Structural Design Strategies for the Production of Internal Combustion Engine Components by Additive Manufacturing: A Case Study of a Connecting Rod

Abstract

Topology optimization and lattice design strategies are excellent tools within the design for additive manufacturing (DfAM) workflow as they generate structurally optimal, lightweight, and complex features often difficult to produce by conventional manufacturing methods. Moreover, topology optimization approaches are quickly evolving to accommodate AM-related processes and geometric constraints. In this study, the re-design of the connecting rod of an internal combustion engine (ICE) is explored by topology optimization and lattice structures. In both topology optimization and lattice design, the objective is to maximize their structural performances while constraining material usage. Structural analyses are carried out on the optimized topologies to compare their mechanical performances with a benchmark design. Results show that the redesign of the connecting rod through topology optimization alone can realize 20% material savings with only a 5% reduction in the factor of safety. However, the combination of topology optimization and lattice structure design can result in over 50% material savings with a 21–26% reduction in the factor of safety. For manufacturability, the fast-predictive inherent strain model shows the designs through topology optimization and lattice design gives the lowest process-induced deformations before and after support structure removal.

1. Introduction

There is a projection that about 80% of powertrains will still depend on carbon-based fuels by the year 2050 [1] despite the advancements in other fuel and energy sources such as hydrogen, biofuels, solar, batteries, etc. To reduce the adverse impact of carbon-based fuels on the environment and to comply with the Fourth Industrial Revolution, it is imperative to reduce $\mathrm{C}{\mathrm{O}}_2$ emission levels to the barest minimum and deploy autonomous and advanced manufacturing processes. Two of the several ways to meet up with the objectives for ICEs are the use of lightweight materials and additive manufacturing. Considering lightweight materials, alloys of aluminum, titanium, magnesium, and other rare earth metals [2, 3] are great for ICEs however they come at high prices [4]. Some efforts have yielded the development of hybrid aluminum composites with inexpensive locally sourced materials made from palm kernel shell (PKS) and periwinkle shell (PS) [5].

Additive manufacturing (AM) has been identified as the manufacturing technology for Industry 4.0 by the World Manufacturing Forum for several reasons not limited to the potential savings in material and energy costs, the elimination of tooling, process flexibility, and the ability to produce intricate features. Considering this, deliberate efforts have gone into designing and redesigning parts suitable for AM. The rapid technological developments of additive manufacturing (AM) processes due to intensive research efforts have sustained their capabilities to produce functional end-use parts. Being a layer-by-layer production process, the realization of complex structural features is now possible, enabling the exploration of new design strategies for optimal structures. Design for additive manufacturing frameworks have been developed, and they enable the user or designer to effectively explore new design pathways. Over the years, research and industry have shown that topology optimization and lattice design form the most popular structural design strategies for additive manufacturing. While design for additive manufacturing is presented in Section 2, topology optimization and lattice designs are presented in Section 3. AM process simulation is presented in Section 4 while the original and optimized designs are compared for functionality, in the static condition, and manufacturability.

2. Design for additive manufacturing

In 2007, Rosen [6] succinctly defined Design for Additive Manufacturing (DfAM) as the “synthesis of shapes, sizes, geometric mesostructures, and material compositions and microstructures to best utilize manufacturing process capabilities to achieve desired performance and other life-cycle objectives”. There are four key aspects of the definition: multiscale structure/geometry, material, manufacturing process, and performance. To put it simply, design for additive manufacturing utilizes multiscale structures (micro, meso, and macro) and materials to achieve the desired performance while considering the additive manufacturing process. DfAM essentially sets itself apart from other design-for-manufacturing frameworks due to its ability to apply multiscale structural design approaches. Several DfAM frameworks have been introduced in the past decade, and several of these frameworks are summarized in [7]. Tang et al. [8] proposed an integrated topological and functional optimization framework, Yang and Zhao [9] suggested an additive manufacturing-enabled design framework, Primo et al. [10] developed a product design framework for AM with the integration of topology optimization, Orquera et al. [11] proposed a multifunctional optimization methodology for DfAM while Lei et al. [12] implemented an AM process model for product family design. While there are several other proposed frameworks, one thing common to most models is the adoption of an optimization tool. Taking a critical look into the models will reveal that topology optimization and lattice design form the basis of structural design methodologies. The AM-enabled design framework developed by Yang and Zhao is shown in Figure 1.

Figure 1.

Additive manufacturing-enabled design methodology. Source: Redrawn and adapted from [9].

Several researchers have applied DfAM in the design/redesign of automotive components. Bikas, Dalpadulo et al., Vaverka et al. [13, 14, 15, 16, 17] redesigned the wheel knuckles of formula race cars through topology optimization and considered laser powder bed fusion (LPBF). The redesign of a gearbox housing and automotive fixtures is seen in Barreiro et al. [18] and Naik et al. [19] respectively. For works focused on engines, Marchesi et al. [20] redesigned a diesel engine support while Barbieri et al. [21, 22] proposed that steel pistons could be used in place of their aluminum counterparts if they are “lightweighted” through topology optimization. In this chapter, topology optimization and strut- and sheet-based lattices are applied to redesign the connecting rod while making comparisons to a traditional design.

To combat expensive print runs to investigate the manufacturability of a part within the DfAM framework, developing a process model is important. A process model essentially predicts the response of a participating component or part by defining and solving the enabling physical phenomena usually through numerical or analytical mathematical approaches. In thermal-based powder-bed metal AM processes such as LPBF, being the most popular, complex multiphysics scenarios are involved. Some of the governing physics are mass transport within the meltpool, heat transfer by conduction, convection, and radiation, phase transformations, chemical reactions, Marangoni, capillary effects, etc. [7]. It is cumbersome and oftentimes impossible to consider all these physics in a high-fidelity process model to predict print success. Two major challenges in developing a high-fidelity process model are the ease of generating ideal, reliable, and validated mathematical solutions that incorporate all physics and the computational cost involved in running the models. While there have been decent attempts at developing full-scale thermal, thermo-mechanical, and thermo-metallurgical models at micro, meso, and macro scales [23, 24, 25], the need for fast-predictive models is pertinent. A widely accepted and applied fast-predictive model to calculate part deflection in LPBF is the inherent strain model (ISM) first developed by Ueda [26]. It requires strain values (often called inherent strains) calibrated from experiments run with exact process parameters intended for the actual part. These calibrated inherent strain values are utilized in a layer-by-layer pure elastic mechanical simulation to predict in-situ deformation and residual stresses within the part [27, 28, 29].

In Figure 2, a schematic diagram shows the typical workflow for ISM in a layerwise AM process. Initially, the base plate or substrate has zero displacements, strains, and stresses. When the first layer is deposited by the scan of a laser or electron beam on the powder bed, in the case of powder-bed processes, the experimentally calibrated inherent strain values are recorded to represent the load effects offered by the initial scanning process. With these strain values, equivalent force, displacement, strain, and stress vectors can be calculated. The calculated strain is referred to as the total strain with contributions from the elastic, plastic, thermal, and phase transformation strains. The inherent strain is a total of all strains except the elastic strain. Since ISM is strictly an elastic mechanical analysis, the elastic strain is obtained by subtracting the inherent strain from the total strain. The calculated stress, equivalent to the residual stress, is based on this elastic strain. The newly computed responses (displacement, strain, and stress) are added to the initial responses. Just before the analysis for the second layer commences, the summed responses are adopted as the initial responses and the whole workflow is carried out in this iterative manner till the last layer is done. To evaluate manufacturability, the maximum absolute deformation and/or residual stress are compared with acceptable values (e.g., yield/ultimate strength of the material for residual stress). For unacceptably high values, a redesign of the part can be done to either mitigate the deformation or residual stress or both. In this study, process simulations using ISM are carried out on all the connecting rod structures investigated and their maximum deformations are compared before and after support structure removal.

Figure 2.

Typical workflow of the inherent strain model for the prediction of deformation and residual stress in thermal-based metal AM processes.

3. Topology optimization and lattice design

Topology optimization is a mathematical approach that finds the best structural layout of a design problem based on an objective and subject to one or more constraints [30, 31, 32]. It is one of three major structural optimization methods with size and shape optimization being the others. It is the preferred structural optimization method because it relies the least on the user’s or designer’s experience. While an initial structural configuration must be done for size and shape optimization, only the external structural boundary is required for topology optimization. More recently, topology optimization has given rise to generative design which performs the optimization without a strict single initial boundary or objective function. Nonetheless, defining the loads, boundary conditions, optimization objective(s), constraints, and material model is common to all optimization methods. There are several topology optimization methods: density-based, boundary/phase-field methods, evolutionary, etc. Several of them are based on the finite element method and a popular formulation to optimize structural rigidity is the compliance or strain energy objective formulation subject to a material volume constraint. Although the detailed formulation is not given here, a workflow of the optimization process is shown in Figure 3.

Figure 3.

A schematic showing the main sections of the topology optimization process.

Another important tool in the DfAM framework is lattice design. A lattice structure can briefly be defined as one that comprises interconnected struts, or unit sheet features that are uniformly, pseudo-uniformly, or randomly patterned resulting in a cellular-like or meta-like design. Typically, for a lattice structure to be formed, the unit cell type, cell parameters (size, thickness, etc.), and lattice framework should be known. There are a host of benefits when lattice structures are used. They possess high strength-to-weight ratios, good heat transfer capabilities (especially surface-based lattices), negative Poisson’s ratio, enhanced osseointegration capabilities (the functional and structural connection between a bone and an implant) when used for bio-implants, and low thermal expansivities. For their application in the re-design of the connecting rod in an ICE, their high strength-to-weight ratio is leveraged.

Three design iterations using the two tools outlined in previous paragraphs are performed. The original and initial designs are shown in Figure 4a and b respectively. Figure 4c and d show the two load cases considered for the structural design where $F_c$, $F_t$, and $F_b$ are the compressive, tensile, and bending forces respectively. Table 1 shows the design material specifications and load values as obtained by Alkalla et al. [33]. The original design was obtained from Grabcad [34] while the initial design is a simpler model that excludes some fine details in the original design. Typically, when performing topology optimization, a larger and simpler initial design domain is recommended.

Figure 4.

Connecting rod showing the (a) original design, (b) initial design, (c) load case 1, and (d) load case 2.

Parameter	Value
Material	Aluminum 7075-T6
Density	2700 $\mathrm{kg}/{\mathrm{m}}^3$
Elastic modulus	70 GPa
Yield strength	450 MPa
Poisson ratio	0.33
$F_c,F_t$	4246 N
$F_b$	1061 N

Table 1.

Material specification and load values.

All design workflows are performed in nTopology (academic license) [35]. There are some open source codes [36, 37, 38] with varying capabilities of performing topology optimization, however, nTopology offers the development of lattice structures with unique interfacing capacities with topology optimization.

3.1 Topology optimization (TO) design strategy

The first design workflow is a strict topology-optimized connecting rod, shown in Figure 5. Like most structural design workflows, the initial design domain is specified along with loads, fixed locations, and preserved regions. A finite element analysis (FEA) is then performed (static analysis in this case) to enable the computation of the objective and constraint(s). The objective is compliance minimization (tantamount to stiffness maximization) while a material volume constraint is imposed at a minimum of 40% and a maximum of 60%. All the information collected is then fed into the optimization module which generates an optimized model after a set number of iterations. This optimized model is usually made up of jagged features, therefore, a post-processing smooth step is established to finalize the model.

Figure 5.

Design workflow of a connecting rod through topology optimization. TO—topology optimization.

3.2 Topology optimization and lattice (TO-L) design strategy

The second design workflow is a combination of topology optimization and lattice design using a graded lattice approach. In Figure 6, there are two color schemes in the design workflow: light brown and blue. The light brown steps indicate topology optimization while the other indicates lattice design. It is observed that the topology optimization steps are the same as the previous design workflow except that the smoothened solution is utilized to generate an implicit thickness parameter field in this workflow. This thickness parameter is needed to establish the lattice gradation. While the thickness parameter is defined, a region that determines where the lattice will be populated must be generated. The thickness parameter and the defined lattice region are both utilized to generate surface and volume strut lattices which are then combined to form the final graded lattice.

Figure 6.

Design workflow of a connecting rod through topology optimization and lattice design (graded lattice strategy). TO—topology optimization.

3.3 Topology optimization and lattice (TO-L) shell-infill design strategy

The third workflow also utilizes topology optimization and lattice design but in a shell-infill approach as displayed in the workflow in Figure 7. The smoothened model obtained from topology optimization undergoes two steps: an outer shell is created and the difference between the outer shell and smoothened model (a core) is obtained creating a shell-core model. A lattice operation is done on the solid core model for transformation to a strut- or surface-based lattice. In this work, two lattice cells are used: a strut-type body-centered cubic (BCC) and a gyroid. The shell and lattice structures are unioned to obtain the final shell-infill connecting rod as shown in Figure 7. The final shell-infill models of the connecting rod in Figure 7 show only half of the shells to reveal the infill structures.

Figure 7.

Design workflow of a connecting rod through topology optimization and lattice design (shell-infill strategy). TO—topology optimization.

4. AM process simulation

As highlighted in Section 2, ISM is used to predict the deformation and stress response of the connecting rod designs during the build process. LPBF is adopted as the AM process in this study and AlSi10Mg is used as material. For process parameters, a power of 200 W, a scan speed of 1 m/s scan, and a layer thickness of 30$\mathrm{\mu m}$ were used. Based on these parameters, two cantilevers were printed and the vertical deformations of the cantilever tips after cutting were recorded as 2 mm each. The printing process and details of calibrating the inherent strain values are beyond the scope of this chapter, and similar studies can be seen in [39, 40]. The inherent strain values used in this study were generated by the Simufact Additive®’s metal AM calibration study and were found to be ${\epsilon }_x=-0.00292,{\epsilon }_y=-0.00297,{\epsilon }_z=-0.03$ based on the deformation values from the cantilever experiments. Simufact Additive^® was also used to solve the elastic mechanical problem for all designs based on ISM.

5. Original vs optimized designs

5.1 Structural performance and material savings

Static analysis was done on all optimized models and a benchmark design. The original and optimized models are shown in Figure 8. Static analysis was performed on all designs according to Figure 4c and d and the von Mises (VM) equivalent stress distributions are shown in Figure 9. For all the designs, the maximum VM stresses are toward the crank pin end, and they all possess similar stress distributions. For the optimized designs, the graded lattice strategy is the most complicated to implement in nTopology while the topology optimized model was the fastest to realize. The reason for this is the dependence of the lattice strategies on the post-processed topologically optimized model.

Figure 8.

Versions of the connecting rod (a) original (b) topology optimized (c) topology optimized and latticed (graded) (d) topology optimized shell-lattice infill (BCC) (e) topology optimized shell-lattice infill (gyroid) design NB: the shell of the designs in (d) and (e) is sectioned along the longest axis to reveal the lattice infill.

Figure 9.

VM stress of the (a) original (b) topology optimized (c) topology optimized and latticed (graded) (d) topology optimized shell-lattice infill (BCC) (e) topology optimized shell-lattice infill (gyroid) design.

To analyze the optimality of the structures for material and function, the volume fractions, maximum displacements, maximum VM stresses, and factor of safeties are investigated and compared. The volume fraction is the ratio of the material volume of a new design to the original/benchmark design. The maximum displacement and VM stresses are the maximum nodal values of these quantities from static FEA while the factor of safety for ductile materials is utilized. Figures 10 and 11 show how the volume fraction, maximum displacement, maximum VM stress, and factor of safety vary from design to design. Typically, the lower the volume fraction, the better the design as less material is used. The higher the VM stress and displacement, the poorer the design while higher factors of safety are favorable.

Figure 10.

Final volume fraction of design (left image), maximum displacement from static analysis (right image).

Figure 11.

Maximum von Mises stress and factor of safety of all designs.

The topology-optimized-lattice (TO-L) graded design is almost equal in material amount as the original/benchmark design while the TO-L shell-gyroid design uses the least material volume. However, the shell-infill designs give considerably higher maximum displacements relative to the others, about twice higher. It is interesting to know that although the TO and TO-graded lattice designs utilize less material volume, they give lower maximum displacements compared to the original design. In Figure 11, the maximum VM stresses and factor of safety are given. The TO-L graded design has almost twice the maximum VM stresses of that of the second-placed design (shell-BCC). Moreover, it gives a factor of safety below 1 which signifies the structure is inadequate even from a static design viewpoint. The graded lattice design can be improved by investigating other cell structures and lattice thickness fields. Unfortunately, since the graded design approach also takes more man-hours to implement, it is less attractive compared to the other approaches.

The final image in Figure 12 shows a radar plot that quite nicely compares the three major design criteria: volume fraction, normalized VM stresses, and normalized maximum displacements. The figure immediately shows that the TO model is comparable in performance to the original design although with almost a 20% reduction (according to Figure 10) in material usage. However, if there are no strict requirements on the maximum VM stress and displacement, the shell-infill strategies are the best approaches since they result in models that use much less material (about 50% less than the original) while keeping their factor of safety above 1.

Figure 12.

Radar plot of all designs comparing volume fractions, normalized VM stresses, and normalized maximum displacements.

5.2 Manufacturability

In this study, residual stress and deformation before and after support removal form the basis of assessing the manufacturability of all designs. The maximum residual stresses for all designs were 370 MPa, and the plots are not shown because of this homogeneity, also, no further assessments are made to judge comparative manufacturability based on the stresses. The displacement plots after support removal are shown in Figure 13 and all designs show quite similar distributions except the topology-optimized and latticed (TO-L) design. The TO-L design shows higher displacements around the middle portion of the connecting rod as opposed to the other designs where they appear at both ends. Figure 14 reveals that the TO design offers the maximum deformation of all designs at 1.6 mm after support removal. The TO-L design also gives a maximum deformation of over 1.2 mm after support removal while the other designs are at around 0.7 mm. It is interesting to note that the TO-L shell-infill strategies give deformations similar to the original design before and after support removal while noting that they both utilize approximately 50% less material. It should be noted that these results are peculiar to the process parameters and build orientation. The deformations can potentially reduce when these parameters are optimized. Moreover, if build orientation optimization is performed for each design type, different build angles might be realized possibly influencing the final deformations.

Figure 13.

AM process-induced displacements after support removal for the (a) original (b) topology optimized (c) topology optimized and latticed (graded) (d) topology optimized shell-lattice infill (BCC) (e) topology optimized shell-lattice infill (gyroid) design.

Figure 14.

Maximum AM process-induced displacements in the designs before and after support removal.

6. Conclusion

To achieve lower carbon footprints in ICEs, reducing the weight of the components is paramount. Additive manufacturing, new design tools, and techniques can help to attain this. In this study, the use of topology optimization and lattice structures to redesign a connecting rod has shown success in this direction. Also, with the advancement of structural design tools such as nTopology, which uses field- and implicit-driven modeling concepts, these new design pathways can be implemented quite easily. This study shows that the integration of topology optimization and lattice design through a shell-infill approach can be adopted to redesign a connecting rod with up to 50% material savings while keeping the structural performance in static conditions at an acceptable level. Also, results from the ISM-based AM process simulation show lower deformations in the shell-infill designs before and after support removal compared to other designs. It should be noted, however, that the overall performance of a connecting rod, like several other components in an ICE, is dictated by other parameters/conditions such as fatigue life, damage, eigenfrequencies, etc. It is therefore recommended that further analysis be carried out on the model outcomes of these nascent design approaches for either validation or model calibration. Moreover, a broad design of experiment study related to build orientation optimization for the AM process should be investigated. Moreover, lattice-alone designs and varying lattice cell parameters should be explored in the future.