4.1 Design Approach

In Step 1 of the AM Design Workflow, a 3D model must be created. This step, much like conventional manufacturing, is the foundation of creating an AM part. Within this step, 3D solid models can be generated in any CAD software package. Generation of the 3D model can come from: (1) creating the original design in CAD as shown in Figure 4.2 (2) scanning an existing part to create a 3D model, (3) converting a 2D drawing like that in Figure 4.3 to a 3D model, or (4) downloading a model from an online design repository.

Depending on which of these four methods is selected, different levels of design requirements may be necessary. For example, if one were to create an original 3D design in CAD, the design requirements may come from the designer’s own thoughts, or perhaps the design requirements are being received from a customer. When 3D scanning an existing part, converting an existing 2D drawing or downloading an existing design, the design requirements could have been established many years ago. As such, those antiquated designs may not be the optimal for AM.

Figure 4.2 3D solid model in CAD (credit: Spencer Wright on pencerw.com, CC-BY 4.0 license)

Figure 4.3 2D drawings must be converted into 3D solid models for AM (credit: Spencer Wright on pencerw.com, CC-BY 4.0 license)

There are many commercial software choices capable of creating 3D models. Many industrial companies and universities own specific software licenses, each having unique characteristics of functional performance. A list of major 3D CAD companies at the time of writing is found in Table 4.2. Note that product names and companies change often, and new or application-specific versions may be more appropriate for your needs.

The Sculpteo Blog maintains a relatively up to date list of open source CAD software that you can use, as well.

Since each software outputs its own coding structure, AM’s primary working solution has been to use a common file language that AM machines can understand. To achieve this, 3D solid models are typically converted into a neutral .STL file that any AM build software can read and interpret. A .STL file is a faceted representation of the part that is created by tessellation, a process that approximates the boundary of an object with many triangles. Tessellation is an arrangement of polygons closely fitted together in a repeated pattern without gaps or overlapping, in this case triangles.

Historically, the .STL file was the first type of digital format developed for the rapid prototyping technology, Stereolithography. This was first developed in the 1980’s and the file protocol hasn’t changed since.

Figure 4.4 shows different levels of detail in triangulations in an .STL file.

Figure 4.4 Triangulation of a .STL file. (credit: Modification of “STL triangulations of the pipe wall,” in Morente, A. et al, “ A Highly Scalable Direction-Splitting Solver on Regular Cartesian Grid to Compute Flows in Complex Geometries Described by STL Files.” Fluids, Volume 8. 2023. CC BY 4.0)

Balancing .STL file size and accuracy is a challenge for complex 3D CAD models, especially, if the models have cellular features incorporated in them.

When exporting .STL files there can be a number of errors.

• Export Issues
  • Resolution tolerance too low, leads to low quality surface curvature
  • Multiple bodies/faces present during export – multiple shells, overlapping faces
• Conversion issues
  • Inverted normals
  • Bad contours, mesh holes
  • Overlapping/intersecting triangles

Though not all AM machines use them, the good news is that there are alternative formats to the .STL file that have been developed recently. One alternative format is the Additive Manufacturing File (AMF). This file format is an XML-based format that includes information beyond the .STL file. Specifically, the AMF includes 1) units and object metadata 2) color and texture 3) multiple materials/gradients 4) optional curved triangles.

Another alternative is the 3D Manufacturing Format (3MF) file format. It is also an XML-based format that includes 1) units and object metadata 2) color and texture 3) multiple materials/gradients 4) beam lattice elements 5) support structure information. Persuading AM machine equipment manufacturers to adopt these alternative types of formats can be difficult. Some machine manufacturers do and others do not.

Many of the large 3D CAD software companies are directly working with the AM machine manufacturers to export the native 3D CAD solid model directly to the machine code, thereby skipping the intermediate step of the .STL, AMF or 3MF formatting altogether. Though an incredibly promising advancement, only specific software companies have partnership agreements with specific machine manufacturers. This results in an unclear definition of which file format from which specific 3D CAD software is compatible with which AM machine type. Though incredibly inefficient and antiquated, the .STL file is the common language ubiquitous to software and machine manufacturers until further development.

Topology Optimization/Generative Design

Powder bed fusion allows for incredibly complex metallic structures to be created. This creativity unlocks design potential that is virtually impossible to fabricate using traditional manufacturing methods. There are two basic approaches to create organic structures for AM:

Topology optimization is a mathematical output from FEA (Finite Element Analysis). The FEA drives where material is placed such that the material layout is optimized within a design space, for a given set of loads, boundary conditions and constraints with the goal of maximizing the performance of the system for the given application. While accurate, topology optimization output can pose issues in terms of file format and detail resolution which limits viability, especially for very complex designs.

Generative design also encompasses the FEA behind topology optimization acknowledging the same criteria to optimize material, which it then utilizes to generate a viable design via an algorithm. It is often used as a design exploration strategy in which designers input multiple design goals in the form of parameters such as performance or spatial requirements, materials, manufacturing methods, and cost constraints. Historically, exploring a series of viable designs would require many hours from engineers to build them manually in CAD, but generative software uses machine learning to explore all of the possible permutations of a solution and quickly generates a number of design alternatives. The software tests and learns from each iteration based on what works and what does not.

Most of the software found in Table 4.3 offers either topology optimization or generative design as an approach to reduce weight, increasing functionality during 3D CAD model development in Step 1. Either approach is effective in creating structure that solves the objective function of design with the least amount of material necessary. Details of topology optimization and generative design can be found below.

Analysis

After a part concept has been generated using topology optimization or generative design, it is then critically important to validate that the design is structurally robust using finite element analysis (FEA) software. This software is a computerized method for predicting how a product reacts to real-world forces, vibration, heat, fluid flow, and other physical effects. Finite element analysis shows whether a product will break, wear out, or work the way it was designed. It is called analysis, but in the product development process, it is used to predict what is going to happen when the product is used.

Note that generative design software often uses artificial intelligence to perform multiple iterations of a design so that it is inherently performing FEA while the part is being generated. However, it is worthwhile to check structural robustness relative to the design intent using a different FEA software that would offer a non-biased perspective.

When the design is validated using FEA software, the software identifies areas in the structure that need to be changed to ensure safe structural integrity to the overall part. This input is then fed back to the 3D CAD systems, and another topological optimization simulation is started. This iterative approach allows for a part to maintain a minimal amount of weight while concurrently being structurally validated, although iterating is sometimes easier said than done, and the workflow is ever-evolving. Certain topology optimization and generative software programs have issues resolving models and smoothing the 3D bodies for export to analyze, especially when the export output is .STL, since most analysis tools prefer CAD or .STEP type files. STEP, which stands for STandard for the Exchange of Product Data file, is a text file that contains three-dimensional model data in a standard format.

Fortunately, optimization programs such as ParaMatters and MSC Apex run the FEA simulation as a part of the generative design process and then automatically apply smoothing for better model resolution. MSC Apex, for example, has aimed to rebuild the optimization platform in such a way that it intelligently smooths surfaces within a CAD environment to yield high-fidelity export options. Once the designer and structural engineer have agreed on the design concept, and the optimized model quality is acceptable, the part must be prepared for fabrication. A .STL file (or compatible build set up file such as .STEP or .3MF) is created and imported into build

Build Preparation

Step 2 of the AM Design Workflow is a digital build plan that prepares the part for fabrication in the AM machine. A build plan specifies the orientation and layout of the part(s) on the build platform, including layer thickness, support structures (if needed), and the toolpath for each layer. See Figure 4.5 for examples of build plans.

Often the build plan is fulfilled by a manufacturing engineer, but it is important for design engineers to be aware of the implications of the build plan on the AM part being produced. Build orientation and layout decisions will impact:

• Build time and material utilization
• Build height and number of layers
• Surface roughness, including staircasing, which is a regular interruption in the part surface that looks like a staircase.
• Thermal cycling and residual stresses
• Microstructure and mechanical properties
• Post-processing (e.g., support removal)

AM build plan software is in development with most of the large commercial 3D CAD companies.

Figure 4.5 In the preparation step, the designer can choose to display and adjust the entire build, including the supports. In this example from Flashforge, a simple interface allows the user to control which supports to display. The two Benchy (boat) examples that lie flat require no extensive supports, but the one that is tilted upward at an angle requires supports to print in that orientation. (credit: Modification of “Flashforge_Dreamer_FlashPrint_#3DBenchy_Screenshot_GCode_preview” by Creative Tools/Flickr, CC BY 2.0)

Build Orientation and Nesting

In the build plan, it is critically important to understand how AM part orientation and nesting can influence part cost. In general, the more parts one can nest in the build volume, the less costly the parts become. In a similar fashion, the shorter the build height and the less sacrificial support material required, the less costly parts become. The less time the laser/extruder is active during the build process and more minimal support structures thus post-processing steps are, the less costly parts become.

With the above general cost considerations in mind, there are also a number of process specific build preparation rules of thumb.

The table below highlights the differences in how build preparation can be different relative to the type of AM process. Note that these are general guidelines and that each type of AM machine and material combination can have special requirements that limit or enhance design freedom.

Process Simulation

Once a build plan is established and the part build orientation and support generation strategy are defined, designers need to understand if the actual fabrication process could produce issues due to common failures or issues specific to the machine, material, or process. Even if everything is planned correctly, an issue in fabrication will not only waste time and material, but could actually damage the machinery.

For example, with AM metallic processes, internal residual stress accumulates with quick heating of feedstock to melting and cooling back to solid metal during AM fabrication. As a result, parts will tend to distort during and after printing and after thermal post-processing. In addition, improperly designed support structures can delaminate from the build platform during fabrication if too much residual stress accumulates. This is especially true with large titanium structures, which have been known to distort and even crack during the build. If this happens during the powder bed laser process, there is a significant risk of the powder recoater feature in the AM machine to collide with parts that have distorted and separated from their respective support structure while printing, causing build failure. Therefore, AM Process Simulation is used to help understand the effects of the AM process on the part geometry and its fabrication.

AM Process Simulation is the analysis of the complex AM machine and material interaction using 3D CAD software specially designed to predict part distortion to parts during AM fabrication. Figure 4.6 highlights a process simulation that allows precise control over which aspect of the build is being considered, including its relationship to other parts. The simulator used here – by Siemens – accounts for collision avoidance with the laser deposition head.

Figure 4.6 AM Process Simulation Example. (credit: “NX CAM Hybrid Additive Manufacturing – Simulation and Machine Support” by Siemens PLM Software/Flickr CC-BY-ND 2.0)

The distortion of features during an AM build depends on many factors such as the feature orientation, thick to thin transition zones, thin walled structures, material type, layer thickness, process parameter settings, heat treatment selected, etc. As such, software is used to simulate the amount of residual stress inherent to the AM build and digitally pre-compensate the geometry so that after printing and heat treatment, the parts will be closer to original design intent in the 3D CAD model. Table 4.5 lists several process simulation software packages that are now readily available.

Inspection

After AM parts are fabricated, they must be inspected for dimensional compliance and mechanical integrity (step 6 of the AM Design Workflow). In addition, mechanical and metallurgical samples that were constructed concurrently with the part are collected and tested as a witness to the AM process. AM parts can be difficult to inspect with conventional coordinate measuring machines (CMM) due to part complexity and presence of internal features. AM parts are often inspected using light inspection techniques, or Computed tomography (CT) using x-rays, for instance (See Radiographic Inspection section in 5.2 Nondestructive Testing (NDT)).

There are many aspects to part inspection, and a sample list of common inspection techniques used for AM follows.

• Surface Roughness
  • Profilometry, which is the measurement of surface roughness at a fine scale – affordable and fast, but limited effectiveness at higher surface roughness values
  • 3D laser scanning microscope – expensive and slow, but highly accurate
• Dimensional Inspection and Metrology
  • Contact methods (e.g., CMM with touch probe)
  • Non-contact methods (e.g., 3D scanner with laser, structured light)
• Fluorescent Penetrant Inspection (FPI)
  • Uses fluorescent die to highlight surface irregularities – AM surface finish makes inspection difficult
• Radiographic Inspection
  • Used to find internal cracks, voids, or trapped powder – Limited use with dense parts
• Computed Tomography (CT) Scanning
  • Able to see in high-density materials
  • High-resolution digital data set for interpretation or measurement

From a design engineering perspective, it is important to understand that using extreme design complexity that AM affords may require more expensive inspection techniques.

AM Features

AM feature modifications are unique depending on the specific AM process used. Table 4.6, highlights rules of thumb for AM feature modifications needed to fabricate AM parts successfully.

Concept Ideation

The first step to beginning re-architecture of a design for AM is to understand clearly the product requirements. A good way to begin understanding these requirements is to develop a morphological analysis, a method for exploring all possible solutions to a multi-dimensional, non-quantified problem. “A Morphological Analysis defines a process for generating solutions to problems by first breaking down a problem into its parts (or subcomponents or subfunctions), generating ideas for each part, and then exploring combinations of the resulting ideas to develop a solution or concept” (Daley et. al, 2016) This approach works well for system-level design for DfAM. To create a morphological chart, you begin by listing the identified subfunctions that satisfy customer requirements. Next, you develop solutions for each subfunction.

To efficiently generate ideas for new concepts, it is helpful to conduct the exercise in a team brainstorming session to create a list of subfunctions to build the morphological chart. Ideally, the skillset required within the team should include AM design, conventional product design, AM process manufacturing, and customer representation. As DfAM structures can often become organic in nature, the AM designer shall also be well versed in the skills of biomimicry.

Biomimicry

According to the Biomimicry Institute, biomimicry can be defined as “an approach to innovation that seeks sustainable solutions to human challenges by emulating nature’s time-tested patterns and strategies. The goal is to create products, processes, and policies—new ways of living—that are well-adapted to life on earth over the long haul.” By integrating biomimicry principles that emulate nature’s structures into concept ideation, products will inherently become more structurally efficient.

DfAM of the same part shown in the MfAM section, except now the design has been completely overhauled and optimized via light-weighting that utilized biomimicry in the upper web-like structures. The ratio at which the branches grow uses the same strength ratios seen in the natural log of a tree.

Figure 4.7 Biomimicry is exemplified in the creation of artificial feathers, initially at a much larger scale than actual feathers. This design includes most aspects of the feathers, including the hooks at the ends of the barbules, which in natural feathers are invisible without magnification. (credit: Modification of “20180322-MeyersLab-Feathers-3492-export-8MP” by UC San Diego School of Engineering/Flickr, CC BY 2.0)

Functional requirements communicated directly from the customer are known as the Voice of the Customer (VOC). Based on this input from the brainstorming meetings where customer feedback is directly communicated, concept ideas are refined and modified to reflect true requirements. These requirements can be weighted within the Pugh Concept Selection Matrix, or Pugh Matrix for short, based on VOC rankings. The Pugh Matrix is a rating analysis tool that results in ranking of ideas to generate an optimal concept or selection.

Once ideas are generated, the concepts should be evaluated relative to the customer-defined requirements. Using a Pugh matrix, each idea is compared to a baseline design for each requirement as shown in Table 4.7. Ideas are then scored as better (+1), similar (0), or worse (-1) than the baseline design for each requirement. The scores are then weighted and summed, and the best concept has the highest consolidated score. With this approach, the team not only selects the best concept but also keeps track of their decision-making process should questions arise later.

In the example provided in Table 4.7, concept ideas generated in the morphological analysis brainstorming sessions are listed at concept A, B, C and D. The VOC requirements are shown on the left. Also, note the VOC relative weighting of each of the requirements on a scale of 1-5 with 5 being the most important. An existing baseline product (either a competitor’s or an older design) is listed with a baseline of 0 for all requirements. Under each concept idea the team provides an assessment score for each concept as being worse than the baseline (-1) or better than the baseline (1) or the same (0).

Next, each assessment score is simply multiplied by the VOC weights to create a sum of weighted scores as shown below.

In this case, concept idea A is far worse than the baseline design, with concept design B being better. Concepts C and D are slightly worse and better than the baseline, respectively.

Topology Optimization and Generative Design

Now that VOC requirements are understood, ideas are generated, and concepts are sorted using the Pugh Matrix, generative design or topology optimization tools can be used to embody the concept. Alternatively, biomimicry or other approaches could be used to generate different embodiments for each concept. Keep in mind that just because one can produce complex structures with AM, does not mean one should do so unless it helps satisfy the VOCs.

Concept B appears to have an edge over the baseline design on paper, how long will it take to optimize for topology or use generative design? On occasion the objective requirements provided by the VOC might be satisfied solely with traditional fabrication methods. However, if they cannot be satisfied with traditional fabrication, then one must turn to additive manufacturing.

Below is the process flow that takes digital definition from VOC 3D CAD definition all the way to inspection. Many types of software exist at each of the steps of the digital chain.

The existing CAD will be assessed, FEA analysis run (within the optimization software), optimization applied, and then additional analysis run on the optimization. The optimized design will be prepared for build. The build plan will be simulated for build feasibility. The part will be built, and later inspection will provide empirical data to help the designer understand how accurate their digital tools are, and design improvements that can be made in the future. What is most important from a DfAM standpoint though, is the transition from CAD Definition to Topology Optimization (or Generative Design).

Another DfAM nuance is applying systems-level thinking when using generative design tools. Designer’s concepts can be used for part consolidation as well as making a structure lighter.

Figure 4.8 Using a generative AI process, a part can be optimized for weight while maintaining strength and shape. (credit: “NX8 Simulation - Opt” Siemens PLS Software/Flickr CC BY ND 2.0)

Generative design offers a glimpse into the future direction of hardware design. Using AI inspired software to derive a large number of design permutations will lead to faster product design cycles that are more optimized for the hardware requirements.

The basic generative design or topology optimization workflow is:

1. Define retained bodies – in other words, volumes on the part that should not be optimized
2. Assign material type – this loads the property inputs for the optimization algorithms
3. Set load cases and boundary conditions – these are the forces or loads the part will see, and surrounding boundaries that contain the part to a certain location in space
4. Set the objective – usually mass reduction
5. Set constraints such as volume fraction, model resolution, design speed, and manufacturing overhang angle
6. Optimize! Sometimes this can take hours or days depending on complexity and computing power

The digital chain is completed when the chosen design is input into build preparation software to create a build plan, which is then run through a process simulation tool to assess buildability. Once deemed feasible to build, the part can be sent to the machine to be built. After building, empirical results from inspection are then compared to simulated process results to advise and correct future designs and process simulations.