Additive Manufacturing (AM) is a process for creating three-dimensional parts by layering material, typically guided by digital 3D model data. Compared to conventional manufacturing techniques, AM offers several benefits:

• Design freedom: AM allows the production of designs with complex geometries.
• Efficient material usage: This technique often requires less material than traditional subtractive methods.

With advancements in metal AM, significant cost reductions and enhanced precision have bolstered practical applications. For example, General Electric (GE) has implemented metal AM in fabricating aircraft engine components, reducing part counts and achieving lighter assemblies (Game Changer: Four Parts Proving Additive Manufacturing Can Compete with Casting on Cost | GE Aerospace News).

Additionally, the development of new materials and processes has spurred innovative transformations within the manufacturing sector. Notably, applications related to thermal management and cooling are attracting attention. New designs for cooling devices and heat exchangers leverage the design flexibility provided by AM. For instance, cold plates for automotive power electronics now offer three times the surface area compared to traditional designs. Consumer products like cameras have experienced a 24% improvement in heat dissipation, showcasing AM's revolutionary impact on thermal management and cooling technologies.

Technical challenges in designing cooling devices using additive manufacturing

There are several technical challenges in designing cold plates and heat exchangers for thermal management and cooling applications using AM, including:

• Optimising flow: To achieve efficient cooling, it is essential to optimise the flow channel shape and arrangement. Even if a flow channel has high local heat transfer, cooling performance may decrease if the flow is not uniform. For example, reducing the pitch of lattice structures can improve the uniformity of the flow, but narrowing the flow channels increases pressure loss, thus requiring a balanced design based on specific requirements.
• Optimising heat transfer performance: Heat from the cooling target is dissipated through the metal lattice structure into the fluid. A thicker metal layer reduces thermal resistance and improves heat dissipation, but narrowing the flow channels increases pressure loss. The basic lattice structure shape, called a unit cell, offers various choices such as gyroid, diamond, and Schwarz primitive structures. The challenge lies in finding a design that is both flow and heat-transfer-efficient from a vast design space.

In addition to fluid and thermal performance, manufacturability must also be evaluated. Even if thermal performance is high, if excessive support material is needed during production, productivity may drop, and costs may rise. Additionally, thermal stress during the sintering process may cause deformation, resulting in poor performance. Therefore, it is crucial to find a design that considers the impact of manufacturability.

Challenges in CAE analysis for additive manufacturing

In most Computer Aided Engineering (CAE) workflows, a 3D shape is created in CAD and then imported into CAE tools to generate a volumetric mesh. However, AM-specific modelling tools use implicit representation instead of boundary representation (B-rep) of the shape of the part. Implicit representation expresses objects using a field that determines whether a point in space is inside or outside the object. This allows for very complex and flexible designs, making it an ideal modelling method for AM.

However, using implicit representation in CAE presents challenges. Many CAE tools require a surface mesh to generate a volume mesh. Therefore, a surface mesh, such as an STL file, is created in CAD and imported into the CAE tool, or the CAE tool itself has the ability to generate a surface mesh from CAD data.

In the case of implicit representations, it is also necessary to create a surface mesh. However, since CAE tools do not support implicit representations, it is necessary to generate a surface mesh-like STL using a modelling tool for AM or another tool.

While B-rep allows for adjustments in mesh density and surface reproduction, implicit representation generates surface meshes solely from a field representing the object.

Figure 1 shows the workflow of modelling with nTop and performing CAE analysis. If high-resolution B-rep surface output (STL mesh) is required, it may take several hours to generate, and the resulting large data sets increase CAE processing time. On the other hand, reducing the resolution shortens output time but compromises the accuracy and reliability of the CAE analysis.

Fig 1: CAE workflow in additive manufacturing design

Development of a CFD mesh generation technology without STL

A new method for generating CFD meshes without the use of the STL mesh has been developed. The tools utilised in this approach are nTop for modelling and Hexagon’s Cradle CFD scSTREAM for CFD analysis. A brief overview of their respective characteristics is provided.

nTop is a modelling tool based on implicit representation, enabling the easy creation of shapes tailored for additive manufacturing. Additionally, it offers a third-party library called nTop Core, which allows querying the presence or absence of objects at arbitrary spatial coordinates.

Cradle CFD scSTREAM is a CFD tool employing orthogonal grids, commonly used in thermal analysis of electronic devices and HVAC (heating, ventilation, and air conditioning) systems in buildings due to its efficient meshing capabilities. It supports voxel elements based on orthogonal grids and cut-cell elements, which are created by slicing voxel elements along CAD geometry.

The orthogonal grid generation and the element property determination mechanisms in scSTREAM exhibit high compatibility with the implicit representation concept. By utilising the object presence information obtained via nTop Core during the CFD meshing process in scSTREAM, CFD mesh elements can be generated without the use of STL. Further optimizations enable high-resolution analysis while maintaining fast processing speeds, even when the number of mesh elements becomes significantly large.

The procedure for using this method is as follows:

1. In nTop, lattice structures are generated for designated parts within CAD models.
2. In scSTREAM, the usual settings are applied, and CFD meshing is performed on the CAD model before generating the lattice structure.
3. The "Export Implicit Body" block in nTop is used to output an implicit file.
4. Finally, the implicit file is associated with the part using the nTop-scSTREAM interface, and the CFD mesh elements within the specified part are converted into a lattice structure.

It is also possible to utilise multiple implicit files, enabling applications beyond cold plates, such as heat exchangers with multiple fluids and metals.

This development facilitates the rapid generation of CFD meshes and analysis without relying on STL. In addition to the method presented above, we have developed a more advanced technique utilising cut-cell elements. As with the previous method, the presence of the object is obtained from nTop Core for element generation, but in this case, intersection information between the Cartesian mesh and the object is retrieved to create cut-cell elements. The cut-cell elements have the capability to divide the elements within the Cartesian grid into polyhedra composed of planar surfaces, thereby enhancing shape fidelity with fewer elements. Users can choose between a robust voxel element-based method and the more advanced cut-cell method.

Application example

This method was applied to Ricoh's aluminium binder jet case. The application is a cold plate and its thermal fluid analysis is presented. The cold plate measures 80 mm x 80 mm x 20 mm and is made of aluminium, with water as the cooling fluid. A lattice structure is generated by periodically arranging gyroid unit cells in the centre of the flow path. The mesh resolution for CFD is set at 0.2mm, with a total element count of 31 million, using voxels for element generation.

Fig 2: Mesh generation with implicit modelling.

	Old(STL)	New(implicit)	Comparison
Mesh generation time	260[min]	12[s]	1300x faster
Mesh volume of cold plate[cm3]	85.5	85.8	Error 0.4%

Table 1: Comparison of meshes by method

In Table 1, the mesh generation time is compared, showing the time required from data output in nTop to CFD mesh generation in scSTREAM. While the conventional method took 260 minutes, the new implicit workflow reduced this time to twelve seconds. This improvement is attributed to skipping the STL generation process and the significantly faster CFD meshing enabled by the new method.

The CFD mesh volume of the cold plate's analysis mesh generated by each method is also compared. A slight difference in volume is observed due to variations in element determination, but the error is only 0.4%, indicating that the shape is almost identical to the conventional method.

Fig 3: Temperature and pressure distribution in lattice structure

	Old(STL)	New(implicit)	Error[%]
Thermal resistance[K/W]	0.02877	3499	0.1
Pressure drop[Pa]	0.02881	3416	1.0

Table 2: Thermal resistance and pressure drop by method

Figure 3 shows the temperature and pressure distributions predicted by the CFD analysis using the new implicit workflow and the conventional method, both of which display nearly identical results. Table 2 presents the thermal resistance and pressure drops at observation points, and it is suggested that the minor errors stem from the slight volume difference. Nevertheless, both thermal resistance and pressure drop exhibit very small errors.

The significant reduction in time from lattice structure shape generation to completing the CFD analysis facilitated a design exploration to optimise the lattice structure by adjusting its parameters. Two parameters were set: the cell size of the lattice structure and the thickness of the aluminium in the gyroid structure as shown in Figure 4. Based on the pressure distribution results from the previous analysis, the thickness was made variable, with high-pressure regions having thinner walls and low-pressure regions having thicker walls. In the design exploration, the maximum thickness ranged from 1 to 3 mm, while the minimum thickness was uniformly set at 1 mm. As the two parameters were set to three levels each, a total of nine cases were analysed.

Fig 4: Lattice structure design space for optimization

The upper part of Figure 5 illustrates the thermal resistance within the design space, indicating that the smallest thermal resistance is achieved when the cell size is 8 mm and the maximum thickness is 3 mm. The lower part of Figure 5 shows the temperature distribution for the cases of minimum and maximum thermal resistance, demonstrating that greater temperature diffusion occurs in the case with lower thermal resistance within the lattice structure. This thermal resistance would be even lower with the larger design space, smaller cell size, and larger thickness ratio. However, the pressure loss will be extremely high due to the reduced space for liquid flow, and the smaller space will make the depowdering process difficult in terms of manufacturing.

Fig 5: Thermal performance of the design space

A performance comparison was conducted between a traditional M-shaped channel and a gyroid-structured channel optimised through CFD analysis. The traditional design exhibited higher thermal resistance and a wider temperature distribution, whereas the gyroid-structured channel showed lower thermal resistance and a more gradual temperature distribution. Furthermore, it was demonstrated that thermal resistance was reduced by 30% with the same pressure loss. When applied to electric vehicle (EV) IGBTs, this improvement is expected to contribute to motor miniaturisation, enhanced power efficiency, and reduced CO2 emissions. Additionally, motor output was increased, suggesting enhanced acceleration performance. This application case study indicates that adopting gyroid channel structures can significantly improve cooling performance.

Fig 6: Comparison of thermal performance between conventional and additive manufacturing

Conclusion

Advancements in additive manufacturing technologies are bringing the development of innovative cooling devices closer to practical realisation. However, design exploration using CAE tools is required for design optimisation, and the time-consuming conversion of the complex shape data to STL data remains a challenge.

To address this issue, a new method for CFD mesh generation without the need for STL was developed. By utilising the implicit representation in nTop and combining it with scSTREAM’s CFD meshing technology, fast CFD analysis was achieved.

In the application case study of CFD analysis for an aluminium cold plate with a gyroid lattice structure presented, the time required for CFD mesh generation, which previously took 260 minutes using the conventional method, was reduced to 12 seconds with the new method. More importantly, this new method enabled design exploration of the lattice structure through CFD analysis resulting in an optimal shape, with experiments confirming that cooling performance improved by 30% compared to the traditional M-shaped channel, under the same pressure loss. This technology is expected to contribute to motor miniaturisation, improved power efficiency, and reduced CO2 emissions when applied to EV IGBTs.