Generative Slicing: The Future of Additive Manufacturing?

Generative Slicing Design

Generative Slicing: The Future of Additive Manufacturing?

By: Greg Vialle

The design-manufacture paradigm is undergoing a change. For most of my 25-year engineering career, there has been a pretty clear line between engineering design and manufacturing. That is not to say that the best engineers don’t know a thing or two about machining (and other manufacturing methods) and that the best CNC programmers don’t have to understand a good bit of engineering. And to some degree, there has been Design for Manufacturing (DfM) to fill the gap. However, the players have always been trained on different tools- the engineers in Computer Aided Design (CAD) and the machinists in Computer Aided Manufacturing (CAM), often separated not just by knowledge and training gaps, but by organizational structures (and typically even separate companies). However, due to the latest tools and impending advances in AI, I predict this status quo is about to change. That change will expand to envelope many design & manufacturing industries and transcend the traditional business silos. We begin to examine what is generative slicing.

Advancement is being driven in large part due to the accelerating developments in additive and hybrid manufacturing, but also by the advances in implementing special AI into the design process. Altogether, the various software steps in going from concept to physical part are referred to as the “toolchain”. Much of the toolchain for additive manufacturing was built upon the very same processes used in CNC. In diagram form, it looks something like this:

Toolchain
Toolchain

Design – Build Process:
Modern product design flow, from input requirements to output raw physical part. The middle three steps are the software toolchain.

Of course, we are already seeing the integration of iterative modeling and simulation software directly into CAD design packages as the result of Agile development efforts to speed the process and reduce cost. Modeling and simulation are software tools as well, and the trend seems to be moving toward integrating them with the CAD software. Most professional-level CAD systems nowadays include Finite Element Analysis (FEA) and other sim packages.
These design advances have occurred at the same time as advances in manufacturing capability where 3D printing (3DP) technology has been removing a lot of constraints that have historically limited machined parts. And this, despite the fact that most of what is currently called 3DP, is really only 2.5D (more on this later). Even so, engineering design has not yet caught up, due to 1) lagging preconceived notions about 3DP capabilities and cost, 2) technical debt of heritage designs, and 3) locked-in supply chain relationships.

The result of the recent advances in AM is that geometric complexity is not only now ever more feasible, it also no longer needs to be cost additive. In subtractive machining processes, complex geometry means more material to remove (equaling more machine time). In additive, it just means less material is used. This has opened up a world of possibility generative design methods seeking to apply AI to the iterative design-model-simulate cycle, rapidly evolving the part design.

What is generative design?

Generative design is a design exploration process where the designer designates performance goals, required features, keep outs, and other optimizable parameters (such as materials, manufacturing methods and cost constraints) into a machine learning algorithm. Built into a software program,this generative design process then explores the possible solution space, generating a collection of design alternatives. It tests and learns from simulation iterations based on what works and what does not. It still requires the designer to select from the solutions based on their engineering experience, additional modeling and simulation, aesthetics, and manufacturability. However, one can easily imagine where machine learning would eventually be adapted to optimize for those criteria as well. With the removal of many constraints by additive manufacturing methods, DfAM optimizations to the generative design algorithm become relatively low hanging fruit for programmers at CAD software companies. Already, for example, Autodesk Fusion 360 has a generative design module addressing some manufacturability optimization.

Generative Slicing Design
Example of [Raw] Generative Design Outcomes from Autodesk Fusion360

So far, however, only the simplest DfAM rules are included, such as overhang and bridging constraints, and some of the artifacts of 3D printing—primarily the anisotropic material properties in FDM type printers and the infill parameters—are left unaccounted for in the generative designs. These decisions, despite their very significant impact on the physical characteristics of the design/part, are currently left to the slicing step.

What is slicing?

Slicing is the 3D printing term for the CAM step. CAM software is a human-machine interaction that determines the tool pathways in which a machine (whether CNC, SLA, FDM, or other) can move to create the physical part bound by the STL mesh. Transferring from the model into printer movements is called slicing and is an art unto itself. It is called slicing because the simplest way to create a tool path is to first break the 3D part into a stack of 2D slices. The resulting instructions are typically transferred to the machine in GCODE format that captures the tool path as well as other parameters that depend on the material and method (like spindle speed in the case of CNC or extruder temperature in the case of FDM).

AM done in this simplistic way is really only 2.5D, not 3D. Every XY layer is fully two dimensional, but the Z only goes up in discrete steps and never back down. Consequently, most so-called 3D printers utilize only 2.5D. One of the artifacts of slicing in discrete layer thicknesses is the well-known look of stair-step layer lines visible in FDM prints, particularly (SLA also has layers but is typically done in thinner slices so is far less apparent). Also, you can’t put down printed material without something under it as support. This artifact results in bridging and overhang constraints in the z-direction. Nonplanar slicing will improve upon this and result in true 3DP for depositional manufacturing (FDM and DED) systems; various incarnations of it are just now becoming available in 2020. Powder bed systems are less affected by these constraints but have other constraints, and, as with SLA, NP slicing will not be an option.

The other notable artifact of depositional manufacturing is that the material properties, like tensile strength (especially in continuous fiber materials), are dependent on the deposition path. If the path lies only in XY, then Z has relatively poor tensile strength. Nonplanar slicing can address some of that as well, but not with the current tool chain.

Generative Slicing
Slicing / Deposition Artifacts

Still a hitch or two…
Why do I think that a couple of problems remain in the toolchain flow?

  1. The way things work now, the machinist/CAM programmer likely doesn’t know the full engineering intent of the design, where and how much the thermal and structural loads will be. But… they are the ones controlling the local material characteristic through tool path selection (i.e., slicing/CAM).
  2. Another significant variable in slicing is the infill selection and planar pattern. That is, how much of each layer to fill with material, and what pattern to use in toolpath planning for that layer. Infill percentage impacts weight and strength of the finished part, giving AM a big advantage over subtractive methods in light-weighting parts. Infill percentage could be a secondary design consideration specified in an accompanying drawing in the same way as surface finishes and dimensional tolerances are currently specified. But it requires the engineer to add design margin to account for variation in how slicers plan the tool path. The planar pattern selection not only affects the tensile strength as previously discussed, it also affects EMI/radiation shielding, thermal conductivity, thermal expansion characteristics, vibrational modes, and stiffness. Such things are the realm of metamaterial* design, beyond even the capabilities of most product design engineers, and certainly not to be left to the discretion of the CAM programmer who is far removed from the driving requirements.

An integrated solution


If metamaterial design is currently beyond most engineers, it seems the engineers merely need the right tool. Already simplistic metamaterial designs done in academic settings are done with custom software simulations. To properly implement that into the design-build toolchain however, the generative design/simulation software needs to be able optimize tool paths (i.e., slicing) within the limits of the machine. It must do this in order to capture the significant impacts of the various slicing/fabrication artifacts previously discussed. These artifacts can then be included in the simulations used in generating alternative outcomes. Enter the concept of Generative Slicing: the further consolidation of the toolchain, finally uniting CAD and CAM. Perhaps it will be called something else, but the concept will ulimately be similar, and likely arrived at incrementally.

A few repercussions

The generative slicing consolidation will drive some significant changes in the design-build industry. The timing of these changes will largely depend on when and how usable the generative slicing software is as it evolves.
For certain, the significant benefits of the cost, weight, and functional performance optimizations resultant of generative slicing will spur the continued adoption of AM and hybrid manufacturing. Machining will likely be the first industry impacted, but it’s likely that forming technologies will ultimately be impacted as well. No mistaking there will always be a need for them, but it’s an easy bet they will be seeing a reduction in market percentage compared to AM.
With the merging of design, prototyping, and manufacturing, we can also expect the design cycle to contract further. More on this in another article.
Perhaps less obvious is that generative slicing will motivate the designer and manufacturer to integrate at least enough that the engineer can design for the specific printer type. We are still very much in the Wild West of AM tech development (convergence/divergence of these technologies is another topic for another article). The consequence, however, is that it’s less likely we’ll see a standardization of toolpath specification, and more likely we can expect a lot more vertical integration between engineering and manufacturing. Ultimately, we believe a lot of engineering companies will either want to bring printers in-house, or outsource the engineering of those parts to a one-stop design/AM shop.

Moreover, once the design ultimately includes tool path instructions, there will be no real need for the intermediate STL files. What is less clear is how the roles of engineers and machinists will evolve to meet these changes. For now, I certainly expect both will be needed to advance the technology.

*Metamaterial: A metamaterial is a material engineered to have properties not found in naturally occurring materials. Such metamaterials are often associated with “smart” acoustic, RF, or optical characteristics resulting from the arrangement of repeating patterns on the scales of the wavelengths of the phenomena they influence (yet larger than the molecular structure of the constituent materials). Metamaterials derive their properties not just from the properties of the base materials, but from their designed structures. While so far research has been primarily in acoustic and RF applications, there are many more possibilities to explore with modern AM capabilities. Metamaterial design is a nascent, but promising field, which we’ll be discussing in much more detail in the future.

TERMINOLOGY
2D Two-Dimensional
3D Three-Dimensional
3DP Three-Dimensional Printing
AI Artificial Intelligence
AM Additive Manufacturing
CAD Computer Aided Design
CAM Computer Aided Manufacturing
CNC Computerized Numerical Control
DED Directed Energy Deposition
DfAM Design for Additive Manufacturing
DfM Design for Manufacturability
FDM Fused Deposition Modeling
GCODE Numerical Control language/format
GS Generative Slicing
NP Non-planar
SLA Stereolithographic Laser Apparatus
STL STereoLithograhy (mesh file format)
TO Topological Optimization
XY Horizontal plane parallel to base
Z Vertical direction normal to XY plane

No Comments

Add your comment