracking AR markers yaw
pitch
Ladle (crucible)
x roll y
[x y z] z

Mold
Molten metal
y
Video camera (front side)
x
Cavity Mold

Fig. 4 Video cameras placed to record ladle motion Fig. 5 Simulation setting

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

376 Int J Adv Manuf Technol (2020) 111:371–382

Table 1 Chemical composition

of Al-Si7Mg Al (wt%) Cu (wt%) Fe (wt%) Si (wt%) Mn (wt%) Mg (wt%) Ti (wt%)

92.47 0.01 0.11 6.90 0.01 0.33 0.17

These noises show errors of AR marker detection. These 3.2 Comparison of experiment and simulation results
noises of the motion data were smoothed and applied to the
simulations. To demonstrate the usefulness of the proposed method, this
paper presents a comparison of the simulation results and the
experiment results. As described above, the time intervals of
12
the molten metal outflow were about 1.534 s in the first ex-
11 periment and 1.634 s in the second experiment. For simula-
10 5 J/kg

10 tions conducted under the inflow condition of a constant flow

9 rate, it is necessary to give the area of the cross-section and
8 fluid velocity at the inflow position. The 15-mm average di-
TL: Liquidus
Enthalpy,

7
TS : Solidus
ameter of the molten metal was derived from the recorded
6 video. The area was calculated from the diameter. The fluid
5
velocities were calculated using Eq. (9) from the area, the time
4
500 550 TS 600 TL 650 intervals, and the amount of the molten metal (about 1.36 ×
Temperature, ºC 10−4 m3). The fluid velocity of the first experiment was about
0.501 m/s. That of the second experiment was about
2600
2580 0.470 m/s.
2560 As described in Sect. 2.3, because the experiments were
Density, kg/m3

2540
2520
planned so that solidification of molten metal was avoided
2500 to the greatest extent possible and because the simulations
2480 shown below predicted that the temperature of the molten
2460
TL: Liquidus
2440 TS: Solidus metal did not fall below liquidus until completion of filling,
2420 we assume for these analyses that the alloy was in the liquid
2400
TS TL state until the completion of filling. Therefore, only the flow
500 550 600 650
Temperature, ºC behavior of the molten metal is discussed below.
25
(1) Comparison of contour shapes of molten metal
Dynamic viscosity, P a s

20

15 Figure 9 shows a front view of the molten metal flow in the

first pouring experiment and that of the corresponding simu-
10
lation result. Here, particle colors in the simulation represent
TL: Liquidus velocity values.
5
TS: Solidus
The upper panel of Fig. 9 presents the pouring experiment
0
500 550 TS 600 TL 650 result. The molten metal surface gradually fell from the inflow
Temperature, ºC position on the left side toward the center and rose again
180
toward the mold wall on the right side. The middle panel of
Thermal co nductivity, W/(m K)

170 Fig. 9 presents results of the simulation based on the pouring

160
150
operation. This result demonstrates that the proposed method
140 can well predict the molten metal contour shape obtained in
130
120 the pouring experiment. The lower panel of Fig. 9 presents
110
TL: Liquidus simulation results obtained under the inflow condition of con-
100
90 TS: Solidus stant flow rate. This simulation result did not correspond to the
80 inflow behavior and the shape of the molten metal surface
70
500 550
TS 600
TL 650 obtained in the pouring experiment.
Temperature, ºC Figure 10 portrays a front view of the molten metal flow in
Fig. 6 Physical properties of Al-Si7Mg calculated using simulation the second pouring experiment and that of the corresponding
software simulation result. The upper panel of Fig. 10 shows that the

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Int J Adv Manuf Technol (2020) 111:371–382 377

Fig. 7 Ladle motions measured in

experiments
0.2
z 0.2
z

Displac e me nt, m

Displac e me nt, m
0.1 0.1
0 x 0 x
-0.1 -0.1
-0.2 y -0.2
y
-0.3 -0.3
0 1 2 3 4 5 0 1 2 3 4 5
Time, s Time, s

25 25
pitch pitch
0 yaw 0
yaw

Rotation angle
Rotation angle
-25 -25
-50 -50
roll roll
-75 -75
-100 -100
0 1 2 3 4 5 0 1 2 3 4 5
Time, s Time, s

N1 N2

molten metal surface of the second experiment was less dis-
turbed than that of the first experiment. The middle panel of
Fig. 10 shows that the proposed method can well predict the
molten metal contour shape in the second experiment and
shows the difference of the contour shape from that of the first
experiment. The lower panel of Fig. 10 shows that the simu-
lation under the inflow condition of constant flow rate can also
predict the molten metal contour shape. However, comparison
of the lower panel of Fig. 9 with the lower panel of Fig. 10
reveals that the simulation under the inflow condition of con-
stant flow rate was not able to predict the tendency of the
difference between the first and second molten metal contour
shapes.
In this example, as shown in the middle and lower panels of
Fig. 10, no significant difference was found between the sim-
ulation results under the inflow condition of constant flow rate
and the results obtained using the proposed method.
Figure 11 portrays the upper view of the molten metal flow
in the first pouring experiment and that of the corresponding
simulation result. As presented in the upper panel of Fig. 11,
in the experiment, because of the ladle motion by the operator,
the molten metal flowed into a position that deviated from the
center line of the mold cavity. Results show that the contour
shape of the molten metal was not symmetrical to the center
line. The middle panel of Fig. 11 presents the result of the
simulation based on the pouring operation. This result dem-
onstrates that the proposed method can well predict the inflow
position and the contour shape of the molten metal obtained in
the pouring experiment. By contrast, as shown in the lower
panel of Fig. 11, the simulation under the inflow condition of
constant flow rate was unable to predict such behavior of the
molten metal.
In the second pouring experiment, the molten metal flowed
almost onto the central line of the mold cavity, such that the

Fig. 8 Comparison of ladle 0.2 25

motions pitch
z
Displac e me nt, m

0.1 0
yaw
Rotation angle

0 x -25
-0.1 -50 roll
-0.2 y -75
-0.3 -100
0 1 2 3 4 5 0 1 2 3 4 5
Time, s Time, s

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

378 Int J Adv Manuf Technol (2020) 111:371–382

Fig. 9 Experiment and simulation

results (front side, N1)

Experiment

Velocity, m/s
1.0
Simulation
0.5
(hand-pouring)
0.0
Inflow condition Velocity, m/s
1.0
Simulation
0.5
(constant rate)
0.0

molten metal did not collide with the wall surface and the
contour shape of the molten metal was symmetrical to the
center line. The proposed simulation method predicted an al-
most identical contour shape of the molten metal. No result
was found to be significantly different from the simulation
results under the inflow condition of the constant flow rate.
Figure 12 presents cross-sections of the results of the mol-
ten metal flow prediction obtained at a certain time of the first
pouring experiment for the simulation using the proposed
method and for the simulation using the inflow condition of
the constant flow rate. The simulation result of the proposed
method demonstrates that the molten metal flowed into the
mold in an oblique direction, unlike the simulation of constant
flow rate. The difference in the flow velocity in the lateral
direction might affect the difference in the contour shape of
the molten metal shown in Figs. 9 and 10, i.e., the difference
of the molten metal behavior is based on the ladle motion. For
that reason, this paper describes results of comparison of the
molten metal tip positions as shown below.
(2) Comparison of molten metal tip positions
A comparison can be made of the positions of the molten
metal tip obtained from the front view image of the molten
metal flow in the pouring experiment and the front view image
of the corresponding simulation result. Here, the molten metal
tip position was normalized with the left wall position in the
mold cavity as 0 and the right wall position as 1, as presented
in Fig. 13. The time was normalized with the time when the
molten metal started flowing into the cavity as 0 and the time
when it reached the right wall in the cavity as 1.
Figure 14 shows the molten metal tip positions from the
moment at which the molten metal enters the cavity to when it
reaches the right wall surface in the first pouring experiment,

Fig. 10 Experiment and

simulation results (front side, N2)

Experiment

Velocity, m/s
1.0

Simulation
0.5
(hand-pouring)
0.0

Inflow condition Velocity, m/s

1.0
Simulation
0.5
(constant rate)
0.0

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Int J Adv Manuf Technol (2020) 111:371–382 379

Fig. 11 Experiment and

simulation results (upper side,
N1)
Experiment

Velocity, m/s
1.0

Simulation
0.5
(hand-pouring)
0.0

Inflow condition
Velocity, m/s
1.0
Simulation
0.5
(constant rate)
0.0

in the corresponding simulation based on the measured ladle
motion, and in the simulation under the inflow condition of
the constant flow rate.
The result of the first pouring experiment demonstrates that
the molten metal started to flow vigorously, but that flow
slowed and speeded up again. This molten metal behavior
reflects the characteristics or habits that produce the hand mo-
tions of the operator who poured the molten metal.
This result demonstrates that the simulation based on the
ladle motion can predict the tendency of the molten metal
behavior in the pouring experiment described above.
However, simulation under the inflow condition of the con-
stant flow rate predicted that the molten metal tip position
increased along with the elapsed time. This simulation was
unable to predict the tendency of the molten metal behavior
described above.
Figure 15 shows the molten metal tip positions in the sec-
ond pouring experiment, in the corresponding simulation
based on the measured ladle motion, and in the simulation
under inflow condition of constant flow rate. Similarly to the
simulation for the first experiment, the simulation based on
ladle motion predicted the tendency of the molten metal
behavior in the second pouring experiment. However, the
simulation under an inflow condition of constant flow rate
was unable to predict the tendency of the molten metal behav-
ior because of the operator’s ladle motion.
As presented in Fig. 10, in the second pouring experiment,
no significant difference was found between the simulation
results under the inflow condition of constant flow rate and
the results obtained using the proposed method. However,
from the viewpoint of the time variation of the arrival position
of the molten metal tip, the proposed method predicted the
molten metal behavior more accurately than the simulation
under the inflow condition of constant flow rate did.
4 Discussion
The comparisons described in the preceding section clarified
that the simulation method proposed herein can predict the
molten metal behavior resulting from the operator’s ladle mo-
tion better than the conventional simulation under the inflow
condition of a constant flow rate. These results confirmed that
the inflow condition of the constant flow rate is insufficient to

Fig. 12 Comparison between Molten metal Molten metal Inflow condition

simulation based on measured
ladle motion and simulation with Cavity Cavity Velocity, m/s
constant rate inflow condition 1.0

0.5
Mold Mold
0.0
Simulation (hand-pouring) Simulation (constant rate)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

380
Position of molten metal tip
0 1
Fig. 13 Normalized position of molten metal tip
predict the molten metal behavior accurately. Moreover, the
results confirmed the importance of handling the inflow con-
dition of molten metal considering the pouring operation and
ladle motion.
Some differences remain between the behavior of the mol-
ten metal in the experiment and the simulation results based
on the ladle motion. For example, as shown in Fig. 9, the
highest point on the molten metal surface in the first experi-
ment was to the left of the inflow position. However, in the
simulation, the highest point on the surface did not exist at the
same position. These differences of the experiment and the
simulation result were probably attributable to some factors
that were not examined. For example, in the experiments,
oxide films were formed on the molten metal. Furthermore,
the marker tracking method of AR entails measurement errors.
However, the simulations did not incorporate them. For more
accurate simulation, they should be examined further.
As described herein, pouring experiments and simulations
were conducted using a simple mold. However, the actual
mold shape is expected to be more complex. The experiments
demonstrate that the ladle motions strongly influenced the
molten metal flow behavior, even for this simple mold. The
ladle motion is expected to exert stronger effects on actual
product castings.
Improvement of the prediction accuracy of molten metal
flow behavior can be expected to enhance the prediction
1
0.9
0.8
Normalized position
0.7
0.6
0.5
0.4
0.3 Experiment N1
0.2 Simulation (hand-pouring) N1
0.1 Simulation (constant rate) N1
0
0 0.2 0.4 0.6 0.8
Normalized time
Fig. 14 Comparison of molten metal tip position (N1)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int J Adv Manuf Technol (2020) 111:371–382
1
0.9
0.8
Normalized position
0.7
0.6
0.5
0.4
Experiment N2
0.3
0.2 Simulation (hand-pouring) N2
0.1 Simulation (constant rate) N2
0
0 0.2 0.4 0.6 0.8 1
Normalized time
Fig. 15 Comparison of molten metal tip position (N2)
accuracy of the temperature distribution of molten metal and
the prediction accuracy of casting defects. Consideration must
be given to the effects of molten metal flow behavior on the
resulted casting quality. As described herein, as the first step,
the effects of ladle motion on the flow behavior were com-
pared between experiments and simulations. The flow behav-
ior is expected to influence the temperature distribution of the
filled molten metal, solidification, and consequently the cast-
ing quality. The influence of the ladle motion on the casting
quality is left as a topic for future discussion.
Simulation results of the molten metal tip position captured
the characteristics or habits of the operator’s hand motions
used when pouring the molten metal. The proposed method
can evaluate an operator’s hand motions in terms of casting
flow. If a standard motion of each person, each machine, and
each company could be derived in future studies, casting de-
sign could come to be performed based on the standard mo-
tion. Moreover, casting defects could come to be analyzed
based on standard motions. Furthermore, the proposed meth-
od will support pour training based on simulation. These tech-
nologies are expected to engender higher-quality casting in the
future.
5 Conclusions
This paper described extension of a simulation method pro-
posed by the authors in an earlier paper for use of ladle mo-
tions as input. The usefulness of this method was demonstrat-
ed for measuring operator movements during casting
operations.
1. During pouring experiments, ladle motions were mea-
sured using the authors’ measurement techniques.
Behaviors of the molten metal were observed using a
mold with quartz glass. Thermo-fluid simulations were
1 executed using the proposed method based on the mea-
sured ladle motions. Comparison of the experiments and
simulation results clarified that the proposed method can
Int J Adv Manuf Technol (2020) 111:371–382 381

accurately predict the molten metal contour shape and of the mold particle j, and h expresses the heat transfer coef-
the tendency of the molten metal tip position to vary over ficient between the mold and molten metal.
time: both are results of the ladle motion. Assuming that heat is transferred only in the vertical direc-
2. However, simulations generally conducted under an in- tion at the boundary surface, then Eq. (10) can be described as
flow condition of the constant flow rate led to difficulty Eq. (11).
in predicting differences in the molten metal flow behav-
λi λ j hrv
ior because of differences in ladle operation. Comparison λ¼
 ð11Þ
of the proposed simulation and the simulation under in- λi λ j þ αij λi hrij þ 1−αij λ j hrij
flow conditions of a constant flow rate clarified the im-
Therein, rv denotes the distance in the direction vertical to
portance of executing the simulation based on ladle mo-
the mold surface from mold particle j to molten metal particle
tion for high-accuracy prediction of molten metal
i. Therefore, the heat transfer between the mold and the molten
behavior.
metal is calculable by replacing 2λiλj/(λi+λj) in Eq. (6) with λ
Acknowledgments This paper is based partially on results obtained from
in Eq. (11).
a project commissioned by the New Energy and Industrial Technology
Development Organization (NEDO). The authors thank Dr. Tetsuyoshi
Suzuki, National Institute of Advanced Industrial Science and
Technology (AIST), who cooperated with us in conducting the Mold Surface
experiments.
Particle j =
Funding This paper is partially based on results obtained from a project
commissioned by the New Energy and Industrial Technology (mold)
Development Organization (NEDO). Normal

Appendix 1.
Mold
2λiλj/(λi+λj) in Eq. (6) signifies thermal conductivity between
two particles having different thermal conductivity. Typically, Particle i
heat transfer between a mold and molten metal is calculated by (molten metal)
Temperature

specifying a heat transfer coefficient. However, because Eq.

(6) does not incorporate a heat transfer coefficient h, calculat-
ing the heat transfer using the heat transfer coefficient is dif-
ficult. Therefore, this study derives equivalent conductivity
expressing heat transfer as described hereinafter. As presented
in Fig. 16, one must consider the heat transfer which occurs
when a mold surface exists between a particle expressing a 0
mold (mold particle) and a particle expressing molten metal
(molten metal particle). Figure 16 shows that the distance Distance from particle j
from the center of the mold particle j to the surface can be Fig. 16 Heat transfer model with the boundary surface placed between
described as αijrij (0 ≤ αij ≤ 1), where rij denotes the distance the mold particle and molten metal particle
between mold particle j and molten metal particle i. In this
case, equivalent conductivity λ expressing heat transfer be-
tween the mold and molten metal can be expressed as Eq.
(10). References
rij 1. Lewis RW, Ravindran K (2000) Finite element simulation of metal
λ¼

1 1−αij rij αij rij casting. Int J Numer Methods Eng 47:29–59
þ þ 2. Monaghan JJ (1994) Simulating free surface flows with SPH. J
h λi λj Comput Phys 110(2):399–406
λi λ j hrij 3. Koshizuka S, Oka Y (1996) Moving-particle semi-implicit method
¼
 ð10Þ for fragmentation of incompressible fluid. Nucl Sci Eng 123(3):
λi λ j þ αij λi hrij þ 1−αij λ j hrij 421–434
4. Cleary PW (2010) Extension of SPH to predict feeding, freezing
In that equation, λi stands for the thermal conductivity of and defect creation in low pressure die casting. Appl Math Model
the molten metal particle i, λj denotes the thermal conductivity 34(11):3189–3201

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

382
5. Kazama M, Suwa T (2014) Numerical simulation of the solidifica-
tion of the melted metal by the particle method. In: Proceedings of
WCCM XI - ECCM V - ECFD VI, Barcelona
6. Ichimiya M, Sakai Y (2013) Development of filling and solidifica-
tion simulation using smoothed particle hydrodynamics. J Jpn
Foundry Eng Soc 85(8):481–488 (in Japanese)
7. Hirata N, Anzai K (2008) Heat transfer and solidification analysis
using particle method. J Jpn Foundry Eng Soc 80(2):81–87 (in
Japanese)
8. Hirata N, Anzai K (2017) Numerical simulation of shrinkage for-
mation behavior with consideration of solidification progress dur-
ing mold filling using stabilized particle method. Mater Trans
58(6):932–937
9. Tokunaga H, Okane T, Okano T (2016) Design interface for flow
channel design integrated with highly efficient fluid flow analysis
method – application to runner design of die-casting during casting
flow simulation. J Jpn Soc Precis Eng 82(1):100–105 (in Japanese)
10. Tokunaga H, Motoyama Y, Okane T (2019) Particle method sim-
ulation for formation and flow of cold flakes in high-pressure die
casting. Int J Met 13(4):897–904
11. Iwashita K, Hakuno M (1990) Modified distinct element method
simulation of dynamic cliff collapse. In: Structural Engineering/
Earthquake Engineering (Proceedings of JSCE No. 416) 7(1):
133–142
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int J Adv Manuf Technol (2020) 111:371–382
12. Watanuki K, Hou L (2010) Augmented reality-based training sys-
tem for metal casting. J Mech Sci Technol 24:237–240
13. Iwamoto K, Tokunaga H, Okane T (2015) An instruction method of
3D task motion with stereoscopic video see-through display and its
application to pouring task. In: Proceedings of the Fifth IEEE
International Conference on Control Systems, Computing and
Engineering (ICCSCE2015), George Town, pp. 309–314
14. Motoyama Y, Iwamoto K, Tokunaga H, Okane T (2020)
Measuring hand-pouring motion in casting process using augment-
ed reality marker tracking. Int J Adv Manuf Technol 106:5333–
5343
15. Wendland H (1995) Piecewise polynomial, positive definite and
compactly supported radial functions of minimal degree. Adv
Comput Math 4:389–396
16. Kondo M, Koshizuka S, Suzuki K, Takimoto M (2007) Surface
tension model using inter-particle force in particle method. In:
Proceedings of ASME/JSME 2007 Fifth Joint Fluids Engineering
Conference, San Diego, FEDSM-37215, pp. 93–98
Publisher’s note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

onlineservice@springernature.com