Earlier this year, I wrote about Altair Engineering’s OptiStruct and how it can assist in functionally driven design. It is not, however, the only tool that performs topology and shape optimization. Another CAE software application that falls into the same category is the Abaqus Topology Optimization Module from Dassault Systèmes’ SIMULIA brand. I know, it’s a mouthful. Luckily, people just refer to it as ATOM.
In this post, we’ll take a look at the background for ATOM, the capabilities it provides and then some of my commentary and analysis. From there, we’ll wrap up with a quick summary and a couple of questions to you. Ready? Let’s go.
Background
The Abaqus Topology Optimization Module (ATOM) was launched in the spring of 2011 as part of Abaqus 6.11. The kind of optimization that ATOM offers is complimentary and not overlapping with what is provided by iSight, also from Dassault Systèmes’ SIMULIA.
Capabilities Provided
OK. So what exactly does ATOM do? Like Altair’s OptiStruct, ATOM can suggest shapes for the design of a part or product as well as refine existing designs. At the most simplistic level, a user starts by defining the rough boundary of the design, adds constraints, loads and an objective and away it goes. ATOM comes back with a new shape that meets the criteria. In concept, the workflow in ATOM are similar like OptiStruct in this excerpt from my post on Altair’s OptiStruct.
Let’s say you are designing a part to carry a structural load. With OptiStruct, you essentially model the outer boundary that this part could occupy. From there, you add in your normal constraints and loads. After that, you define your objective and design constraints. From there, OptiStruct essentially iterates on removing material in various locations to meet the objective and stay within your design constraints. Essentially, by performing a structural simulation on the block, OptiStruct understands which material within the block is and is not carrying the load. From there, it’s not to hard to see how OptiStruct could remove the material that isn’t carrying the load and remove it.
At first glance, it may sound fairly unimpressive, but what is key is in defining the outer boundaries that the part could occupy. That starting shape could be as simple as a block or cylinder. But realistically, it could be far more complex. Perhaps you need a part to fit within a number of other assemblies. You could potentially use a CAD model to extract the unfilled volume that the part should occupy along with its attachment points. That becomes the starting point for which OptiStruct could start removing material and tell you what the conceptual shape for the part should be.
Inevitably, the question wandering around your head is: how is ATOM different from OptiStruct. Well, here’s some more details about ATOM.
- ATOM Offers Two Kinds of Optimization: Interestingly, these two optimization approaches are used in a complementary fashion.
- With topology optimization, the shape and volume of the design is changed to meet an objective. This is done by changing material stiffness.
- With shape optimization, the design is changed to address more localized issues. For example, peak stresses are smoothed by moving nodes of the FE model around.
- In both cases, one objective can be defined. Of course, many measures can be combined to create a single multi-measure objective.
- These optimization routines do not change the FE model through parameters. The changes are driven by modifying the mesh of the FE model.
- ATOM Covers a Range of Physics: The kinds of analyses that can be taken into account include:
- Single bodies, assemblies that include contact
- Advanced behaviors like large deformation, contact and non-linear material properties.
- ATOM Offers a Range of Constraints: No design is completely freeform. As such, the user needs to identify what can and cannot change. This includes:
- Geometry that cannot be changed, must maintain a type of symmetry (planar, cyclical, etc.) or must be a certain size.
- Designs that can be released from a mold or die taking draft angles, pull directions and overdrafts into consideration. Similar constraints and more can be applied to parts that are stamped or forged.
- ATOM can be ‘scriptable’ with Python: So even custom actions can be defined and taken into account during the optimization. Scripting also allows you to use ATOM within workflows built using Isight.
That gives you an idea of what ATOM can do. But how exactly is it done? Well, there’s actually a fairly detailed process or procedure that is involved. Here’s the sequence of events.
- To start, a user can utilize an existing FEA model or they opens or imports a geometric model. If the latter, the model could come from any CAD application. Clean up of the geometry for the analysis would occur prior to this step. Again, this doesn’t represent a finalized design but the boundary of the design that can be formed, shaped and molded.
- Next, the user creates the FE model by meshing the model as well as adding simulation constraints and loads.
- From there, the user sets up the optimization. This includes defining the objective that must be met and the variables that can change.
- Then, ATOM runs a topology optimization, which can result in large-scale and dramatic changes in the design geometry.
- Even though the topology optimization in ATOM essentially works by removing and adding elements as necessary, the resulting geometry may not be as prismatic as the user desires. So next, the user then needs to refine the topology optimization geometric results.
- As a next optional step, ATOM runs a shape optimization that tweaks geometry across the model to address localized issues such as peak stresses.
- As a final optional step, the user redefines the geometry as the output is point cloud data from the external element.
We now have an idea of what ATOM can do and how you go about doing it. What about the value you can get from it? Is it worth it? Let’s talk about that next.
Commentary and Analysis
Software applications like ATOM and OptiStruct offer a number of advantageous capabilities for an organization. For those of you that haven’t read my post on Altair’s OptiStruct, let me take a moment to recap those advantages here.
- The emergence of Direct Modeling has enabled more engineers, as opposed to CAD specialists, to use CAD for design activities. However, some designs are almost exclusively functionally driven. In those cases, a functionally driven design method, like using shape optimization, is a better fit than trying to creatively come up with something new.
- Furthermore, shape optimization technology is valuable for engineers in that they can flip it on, walk away and work on another design. Every engineer is pressed for time today. This technology essentially lets them be more productive.
- As good as a fit as it might seem, you wouldn’t use this technology on every single part you design. This is not a technology that is ‘casual user ready’ in the way that simpler more directional analyses are. You’d want to apply this to the really big, nasty and tough design problems where there would be a big payoff in terms of weight or cost.
In addition to those advantages and caveats, there’s another one that needs to be pointed out. Neither OptiStruct or ATOM includes CFD simulations in shape optimizations. Maybe that’s asking a bit too much, but if you’re going to invest a significant portion of time setting this thing up and then a significant amount of time in terms of computational cycles, you really do want to make sure that you cover all of the design’s behavioral characteristics. Maybe that will be a point to support in the future.
Interestingly enough, this area of discussion brings up an important advantage in using ATOM.
ATOM’s Support of Advanced Physics
As it turns out, the ATOM technology covers a broader range of physical phenomena than other similar applications do. Specifically, it handles more advanced physics like large deformation and non-linear material properties. So what? To my point before, if you’re going to invest a lot of time in optimizing a design, you want to ensure that you cover all the cases, not just most of them. Remember, this isn’t some quick and dirty directional type of analysis. The outcome of this kind of shape optimization should be fairly close to the final design.
Does the inclusion of advanced physical behaviors make a difference? Yes. Actually quite emphatically. The results between a shape optimization that includes advanced physics and one that does not can be dramatically different. That can spell the difference between passing or failing a round of prototype testing.
Geometric Output of ATOM’s Shape Optimization
Another point that must be discussed is the export format of the geometry that ATOM optimizes. It’s STL or INP, which is an Abaqus Input format. That’s not exactly the broadest range of formats that are being supported. This is the result of how the technology works. You see, the model that is being optimized here is the Finite Element model, not a 3D geometric CAD model. And yes, that makes a big difference. You see, ATOM removes volume from the model by setting the material stiffness of the element to zero. The elements are there, they just have no impact on the result of the simulation. It’s a great way to make the technology work and actually dramatically change the shape of a design. However, the only geometry in the model is composed of elements. There are no boundary representations. All that is left are the nodes of the elements, which resembles… that’s right… a cloud of points. And essentially, that is what is being exported here.
As much as it would be nice to have other export formats, it’s tough to conceptualize another way to get there from here. On the bring side, a wide variety of new software applications have emerged that work very well with point cloud data. This, for the most part, has come about because of the new technologies in the 3D scanning space, but they can actually be applied here.
Summary and Questions
I know. Lots of content. But with technologies this advanced, it takes a little explanation. Here’s the recap.
- The Abaqus Topology Optimization Module (ATOM) was launched in the spring of 2011 as part of Abaqus 6.11.
- The workflow in ATOM is similar to those of Altair’s OptiStruct for shape optimization. You define the outer boundary of the design, setup the optimization and ATOM removes material to suggest a new shape for your design.
- ATOM offers two optimization routines, one for shape optimization and one for design refinement and tweaking.
- ATOM optimizations are inclusive of simple and advanced FE simulations like large deformation and non-linear materials. It also offers a wide number of geometric and manufacturing related constraints.
- Shape optimization is particularly advantageous for engineers working on functionally driven design, where it can suggest shapes engineers might never have considered. Furthermore, it is an autonomous technology, allowing engineers to walk away and work on other designs.
- Shape optimization, however, does not apply to every part due to its time and computational intensive requirements. Additionally, it is not a ‘casual user’ technology. It is valuable for high impact parts where dramatic weight and cost savings can be achieved.
- A differentiator for ATOM is its support of large deformation and non-linear materials. This can dramatically affect the shapes that an optimization routine suggests. Given the investment this activity requires, covering all cases is important.
- The geometric output of ATOM’s optimization is a point cloud. It is not ideal, but given how the technology works, it is nearly unavoidable. Many new applications have emerged that work well with point cloud data, given the growth in the 3D scanning space. Those tools apply here.
OK. That’s my perspective. What’s yours? Sound off and let us know what you think of shape optimization and ATOM technology in particular.
Take care. Talk soon. And thanks for reading.