International Journal of Heat and Mass Transfer 256 (2026) 127965

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: [Link]/locate/ijhmt

Additive manufacturing assisted investment casting of gyroid structure heat

exchangers with casting-modulus-driven optimization in 316 L stainless
steel castings
Chen-Ming Chen a, Chia-Yu Kao a , Yu-Hao Pan a , Zi-Xiang Hong a , Cheng-Fu Huang b ,
Sheng-Chan Lee b, Chien-Wei Chan b, Yiin-Kuen Fuh a,*
a
Department of Mechanical Engineering, National Central University, Taoyuan City 32001, Taiwan
b
GlobalTek Fabrication Co., Ltd. Xinwu Factory, Taoyuan City 327001, Taiwan

A R T I C L E I N F O A B S T R A C T

Keywords: This study presents the feasibility of fabricating triply periodic minimal surface (TPMS) heat exchangers, spe­
TPMS cifically Gyroid structure, using additive-manufacturing-assisted investment casting. A casting-modulus-guided
Gyroid structure design approach was employed to optimize the gating system, enabling directional solidification and reducing
Heat exchanger
casting defects in 316 L stainless steel. Three gating designs were simulated and evaluated for filling behavior,
Additive manufacturing
temperature distribution, and shrinkage prediction, then validated by experimental casting observations. The
Investment casting
316 l stainless steel results showed close agreement in defect location and geometry retention. Flow and heat transfer tests were
Casting modulus conducted using water at laboratory scale, with equal hot and cold flow rates varied between 2 and 8 L/min. The
measured pressure drop and temperature difference trends matched steady-state COMSOL simulations, with
deviations mostly within 5–22%. While the relatively large 30 mm unit cell size limited the specific surface area
and led to lower Nusselt numbers and thermal effectiveness compared to literature benchmarks, the cast
structures exhibited reduced pressure loss and robust hydraulic stability. The use of additive manufacturing in
patterns making successfully addressed the geometric limitations of traditional wax molds, enabling the repli­
cation of complex TPMS structures. These results confirm that investment casting, when guided by casting
modulus principles, provides a viable pathway for producing complex TPMS heat exchangers, with future im­
provements in mold material and unit cell miniaturization expected to unlock higher thermal performance.

surface that partitions space into interpenetrating labyrinths. This ar­

chitecture yields a high specific surface area and isotropic mechanical
1. Introduction
properties, which contribute to improved heat transfer performance and
mechanical robustness [6,7]. Several studies have demonstrated the
Due to the rapid development of advanced thermal regulation
superior flow uniformity and reduced pressure drops in Gyroid-type
technologies, the growing demand for compact and high-efficiency heat
TPMS exchangers compared to conventional fin or plate designs [8,9].
exchangers (HE) has increased in critical sectors such as renewable en­
Current fabrication techniques for TPMS-based heat exchangers
ergy, aerospace propulsion, and high-density electronics cooling [1–3].
predominantly rely on metal additive manufacturing (AM), particularly
Compact heat exchangers offer substantial advantages due to their
electron beam melting (EBM) or selective laser melting (SLM), which
minimized volume, enhanced surface area/volume ratio, and improved
can produce geometrically intricate structures with minimal post-
thermal performance, which renders them particularly valuable in ap­
processing [10–12]. Despite their geometric freedom, these additive
plications where spatial constraints and energy efficiency are key con­
manufacturing processes face critical challenges, including high opera­
siderations [4,5].
tional costs, long build times, and limited scalability for mass production
Among the various geometrical configurations explored, HE utilizing
[13,14]. Additionally, the inherent layer-by-layer deposition method
Triply Periodic Minimal Surfaces (TPMS) architectures have garnered
can introduce anisotropic material properties and residual stresses,
growing research interest. Among these, the Gyroid structure stands out
compromising structural consistency [15].
for its mathematically defined, continuous, and non-self-intersecting

* Corresponding author.
E-mail address: mikefuh@[Link] (Y.-K. Fuh).

[Link]
Received 30 July 2025; Received in revised form 3 October 2025; Accepted 12 October 2025
Available online 23 October 2025
0017-9310/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Nomenclatures Greek Letters

ϕ Implicit gyroid surface function
Symbols Descriptions ω Wavenumber
L Unit cell length μ Molecular (dynamic) viscosity (Pa.s)
C Isovalue constant controlling wall thickness μt Turbulent (eddy) viscosity (Pa.s)
ρ Fluid density (kg/m3) σk Turbulent Prandtl number for k
t Time (s) β* Model constant
→v Velocity vector of fluid flow (m/s) ωʹ Specific dissipation rate (1/s)
→u Mean velocity vector (m/s) σω Turbulent Prandtl number for ωʹ
P Pressure (Pa) αω Model constant
k Turbulent kinetic energy (m2/s2) β Model constant
Pk Rate of turbulence generation (J/(kg.s)) σ ω2 Model constant
F Blending function λ Thermal conductivity (W/(m.K))
cp Specific heat capacity (J/(kg.K)) ε Heat exchanger effectiveness
T Temperature field ( ◦ C or K) λω Thermal conductivity of fluid (W/(m.K))
Dh Hydraulic diameter (m)
Subscripts
Vs Wetted volume (m3)
h Hot fluid
As Wetted surface area (m2)
c Cold fluid
ṁ Mass flow rate (kg/s)
i Inlet
Q̇ Average heat transfer rate (W)
o Outlet
U Overall heat transfer coefficient (W/(m2.K))
act Actual
ΔTlm Logarithmic mean temperature difference (K)
min Minimum
d Wall thickness (m)
max Maximum
Nu Nusselt number
h Average convective heat transfer coefficient (W/(m2.K)) Abbreviation
uQ̇ Expanded uncertainty of heat load (W) TPMS Triply periodic minimal surface
c Coverage factor HE Heat exchanger
uV̇ Standard uncertainty of volumetric flow rate (m3/s) AM Additive manufacturing
V̇ Volumetric flow rate (m3/s) EBM Electron beam melting
uρ Uncertainty of fluid density (kg/m3) SLM Selective laser melting
ucp Uncertainty of specific heat capacity (J/kg.K) FDM Fused deposition modeling
PLA Polylactic Acid
uT Standard uncertainty of temperature measurement (K)
RANS Reynolds averaged Navier–Stokes
NTU Number of transfer unit

To mitigate these limitations, hybrid manufacturing approaches application scope of investment casting in high-performance thermal
combining additive manufacturing and traditional casting methods have systems.
been proposed. In this study, we explore an additive-manufacturing-
assisted investment casting strategy wherein Fused Deposition 2. Mathematical model
Modeling (FDM) printed PLA patterns are used in place of conventional
wax molds, as traditional wax patterns are incapable of replicating the 2.1. Gyroid geometry
complex geometries of TPMS structures. This approach circumvents the
geometric constraints of traditional patterning while significantly To model the internal geometry of the HE, a TPMS-based Gyroid
lowering production costs and time [16,17]. Moreover, it enables the structure was adopted [20,21]. The surface geometry is defined
application of robust engineering alloys, such as 316 L stainless steel, in implicitly by the function in Eq. (1):
complex geometries that are otherwise difficult to achieve through ad­
ϕ(x, y, z) = sin(ωx)cos(ωy) + sin(ωy)cos(ωz) + sin(ωz)cos(ωx) (1)
ditive manufacturing alone.
Furthermore, we introduce a casting-modulus-guided design meth­
where ω = 2Lπ , and L is the characteristic unit cell length. This implicit
odology to ensure directional solidification and defect minimization
function generates a smooth, continuous surface that divides the domain
during the investment casting process. The casting modulus, defined as
into distinct regions: hot fluid region (ϕ > C), cold fluid region (ϕ < −
the volume-to-surface-area ratio (V/A), serves as a critical parameter for
C), and the solid wall with finite thickness (|ϕ(x,y,z)| < C), as shown in
controlling thermal gradients and feeding efficiency in solidifying
Fig. 1(a).
metals [18,19]. By optimizing gating configurations based on modulus
To establish both the simulation and casting models, a 5.5 × 3 × 1
distribution, we aim to enhance casting quality and reduce common
Gyroid unit configuration (30 mm per unit) was used to form a 165 × 90
defects such as shrinkage porosity and incomplete filling.
× 30 mm structure, Referring to Fig. 1(b)-(d) The wall thickness was
The present study investigates the feasibility of this hybrid approach
defined as 1.5 mm, and baffles were added at both inlet/outlet regions to
for producing Gyroid TPMS heat exchangers with high fidelity and
avoid mixing between two fluids, inspired from reference [8]. And Fig. 1
structural integrity. Through integrated simulation and experimental
(e) shows the internal Gyroid structure of the cast heat exchanger, which
analysis, we evaluate the thermal-hydraulic performance and
was fabricated using the investment casting process, rather than metal
manufacturing reliability of cast 316 L stainless steel structures. The
3D printing. Unlike traditional metal printing methods, investment
findings provide actionable insights for scalable, cost-effective produc­
casting enables the realization of highly precise geometric structures,
tion of complex heat exchanger architectures and broaden the
offering better manufacturing stability and structural strength.

2
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 1. Assembled Gyroid unit cell and its geometric configuration, including its decomposition into distinct regions and the actual casting part. (a) Gyroid structure
showing the hot fluid region (ϕ >C), the cold fluid region (ϕ <− C), and the wall solid with finite thickness (| ϕ(x, y, z)∣<C); (b) Overall structure dimensions: 165
mm × 90 mm × 30 mm; (c) Each unit cell is 30 × 30 × 30 mm and is arranged in a 5.5 × 3 × 1 configuration aligned with the X, Y, and Z directions; (d) Wall
thickness is 1.5 mm, with baffles placed at both inlet/outlet ends to avoid direct contact between hot and cold flow; (e) Mid-plane cross-section of cast gyroid
structured-heat exchanger with the schematic of hot and cold water flow.

2.2. Common governing equations

∇⋅→
v =0 (3)
The fluid flow in both the casting and heat exchanger simulations is Eq. (2) is relevant during the molten metal filling stage in the casting
governed by the law of mass conservation. The general form of the process, where density variations due to temperature may occur. Eq. (3),
continuity equation, applicable to compressible flows, is expressed as on the other hand, is applied to the steady-state flow of working fluids in
Eq. (2): the heat exchanger simulation, assuming constant density. Both for­
∂ρ mulations are fundamental to accurately resolving the fluid behavior in
+ ∇⋅(ρ→
v)=0 (2) their respective domains, and are consistently employed in this study
∂t
using C3P for the casting simulation and COMSOL for the heat
In most practical scenarios of this study, the fluid is considered exchanger analysis, as both software packages solve the same governing
incompressible, and the equation simplifies to Eq. (3):

3
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

continuity equations. Table 2

Parameters used in both the simulation and the actual casting.
2.3. Materials and casting setup Material Mold wall Mold Pre- Metal Pouring Pouring time
thickness (mm) Temp. ( ◦ C) Temp. ( ◦ C) (sec)
The cast TPMS heat exchanger in this study was fabricated using 316 316L 4.5 1100 1680 7
L stainless steel through an investment casting process. Table 1 presents
the chemical composition examination of the solidus material used in
this study. Due to complex geometry of workpiece, the mold pattern [ T
]
making process was assisted using an additive manufacturing method, ρ(→ u = − ∇P + ∇⋅ (μ + μt )(∇→
u ⋅∇)→ u
u + ∇→ (4)
fused deposition modeling (FDM). The printed material and FDM ma­
chine used in this study were polylactic acid (PLA) and Creality 3D where μt represents the turbulence-induced eddy viscosity, determined
printer, respectively. Through accumulative experience in obtaining via the k-ω SST model. The governing equation for the turbulent kinetic
desired toughness structure with minimum printing time, the FDM pa­ energy k is given in Eq. (5):
rameters were set as follows; 200 ◦ C-nozzle temperature, 60 ◦ C-bed
u ⋅∇)k = ∇⋅[(μ + μt σ k )∇k] + Pk − β* ρkωʹ
ρ(→ (5)
temperature, 50 mm/s-print speed, 120 mm/s-travel speed, 0.2 mm-
layer height, and 70 %-rectilinear infill percentage. The printed PLA
where β* is a model constant, σ k corresponds to the Prandtl number for
mold pattern included tree assembly consisting of sprue head, gate, and
turbulent transport of k, and Pk denotes the generation component of
runner. To determine proper tree design, this study analyzes three
turbulent kinetic energy. The equation governing the evolution of the
different designs so-called Case 1, Case 2, and Case 3. Numerical casting
specific dissipation rate ωʹ is given in Eq. (6):
simulation was involved in this stage using C3P CAST-DESIGNER, which
specializes in mold filling and solidification analysis. ρ(→
u ⋅∇)ωʹ = ∇⋅[(μ + μt σ ωʹ )∇ωʹ] +
αω 2
Pk − ρβωʹ + 2(1 − F)
σ ω2 ρ
∇k⋅∇ωʹ
Key process parameters applied in both the simulation and experi­ k ωʹ
mental casting are listed in Table 2, including shell mold thickness, mold (6)
and pouring temperatures, and pouring time. Temperature-dependent
where σω is the turbulent Prandtl numbers for ωʹ, while σ ω2 , αω , and β are
thermophysical properties—such as viscosity, density, enthalpy, ther­
model constants. F denotes the blending function. The default constant
mal conductivity, specific heat, and solid fraction—were calculated
values provided in COMSOL were used in this study.
using JMatPro software based on the actual chemical composition pro­
The modeling of thermal transport within the fluid domain adopts
vided by the industry partner, and implemented into the simulation
energy balance formulation, applicable under steady and incompress­
material model [22], as shown in Table 3.
ible flow assumptions, as given in Eq. (7):
Fig. 2 presents the structural configurations and casting modulus
distribution of three different tree assembly designs. The system is ρcp (→
u ⋅∇T) = ∇⋅(λ∇T) (7)
divided into the sprue heads (A–C), gyroid structure (D–E), gate (F–G),
and runners (H–J). Design considerations were based not only on casting where cp is the specific heat under constant pressure, T refers to tem­
modulus (volume-to-surface-area ratio, V/A) but also on the feasibility perature, and λ corresponds to thermal conductivity.
and scalability of precision investment casting processes. The corre­ The above equations were implemented and solved using COMSOL
sponding modulus values are provided in Table 4. Multiphysics, which employs the finite element method (FEM) and in­
Regions A–C refer to three mold head designs, each corresponding to cludes k–ω SST turbulence formulation [25] for incompressible flow and
a specific case. These sprue heads were derived from existing tooling heat transfer in complex geometries.
used by the industry partner, enhancing practical relevance. Due to In the flow simulation, both hot and cold fluid inlets were defined as
proprietary know-how, the original mold head geometries are not mean velocity (u) where calculated from continuity equation as
shown in full detail; instead, simplified geometries are used in the expressed in Eqs. (2,3). The average cross-sectional flow area is
illustration. Regions D–E are gyroid structures with identical modulus approximated from the volume of fluid domain divided by the length of
settings, while F–J cover gates and runners where geometric arrange­ channel. Thanks to CAD software ability, the parameters such as vol­
ments play a critical role in controlling melt flow and feeding behavior. umes, surface areas, and lengths in the complex geometry of gyroid heat
All three designs were developed to promote directional solidification exchanger could be promptly evaluated. This study used variation of
and minimize casting defects. volumetric mass rate 2–8 L/min or roughly 0.11–0.42 m/s-mean ve­
After the best PLA printed tree assembly design was achieved, the locity. While the inlet temperatures were 353 K and 298 K for hot and
investment casting process was continued with shell mold making, de- cold fluids, respectively. The outlets were set as pressure outlets with a
waxing, metal pouring, cooling, mold breaking, and surface treatment. gauge pressure of 0 Pa, and all walls were assigned no-slip boundary
Important investment casting parameters has been summarized in conditions. The main boundary conditions applied in the HE simulations
Table 2. For more detailed process, we encourage readers to refer to our are summarized in Table 5.
previous study in manufacturing cast product since no significant dif­
ference [23]. Especially for ceramic slurry material, we used our
2.5. Data reduction
4.5-layer original formulation further discussed in our other study [24].

To investigate the thermal and fluid dynamic behavior of the Gyroid

2.4. Governing equations for HE simulation
HE, a series of key physical and dimensionless parameters were calcu­
lated based on the temperature, pressure, and velocity fields obtained
This study employed the steady-state, incompressible form of Rey­
from both simulations and experiments. The relevant equations and
nolds Averaged Navier–Stokes (RANS) equations to describe the mo­
definitions are summarized as follows.
mentum transport [25] as expressed in Eq. (4).
The Reynolds number as Eq. (8) was calculated to characterize the
flow regime:
Table 1
Chemical composition of 316 L stainless steel (wt %, base: Fe). ρcp (→
u ⋅∇T) = ∇⋅(λ∇T) (8)
C Ni Si S Mn P N Cr Mo
where ρ denotes density of the water (ρ = 998 kg/m3), u corresponds to
0.08 12 1 0.03 2 0.045 0.11 17 2.5
the inlet-mean fluid velocity as explained in the previous section, and μ

4
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Table 3
Simulation-related thermophysical data for 316 L stainless steel.
Property 316L

Liquidus temperature ( C)
◦
1338
Solidus temperature (◦ C) 1412

Fig. 2. Casting modulus regions and tree assembly features in three different design cases. (a) Case 1 uses a three-inline sprue head with top pouring; (b) Case 2
adopts a cross-type sprue head and runner; (c) Case 3 features a cylindrical sprue head and three Gyroid structures. Cases 2 and 3 both utilize bottom pouring.

Table 4 Table 5
Casting modulus (V/A) values and casting yield in three different tree assembly Description of main boundary conditions.
designs. Boundary Assigned boundary Physical input values
Region Casting modulus(mm) Applied in Case location condition

Sprue head A 8.57 1 Hot inlet Velocity inlet • Flow velocity = 0.11–0.42m/
B 9.25 2 s
C 10.78 3 • temperature = 353 K(80 ◦ C)
Gyroid Structure D 0.92 1 and 3 Hot outlet Pressure outlet • relative pressure set to 0 Pa
E 0.92 2 Cold inlet Velocity inlet • Flow velocity = 0.11–0.42m/
Gate F 4.74 1 and 3 s
G 3.56 2 • Temperature = 298 K(25 ◦ C)
Runner H 3.48 2 Cold outlet Pressure outlet • relative pressure set to 0 Pa
I 4.84 2 All wall wall • velocity = 0 (no-slip)
J 4.02 3

Casting yield Case1 Case2 Case3

50 % 38 % 45 %

5
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

is the dynamic viscosity which was assumed constant at 0.001 kg/(m⋅s),

U⋅As
corresponding to water at 20 ◦ C, in both experiments and simulations. NTU = ( ) (16)
ṁcp min
This assumption allows for consistent comparison. Dh is the hydraulic
diameter of the flow channel is defined as Eq. (9):
3. Simulation and experimental methods
4Vs
Dh = (9)
As
This chapter presents the simulation and experimental analysis of
Gyroid structures. Section 3.1 describes the casting simulation and
where Vs denotes the wetted (enclosed) volume, while As refers to the
defect validation, including the flow behavior of molten metal, tem­
entire wetted surface are.
perature distribution, and shrinkage prediction. Section 3.2 introduces
The rate of heat transfer is calculated based on the ΔT between the
the experimental setup of the heat exchanger and the simulation anal­
fluid’s inlet and outlet of each hot and cold fluid as expressed in Eq. (10):
ysis of its thermal performance.
( ) ( )
Q̇ h = ṁ h cp,h Th,i − Th,o Q̇ c = ṁ c cp,c Tc,o − Tc,i (10)
3.1. Casting simulation and experimental validation
where ṁ h , ṁ c denote the mass flow rate and cp,h , cp,c represent the
specific heat for the hot and cold fluid, respectively, at constant pres­ 3.1.1. Mold flow and solidification simulation
sure. To evaluate mold filling and solidification behavior, simulations
The thermal effectiveness of the HE is defined as the ratio of actual were performed for three gating designs (Cases 1–3), focusing on melt
heat transfer between hot and cold fluids and the maximum possible flow, temperature distribution, and shrinkage porosity.
heat transfer that occurs. Due to losses or measurement inaccuracies, to As shown in Fig. 3(a), Case 1 exhibited incomplete filling during
reduce the experimental error, the average heat transfer from both hot casting, particularly in the lower Gyroid structure and gate regions
and cold side was calculated and used. In this study, the inlet-mass flow (Regions D and F), as illustrated by a3 and a4 in Fig. 3. In the last step,
rates of hot and cold fluids were set equal and since the temperature the incomplete filling regions existed in Case 1 (a5) as highlighted with
difference was relatively small, the ṁ h = ṁ c and cp,h = cp,c assumptions white arrows. The narrow and complex geometry of the Gyroid chan­
were applied, leading to final formula as Eq. (11): nels, along with their faster cooling characteristics, reduced filling ef­
ficiency. Under the top-pouring method with only a single inlet, the

( ) ( ) ( ) ( )
Q̇ act Q̇ + Q̇ c )/2 ṁ h cp,h Th,i − Th,o + ṁ c cp,c Tc,o − Tc,i Th,i − Th,o + Tc,o − Tc,i
ε= =( h = ( ) ( ) ≈ ( ) (11)
Q̇ max Q̇ max 2⋅ ṁcp min ⋅ Th,i − Tc,i 2⋅ Th,i − Tc,i

molten metal lacked sufficient kinetic energy to overcome viscous

( )
ṁcp min is the minimum heat capacity rate between hot and cold resistance and failed to reach Region F effectively. In addition, poor gas
fluid. evacuation led to the formation of shrinkage cavities, gas porosity, and
The overall heat transfer coefficient of heat exchanger, U, can be filling voids.
directly calculated using Eq. (12). While the overall heat transfer coef­ By contrast, Case 2 and Case 3 (Fig. 3(b), (c)) adopted a bottom-
ficient itself is the inverse of sum of heat transfer coefficient of cold fluid pouring strategy with dual runners, offering more stable and contin­
(hc ), wall conductivity, and cold fluid (hh ). By assuming hc = hh = h since uous filling throughout the structure.
no significant temperature difference in this study, the relationship be­ As shown in Fig. 4, the simulation results reveal the solidification
tween overall heat transfer coefficient (U) with the average convective behavior of the three gating designs. In Case 1 (Fig. 4(a)), thermal
hotspots are observed above and below the Gyroid structure (region D).
heat transfer coefficient (h) is approximated as Eq. (13).
Notably, the lower hotspot region (F) does not connect to the higher
Q̇ casting modulus area (A) as the upper gate region, preventing effective
U= (12)
As ⋅ΔTlm directional solidification and resulting in isolated liquid zones that are
prone to shrinkage defects. In contrast, Case 2 and Case 3 (Fig. 4(b), (c))
1 1 d 1 2 d exhibit better alignment between the feeding paths and the intended
= + + ≈ + (13)
U hc λω hh h λω casting modulus distribution, thereby improving the shrinkage
compensation performance.
where d is the wall thickness and λω denotes the fluid’s thermal con­ Fig. 5 shows the predicted shrinkage porosity for the three gating
ductivity (λω = 0.6 W/m⋅K). The logarithmic mean temperature differ­ designs. Shrinkage porosity is a typical casting defect that forms when
ence is computed as Eq. (14): liquid metal contracts during solidification and there isn’t enough
( ) ( ) feeding. As the molten metal contracts upon cooling, isolated regions
Th,i − Tc,o − Th,o − Tc,i
ΔTlm = ( ) (14) that solidify first cannot be compensated with additional liquid metal,
T − Tc,o
ln Th,i resulting in voids or pores. These defects usually appear in hot spots or
h,o − Tc,i
poorly fed regions of the casting and may compromise the structural
These two heat transfer coefficient values (U and h) are very integrity. In C3P software, the numerical shrinkage porosity results are
important to calculate the Nusselt number and NTU (number of transfer calculated from analyzing solidification results, geometry of workpiece,
units) as expressed in Eqs. (15,16), respectively: and temperature history (to capture hot spots, temperature gradient,
and cooling rate). While no shrinkage defects are found in the Gyroid
h⋅Dh region, a noticeable shrinkage zone appears at the junction between the
Nu = (15)
λω Gyroid structure and the gate area (Regions D to F) in Case 1 (Fig. 5(a)),
caused by poor feeding and isolated liquid zones during solidification. In
contrast, the shrinkage located near the sprue head corresponds to the

6
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 3. Simulated filling behavior near the Gyroid region for the three gating designs. (a) Case1;(b) Case2;(c) Case3.

Fig. 4. Simulated solidification behavior near the Gyroid region for the three gating designs. (a) Case1;(b) Case2;(c) Case3. Transparent regions indicate fully so­
lidified zones; colored regions represent molten or semi-solid metal.

7
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 5. Predicted shrinkage porosity for the three gating designs: (a) Case1;(b) Case2;(c) Case3. Blue regions indicate potential shrinkage defect zones.

Fig. 6. Comparison of casting quality under different casting strategies. (a) Case 1: Casting exhibiting breakout and shrinkage cavity defects; (b) Case 2: Casting with
a cold shut defect; (c) Case 3: Gyroid structure casting with no obvious defects; (d) Bottom pouring case showing surface porosity defects.

expected outcome based on the casting modulus M design. These likely caused by excessive temperature differences between molten
shrinkage areas occur in predefined hotspots, do not impact structural metal streams converging at the top of the structure [26,27]. To verify
integrity, and are therefore not further discussed. this hypothesis, Fig. 7 presents the temperature distribution during the
pouring process.
3.1.2. Comparison of simulation results and casting observations As shown in Fig. 7(a), a significant temperature difference (ΔT ≈ 140
To validate the accuracy of the simulation analysis, actual castings ◦
C) developed at the convergence zone between the two molten metal
from the three gating designs were examined, as shown in Fig. 6. In Case streams in Case 2, resulting in incomplete fusion and a visible cold shut.
1 (Fig. 6(a)), breakout, incomplete filling, and shrinkage defects are This thermal gradient was likely caused by the excessive runner length,
observed in the red-marked areas, with dashed lines indicating the depth which led to asynchronous arrival of the molten fronts and increased
of internal cavities. In Case 2 (Fig. 6(b)), a cold shut defect is identified, localized heat loss to the mold wall. These factors together contributed

8
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 7. Numerical results of temperature distribution during the pouring process. (a) Case2; (b) Case3.

Fig. 8. Schematic of defect formation and related factors: (a) porosity types and mechanisms (red arrows indicate the molten metal flow direction during filling); (b)
defect types and process parameter correlation.

to the formation of a cold zone that hindered proper coalescence. In determining the defect types and their severity.
contrast, the temperature distribution in Case 3 (Fig. 7(b)) shows a Overall, the simulation results of the three gating strategies show
reduced ΔT of approximately 80 ◦ C in the convergence region, indi­ strong consistency with the observed casting outcomes, confirming that
cating more synchronized flow and improved thermal uniformity. This the proposed modeling approach can reliably predict both thermal
improvement is attributed to the shortened runner length and optimized behavior and defect formation during casting.
casting modulus configuration, which facilitated smoother metal
convergence and enhanced fusion performance.
In some bottom-pouring cases, although no porosity was observed 3.2. HE experiment and simulation
within the Gyroid structure, various surface defects of different shapes
and sizes were occasionally found near the top of the castings, as shown 3.2.1. Experimental setup of HE system
in Fig. 6(d). The possible formation mechanisms are summarized in To assess the thermal & hydraulic behavior of the cast 316 L Gyroid
Fig. 8(a), including air entrapment due to insufficient venting, which HE, a test system was constructed. As shown in Fig. 9, the heat
leads to localized gas accumulation and pore formation; partial shell exchanger features a cross-counter-flow design, with hot and cold fluids
spallation caused by thermal stress or uneven coating, which may result entering separate channels and exchanging heat through the Gyroid
in sand particle or oxide inclusions; and during the high temperature structure walls.
process, the evolution of reactive gas species (such as O2 and N2) dis­ The fully assembled heat exchanger, shown in Fig. 10, was sealed by
solved in the molten metal may also lead to pore formation if not welding and bonding to integrate the pipes and casing. To ensure that no
properly vented. The formation of casting defects in this study is pri­ cross-leakage or intermixing occurs across the two fluid pathways, the
marily associated with three key process parameters: shell properties, leak test was performed by preliminary flowing pure water and black-
pouring velocity, and temperature. As illustrated in Fig. 8(b), these pa­ colored ink water in the cold and hot fluid channels, respectively. The
rameters are strongly influenced by the overall design strategy, such as observations on the cold channel outlet clearly show that the pure water
the gating and tree system configuration. The interaction between was not contaminated with black-colored ink indicating no leak on the
design decisions and process conditions plays a crucial role in separating walls between cold and hot fluid channels of our heat
exchanger.

9
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 9. Schematic of the Gyroid heat exchanger model, including geometric dimensions and the flow direction.

A dual-loop system was established to provide stable hot and cold especially flowmeter, which is vital in this data curation, the measure­
flows for the heat exchange experiments. Fig. 11 shows the piping and ment uncertainty needs to be evaluated. In this study, we adopted the
instrumentation diagram (P&ID), and Fig. 12 presents the actual formula of the uncertainty propagation of the heat load measurement
experimental setup with pumps, thermometers, pressure gauges, and proposed by reference [28]. Originated from basic heat load equation,
flow meters installed at each inlet and outlet. We used DC centrifugal Q̇ = V̇.ρ.Cp .(Tout − Tin ), the expanded uncertainty per heat load is
water pumps model XTL, 24 V, 400 W manufactured by Xiamen XTL written as Eq. (17).
with 50 L/min-maximum flow rate. In this study, the operating flow rate √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )2 ( )2 ( )2 ( )2 ( )2̅
was adjusted to 2–8 L/min using a control valve. The thermometers are uQ̇ uV̇ uρ ucp uTi uTo
=c. + + + + (17)
WSS-411 radial bimetallic type manufactured by Shanghai Automation Q̇ V̇ ρ cp ΔT ΔT
Instrument, capable of − 60 up to 500 ◦ C measurements with ±0.4 ◦ C
measurement error. For pressure gauges, an economic dry pressure where c is the coverage factor. In this study, c was inputted as 2 (≈95 %
gauge air compressor with 2″ dial size, 1/4″ NPT center back mount from confidence), which is a common assumption. Since the fluid (water) was
MEASUREMAN was used, range in 0–1.4 bar measurement with ±0.028 assumed to be incompressible, the
uρ
and
uCp
are extremely small. The
ρ Cp
bar-error. Then, we used F101 Rotameter 1/2″ pipe size-type flowmeter values of 0.001 (0.1 %) and 0.005 (0.5 %) are typical assumptions for
manufactured by LORRIC, specialized for 0.2 – 20 L/min range with
both property uncertainties, respectively. In this study, the V̇ was set
±0.10 L/min error. While the power supplies were solely used to operate
varying from 2 to 8 L/min. By recording the Ti and To temperatures data
the DC centrifugal pump and did not influence the data acquisition or
for each hot and cold inlets/outlets, the average heat fluxes were
measurement system. In addition, PEX (cross-linked polyethylene)
calculated having ±24 %-maximum relative errors, showing acceptable
tubing was employed at the hot-water connections between the water
uncertainties for research-purposed small-scale HE experiments.
bath and the heat exchanger inlet because of its flexibility and high
thermal resistance. For other connections with lower thermal demand,
3.2.2. Flow and heat transfer simulation
standard laboratory tubing was used.
To examine the internal thermal and flow field distribution of the
Due to the existence of measurement errors in the instruments,

10
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 10. Photograph of the heat exchanger assembled with bonding and welding. (a) 1 mm Stainless steel plates and structural adhesive;(b) Assembly view showing
internal structure;(c) Front and side views of the final product.

Fig. 11. Schematic diagram of the experimental setup.

fabricated Gyroid heat exchanger, steady-state simulations were con­ flow direction, indicating effective heat exchange.
ducted using COMSOL Multiphysics under flow rates ranging from 2 to 8 Fig. 14 illustrates the streamlines and velocity field of the hot fluid at
L/min. Fig. 13 presents the simulated temperature contours at a flow 8 L/min, corresponding to an average inlet velocity of approximately
rate of 8 L/min. Temperature distribution is shown across both the mid- 0.42 m/s. The flow generally follows the main direction through the
plane and the cross-section 7 mm away from it (Fig. 13(a), (b)), alter­ Gyroid channels, while localized vortices and secondary flows are also
nating hot and cold channels are observed, with noticeable thermal present, which enhance local convective heat transfer. However, due to
diffusion at the interface and a stable temperature gradient along the the relatively large unit size of 30 mm used in this study, the internal

11
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 12. Experimental setup components. (a) Full view of the assembled test system;(b) Close-up around the heat exchanger. (1) thermostatic water bath, (2) pumps
× 2, (3) control valves × 2, (4) flow meters × 2, (5) thermometers × 4, (6) pressure gauges × 4, (7) tested heat exchanger, (8) PEX tubing, (9) power supplied × 2.

Fig. 13. Simulation results of temperature distribution at 8 L/min flow rate: (a) cross-sectional view at the mid-plane of the heat exchanger, and (b) cross-sectional
view 7 mm away from the mid-plane.

12
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 14. Streamline and velocity distribution of hot fluid at 8 L/min (0.42m/s), showing the internal flow behavior of the Gyroid channels.

Fig. 15. Overall HE performance: comparison of simulated and experimental outlet temperatures.

Fig. 16. Comparison of simulated and measured data for hot fluid channel: (a) Temperature difference (ΔTh ) and (b) Pressure drop (ΔPh ) between inlet and outlet.

13
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

passages are comparatively wide and gently curved, resulting in pre­ pressure drop data were further utilized in conjunction with Eqs. (8–16)
dominantly laminar flow with limited disturbances. This is consistent to derive key indicators reflecting the thermal-hydraulic behavior of the
with the moderate pressure drop and relatively low Nusselt number exchanger. A comprehensive analysis and discussion are provided in the
observed in the experiments. subsequent sections.

4. Results and discussion

4.2. Heat transfer characteristics
4.1. Experimental validation of HE
To further understand how thermal energy behaves in the invest­
To assess the general thermal behavior of the heat exchanger, Fig. 15 ment casting Gyroid HE, the experimental results of Nusselt number
illustrates the outlet temperatures of hot and cold fluids obtained from (Nu), heat flow, and temperature effectiveness (ε) were compared with
both simulation and experiment across various flow rates. A consistent the numerical data reported in [8]. In addition, simulation results of G15
trend is observed between the two sets of data, and the Max deviation and G5 were included to estimate the upper-bound thermal performance
appears at 3 L/min for the cold outlet (2.6 K, 22 %) and 4 L/min for the that could be achieved if current manufacturing constraints (particularly
hot outlet (2.1 K, 20 %). To further quantify the heat exchange and flow unit cell size and wall thickness) can be overcome. The terms of “G15”
resistance characteristics, the temperature difference (ΔTh ) and pressure and “G5” refer to gyroid unit cell size where the numbers define the
drop (ΔPh ) between inlet/outlet of hot fluid channel are compared in length in mm for one unit cell. Compared to the previous G30 design,
Fig. 16. where for one-unit cells need 30 mm, the G15 and G5 have more gyroid
As shown in Fig. 16(a), the temperature difference from inlet to unit cells in the same size of heat exchanger design. The structural dif­
outlet on the hot side decreases with increasing flow rate. This trend is ferences and the corresponding variations in surface area and volume
consistent with conventional heat exchanger behavior, where reduced are clearly presented in Fig. 17.
residence time at higher flow rates leads to lower thermal exchange. The Fig. 18(a-b) compare the thermal and hydraulic performance of
simulation results overestimate the temperature drop by approximately gyroid heat exchangers with different unit cell sizes under identical
5–22 %, mainly because the model assumes adiabatic external walls and volumetric flow rates. As expected, reducing the unit cell size (from G30
ideal thermal contact conditions, without accounting for heat loss to the → G15 → G5) enhances heat transfer because of increased surface area
ambient environment that occurs in the actual experiment. Although and flow mixing, as reflected by the higher hot-side temperature drop
this simplified setting is primarily intended for the analysis of internal (ΔTh ). However, this benefit comes with a penalty in pressure drop
convective heat transfer behavior, it also results in a slight over­ (ΔPh ). At a given flow rate, the G5 structure provides the largest ΔTh but
estimation of the predicted temperature difference. also incurs the highest ΔPh , while G30 shows the lowest ΔPh but also the
Fig. 16(b) illustrates the corresponding pressure drop across the hot weakest heat transfer. Importantly, when comparing at equal pressure
fluid side. As expected, the pressure drop increases almost quadratically drops rather than equal flow rates, the G15 structure represents a
with the increase in flow rate, due to the rise in flow velocity and loss of balanced design—delivering a significantly higher ΔTh than G30 while
kinetic energy. The simulation results again show lower values avoiding the excessive hydraulic resistance of G5. This indicates that for
compared to the experiments, which can be attributed to factors such as practical applications where both heat transfer enhancement and
surface roughness, minor manufacturing defects, or additional flow pumping power must be considered, G15 offers the most favorable
resistance from connector joints that were not fully captured in the trade-off among the tested unit cell sizes.
simulation geometry. This trade-off between thermal efficiency reduc­ As shown in Fig. 19(a), the experimental results of the G30 structure
tion and pressure drop increase with higher flow rates is consistent with exhibit a typical trend of steadily increasing Nusselt number with Rey­
typical TPMS-based heat exchanger behavior reported in the literature nolds number, consistent with the results reported in the literature [8,
[29–32]. 30]. However, when the structure is further miniaturized to G15 and G5,
The experimentally and numerically obtained temperature and simulation results show that their Nu values decrease rather than in­
crease, indicating that geometric optimization alone did not yield the

Fig. 17. Gyroid structures of different unit cell sizes (left) and the variation in surface area and single-fluid volume (right).

14
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Fig. 18. Comparison of heat transfer performance for different unit cell size: (a) Temperature difference (ΔTh ) and (b) Pressure drop (ΔPh ) between inlet and outlet
of hot fluid channel.

Fig. 19. Comparison of heat transfer performance between this study and data from [8]: (a) Nusselt number versus Reynolds number; (b) heat flow versus Rey­
nolds number.

This phenomenon can be explained by three key equations. Ac­

cording to Eq. (8), the Reynolds number is directly proportional to the
hydraulic diameter Dh . Therefore, as the structure becomes finer and Dh
significantly decreases, Re drops under the same flow rate, suppressing
flow disturbance and the development of convective strength. Further­
more, since Nu and Re follow a power-law relationship (Nu ∝ Ren), the
reduction in Re has a substantial negative impact on Nu. From Eq.
(12–13), the convective heat transfer coefficient h depends on the total
heat flow Q̇ and heat transfer area As . Although Q̇ increases in G15 and
G5, the surface area As also increases, resulting in only a limited
improvement in h. Finally, as shown in Eq. (15), the Nusselt number is
proportional to both h and Dh ; with minimal increase in h and significant
reduction in Dh , the Nu value fails to improve.
Although the overall Nu values are lower than the literature [8],
Fig. 19(b) shows that the experimental G30 sample in this study reaches
a heat flow of approximately 3 kW at Re=3000, which is significantly
Fig. 20. Temperature effectiveness versus NTU between the present experi­
lower than the 6.4 kW achieved by the Gyroid structure with a unit cell
mental and simulation results of the Gyroid HE data from [8].
size of approximately 7 mm reported in [8], representing nearly a 50 %
difference. In contrast, the simulated G15 and G5 structures, even under
expected improvement in local heat transfer performance.

15
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

lower Reynolds number conditions, achieve maximum heat flows of manufactured via additive-manufacturing-assisted investment casting.
approximately 9.5 kW and 15.5 kW, respectively—substantially sur­ Simulation and experimental validation were conducted to evaluate the
passing both the G30 result and the literature’s data. thermal and hydraulic behavior of TPMS structures in stainless steel
Fig. 20 shows the relationship between temperature effectiveness (ε) applications. Key outcomes derived from this study are outlined below:
and NTU. The effectiveness exhibits a steadily increasing trend, indi­
cating good thermal utilization at higher NTU values, consistent with • TPMS structures were successfully fabricated using investment
the results reported in the literature. However, under comparable NTU casting assisted by 3D-printed molds, achieving high part density and
conditions, the effectiveness values of the G30 experimental sample quality. Compared to conventional metal 3D printing, the proposed
remain significantly lower than those reported in the reference, with ε approach enables the production of geometrically complex TPMS
reaching only about 0.28 at NTU ≈ 0.4, compared to over 0.5 in the components with improved scalability and lower cost, while
literature. This discrepancy is attributed to the larger unit cell size and addressing key drawbacks such as long production time and high
lower specific surface area of the structure, which limit heat transfer material expenses. In addition, the use of FDM molds overcomes the
performance. In contrast, the simulated G15 and G5 designs demon­ limitations of traditional investment casting, which is often inca­
strate notable improvements in effectiveness, with G5 reaching up to pable of producing wax patterns with intricate geometries.
0.68 at higher NTU values. Although the three curves may appear • The overall tree assembly layout was designed based on the casting
visually continuous, they represent independent datasets for different modulus (M) and process feasibility, aiming to achieve directional
structural configurations. The clear separation in their thermal perfor­ solidification and reduce casting defects. The simulation-predicted
mance confirms that reducing the unit cell size significantly enhances defects closely matched the experimental observations, primarily
heat exchange efficiency, especially as NTU increases. These results occurring at flow convergence zones and regions with insufficient
highlight the strong potential of miniaturized TPMS designs in future feeding, confirming the effectiveness of casting modulus as a prac­
high-performance heat exchanger applications. tical tool for structural design and defect control.
The results of this section suggest that, if current limitations in unit • Thermal simulation and experimental results of temperature differ­
cell size and wall thickness can be overcome, the feasibility of fabri­ ence and pressure drop showed consistent trends, with most de­
cating fine TPMS structures via investment casting assisted by additive viations falling within an acceptable range (5–22 %). This confirms
manufacturing can be further validated, revealing the process potential the accuracy of the established model, supporting its use in future
for integrated structural and functional performance. design optimization.
• Although the thermal performance indicators (Nu, heat flow, and
4.3. Manufacturing limitations and future enhancements effectiveness ε) in this study were generally lower than those re­
ported in some studies (primarily due to process-induced limitations
This section discusses the feasibility and current limitations of in­ on specific surface area) the overall trends were consistent, and the
vestment casting for TPMS heat exchangers, based on our experimental simulation results aligned well with experiments, validating the
results, and outlines future research strategies to address these feasibility and practical potential of the proposed structural design
challenges. and modeling approach.
The investment casting of Gyroid-type TPMS structures has been
demonstrated to be feasible; however, technical challenges remain when If the current limitations in size and geometric design of TPMS
fabricating smaller unit cell sizes and thin-walled geometries. Simula­ structures fabricated via FDM-assisted investment casting can be further
tions of the G15 and G5 structures indicate that geometric optimization overcome, it would help unlock the full potential of this process,
may enhance thermal performance. During the shell-making process, enabling the scalable production of high-performance TPMS heat ex­
however, ceramic slurry and sand particles often block the external pore changers and expanding their design flexibility and application scope in
openings, preventing the inner shell layers from forming properly, energy-efficient and industrial systems.
which in turn reduces shell strength. In addition, thin-walled regions are
prone to incomplete mold filling, leading to defects such as shell Data availability
breakout, cold shut, and incomplete filling. To address these issues,
future research can focus on the following improvement strategies. The datasets generated and/or analyzed during the current study are
Recent studies have shown that incorporating micro-scale SiC and not publicly available due to project confidentiality agreements, but are
Al₂O₃ powders into the slurry can significantly enhance the mechanical available from the corresponding author upon reasonable request.
strength and thermal performance of shell molds while reducing
porosity and improving resistance to thermal stress. In a 2024 study, CRediT authorship contribution statement
Wang et al. reported that shell molds reinforced with 15 wt % SiC and
Al₂O₃ achieved post-sintering modulus of rupture values of 8.41 MPa Chen-Ming Chen: Writing – original draft, Visualization, Method­
and 8.10 MPa, respectively—representing increases of 26.3 % and 21.6 ology, Investigation, Formal analysis, Conceptualization. Chia-Yu Kao:
% over the original shell formulation. Additionally, the heat transfer Validation, Methodology, Data curation. Yu-Hao Pan: Visualization,
coefficient (HTC) improved substantially, reaching 960 W/m²⋅K for Software, Investigation, Formal analysis. Zi-Xiang Hong: Visualization,
Al₂O₃ and 985 W/m²⋅K for SiC in the range of 1000–1200 ◦ C, out­ Software, Investigation, Formal analysis. Cheng-Fu Huang: Supervi­
performing the original’s 900 W/m²⋅K [24]. sion, Resources, Project administration. Sheng-Chan Lee: Supervision,
Several studies have also indicated that using vacuum casting, bot­ Resources. Chien-Wei Chan: Supervision, Resources. Yiin-Kuen Fuh:
tom pouring, or other assisted casting methods can effectively improve Writing – review & editing, Supervision, Project administration, Fund­
molten metal flowability and mold filling, while reducing common ing acquisition, Conceptualization.
gravity casting defects such as cold shuts, incomplete filling, and
porosity [33–35]. These process enhancements are particularly effective Declaration of competing interest
for casting complex and thin-walled geometries.
The authors declare that they have no known financial or personal
5. Conclusions relationships that could be perceived as potential conflicts of interest
related to this work.
This study is based on a casting-modulus-guided design to investigate
the fabrication and performance of Gyroid heat exchangers

16
C.-M. Chen et al. International Journal of Heat and Mass Transfer 256 (2026) 127965

Acknowledgement [17] V.K. Tiwary, A. P, A.S. Deshpande, N. Rangaswamy, Surface enhancement of FDM
patterns to be used in rapid investment casting for making medical implants, Rapid
Prototyp. J. 25 (5) (2019) 904–914.
The authors gratefully acknowledge the experimental equipment [18] A.A. Chalekar, A.A. Somatkar, S.S. Chinchanikar, Designing of feeding system for
support the Globaltek Fabrication Co., Ltd provided. investment casting process–A case study, J. Mech. Eng. Automat. 5 (2015) 15–18.
[19] B. He, Y. Lei, M. Jiang, F. Wang, Optimal design of the gating and riser system for
complex casting using an evolutionary algorithm, Materials 15 (21) (2022) 7490.
References [20] A.H. Schoen, Infinite Periodic Minimal Surfaces Without Self-Intersections (Vol. 5541),
National Aeronautics and Space Administration, 1970.
[1] T. Dixit, E. Al-Hajri, M.C. Paul, P. Nithiarasu, S. Kumar, High performance, [21] P. Tran, C. Peng, Triply periodic minimal surfaces sandwich structures subjected to
microarchitected, Compact Heat Exchanger Enabled by 3D Printing, 210, Applied shock impact, J. Sandw. Struct. Mater. 23 (6) (2021) 2146–2175.
Thermal Engineering, 2022 118339. [22] Sente Software Ltd., JMatPro (Version 7.0) [Computer Software], 2015. htt
[2] M. Alteneiji, M.I.H. Ali, K.A. Khan, R.K.A. Al-Rub, Heat transfer effectiveness ps://[Link].
characteristics maps for additively manufactured TPMS compact heat exchangers, [23] K.Y. Chen, S. Khoiruddin, C.F. Lin, M.H. Ho, C.F. Huang, S.C. Lee, C.W. Chan,
Energy Storage and Sav. 1 (3) (2022) 153–161. I. Mardiono, Y.K. Fuh, Comparative evaluation of insulation cotton configurations
[3] D.P. Sekulic, Compact heat exchangers. Handbook of Thermal Science and on the investment casting quality of industrial valve parts, Int. J. Adv. Manuf.
Engineering, Springer, Cham, 2018, pp. 1501–1520. Technol. 134 (2024) 1179–1193.
[4] M.G. Gado, O. Al-Ketan, M. Aziz, R.A. Al-Rub, S. Ookawara, Triply Periodic [24] Y.C. Wang, C.Y. Kao, C.M. Chen, C.F. Huang, Y.C. Liu, S.C. Lee, Y.K. Fuh, Effects of
Minimal Surface Structures: design, Fabrication, 3D Printing Techniques, State-of- Al₂O₃ and SiC micro-scale powder addition on mechanical, thermal and physical
the-Art Studies, and Prospective Thermal Applications for Efficient Energy properties of reinforced shell for investment casting, Ceram. Int. 51 (6) (2025)
Utilization, Energy Technol. 12 (5) (2024) 2301287. 7217–7228.
[5] M. Beer, R. Rybár, Optimisation of Heat Exchanger Performance Using Modified [25] COMSOL AB, COMSOL Multiphysics® CFD Module User’s Guide (Version 6.1)
Gyroid-Based TPMS Structures, Processes 12 (12) (2024) 2943. [Software Manual], COMSOL AB, 2022. [Link]
[6] C. Renon, X. Jeanningros, A numerical investigation of heat transfer and pressure [26] Y.S. Feng, D.M. Liao, T. Chen, Confluence and cold shut computation based on time
drop correlations in Gyroid and Diamond TPMS-based heat exchanger channels, field in casting simulation, China Found. 18 (2) (2021) 101–109.
Int. J. Heat Mass Transf. 239 (2025) 126599. [27] T.Y. Chen, Y.C. Wang, C.F. Huang, Y.C. Liu, S.C. Lee, C.W. Chan, Y.K. Fuh,
[7] X. Su, Y. Zhang, Y. Rao, K. Yeranee, X. Wang, Experimental and Numerical Study Formation mechanism and improved remedy of thermal property of cold shut
on Flow and Heat Transfer Characteristics of Additively Manufactured Triply surface defects in Vortex Flow Meters: numerical simulation and experimental
Periodic Minimal Surface (TPMS) Heat Exchangers for Micro Gas Turbine, verification in investment casting of 316 L stainless steel, J. Manuf. Process. 120
Aerospace 12 (5) (2025) 416. (2024) 542–554.
[8] K. Kus, M. Wójcik, Z. Malecha, Z. Rogala, Numerical and experimental [28] R.A. Tatara, G.M. Lupia, Assessing heat exchanger performance data using
investigation of the gyroid heat exchanger, Int. J. Heat Mass Transf. 231 (2024) temperature measurement uncertainty, Int. J. Eng., Sci. Technol. 3 (8) (2011)
125882. 1–12.
[9] T. Hirokawa, H. Miyata, Experimental investigation on pressure drop [29] B.W. Reynolds, C.J. Fee, K.R. Morison, D.J. Holland, Characterisation of heat
characteristics of adiabatic two-phase flow in a Gyroid-structured channel, Int. J. transfer within 3D printed TPMS heat exchangers, Int. J. Heat Mass Transf. 212
Multiph. Flow 181 (2024) 104982. (2023) 124264.
[10] O. Al-Ketan, R. Rowshan, R. Al-Rub, Topology-mechanical property relationship of [30] M.Z. Saghir, M. Yahya, Convection Heat Transfer and Performance Analysis of a
3D printed strut, skeletal, and sheet based periodic metallic cellular materials, Triply Periodic Minimal Surface (TPMS) for a Novel Heat Exchanger, Energies 17
Addit. Manuf. 19 (2018) 167–183. (17) (2024) 4275.
[11] J. Liu, D. Cheng, K. Oo, T.L. McCrimmon, S. Bai, Design and additive [31] R.J. Singh, S. Vijayavenkataraman, T. Zhang, S. Kumar, Thermal performance and
manufacturing of TPMS heat exchangers, Appl. Sci. 14 (10) (2024) 3970. structural stability of gyroid heat exchanger for supercritical CO2 cycle, Appl.
[12] N. Bastola, M.P. Jahan, N. Rangasamy, C.S. Rakurty, A review of the residual stress Therm. Eng. (2025) 127015.
generation in metal additive manufacturing: analysis of cause, measurement, [32] J. Iyer, T. Moore, D. Nguyen, P. Roy, J. Stolaroff, Heat transfer and pressure drop
effects, and prevention, Micromachines 14 (7) (2023) 1480. characteristics of heat exchangers based on triply periodic minimal and periodic
[13] M. Baumers, L. Beltrametti, A. Gasparre, R. Hague, Informing additive nodal surfaces, Appl. Therm. Eng. 209 (2022) 118192.
manufacturing technology adoption: total cost and the impact of capacity [33] S.L. Sin, D. Dubé, Influence of process parameters on fluidity of investment-cast
utilisation, Int. J. Product. Res. 55 (23) (2017) 6957–6970. AZ91D magnesium alloy, Mater. Sci. Eng.: A 386 (1–2) (2004) 34–42.
[14] H. Piili, A. Happonen, T. Väistö, V. Venkataramanan, J. Partanen, A. Salminen, [34] V.H. Carneiro, H. Puga, N. Peixinho, J. Meireles, Thin-rib and high aspect ratio
Cost estimation of laser additive manufacturing of stainless steel, Phys. Procedia 78 non-stochastic scaffolds by vacuum assisted investment casting, J. Manuf. Mater.
(2015) 388–396. Process. 3 (2) (2019) 34.
[15] A. Fedorenko, B. Fedulov, Y. Kuzminova, S. Evlashin, O. Staroverov, M. Tretyakov, [35] M. Raza, R. Svenningsson, M. Irwin, Experimental study of the filling of thin-walled
I. Akhatov, Anisotropy of mechanical properties and residual stress in additively investment castings in 17-4PH stainless steel, Metall. Found. Eng. 41 (2) (2015) 85.
manufactured 316L specimens, Materials 14 (23) (2021) 7176. -85.
[16] C.M. Choe, W.C. Yang, U.H. Kim, B.G. Ri, M.S. Om, Manufacture of centrifugal
compressor impeller using FDM and investment casting, Int. J. Adv. Manuf.
Technol. 118 (2022) 173–181.

17