We spend a lot of time writing articles about how to use the very capable tools that are available from Ansys, Inc., but tend to skip over the project management side of simulation. But, project management can be as important, especially for challenging simulations, as the technical aspects. We recently completed a series of Non-Linear Dynamics simulations with ANSYS LS-DYNA and ended up learning a lesson or two about how to get such projects done on time and on budget.
What is Non-Linear Dynamics Simulation, and what makes it different?
When materials are deformed so fast that the rate of strain changes material properties, we refer to that as non-linear dynamics. In non-linear structural simulation, the material may be distorting in a non-linear way (usually plasticity), but the non-linear properties are dynamic. Because of this, time gets involved in the equation, as do non-linear material properties. Think car crash, metal forming, drop, bird impact on windshields and jet engine blades, and bullets going through stuff.
There are various software tools in the Ansys family that can do the non-linear dynamics, but our preferred program is Ansys LS-Dyna. It is an explicit dynamics solver that solves structural, fluid, thermal, and other physics. It is an amazing program that does many things. Still, for the class of problem we are talking about here, we only care about time-dependent material non-linearity for structural deformation.
Before beginning a project of any type, it is important to establish goals. Non-linear dynamics is no different. What is different is that you have to be realistic about what goals you can achieve. The events you are modeling are, by their very nature are very, well, non-linear. The answers you calculate can change drastically with mesh, loads, material properties, and solver parameters.
If you need high accuracy, then you need to set the expectation that it will take longer to solve, and you have to be more careful with your model. If you don’t need accuracy and you may be looking for relative improvements, like seeing if one geometry option makes things better or worse in your design, then you can back off and be less detailed. This difference can have a large impact on your schedule and overall cost.
So before you plan, before you start gathering information, decide what you expect to get out of your model.
Planning for the Job
Once you have set your expectations and goals, it’s time to map out the project. It is not that different from most structural or vibration jobs. You still have to get geometry, create a mesh, define loads and constraints, apply material models, run, and post processes.
However, each of those steps can be different for non-linear dynamics. Here are some critical issues to be aware of when producing a schedule:
If you are working with thin, especially sheet metal, parts, you probably want to use shells. They are more efficient and can be more accurate in many situations. You need to not just have a CAD model, but also a model that has the mid-plane surface defined as well as thicknesses.
You also want to look at removing tiny features that don’t impact the solution. The run time in an explicit dynamic solver is driven by the smallest element size. If you have tiny features relative to your overall geometry, capturing them can drive up your run times. So set aside time to remove or simplify them.
As mentioned above, small elements can drive up run time. Also, distorted elements or elements that become distorted can cause your solutions to diverge and fail. You may (probably) need to create a hexahedral (brick) mesh. All of these things require more time to create the mesh, and from a project standpoint, you need to plan for that.
Ansys LS-Dyan rocks at contact. It is pretty much automatic in most cases. So here, you don’t have to set aside time to define and tweak your contacts to get convergence. But there are many options, including erosion and other fancy options. Understand your contact needs and track and manage them.
Everything in LS-DYNA is time-dependent, and loads are no exception. If you are lucky, your load or loads are constant over time. But if not, you need to set aside time to characterize those loads and get them specified in the right format. In addition, loads can be calculated, say the results of an explosion or an airbag deployment. These use Equation of State models to calculate forces on the fly and are a major advantage of the tool.
From a project management standpoint, it is very important to plan for relatively long solves, restarts, and if possible, solving several jobs at the same time. Non-linear dynamics is computationally intense. Do some trade studies on computer resources vs. schedule time. Is it worth investing in more cores to solve faster or just let it chug away on a smaller computer? Also, don’t assume a single run to get the answer you want. Often you need to run the model multiple times before you understand what is really going on.
We are solving highly non-linear events, and understanding what the model is telling us is the whole point of the exercise. Budget time for processing massive amounts of data over time and reducing it into something useful. Also, time is needed to create animations. The analyst may also find themselves buried in the weeds at the post-processing stage, and project management should take on the role of reviewing the results from a big picture perspective and drive what tables, graphs, plots, and animations are created.
Keeping the project on the rails when things are literally blowing up and crashing
The dynamic nature of both the events being modeled and the process of creating and running the models make for a less predictable progression for the project. A project manager needs to pay close attention to what is going on at all times and pull the engineers doing the work back up for air to find out where things are going.
Here are some things to watch out for:
- Building a model that is more complex than needed
- Making sure that the situation being simulated in the model is what the customer needs simulated
- Too much time being spent to fit the model on a limited computer. Get a bigger computer.
- The simulation engineer is fixated on details that don’t impact the solution much
- Oversimplification of components, connections, and loads.
- Science project mode – spending time trying to learn basic information or trying to get something new to work, and not solving a specific problem.
One of, if not the most important roles for the project manager is communication. Constantly interacting with the engineers (without nagging) and with the customer (the person who will consume the results) is critical. This is not a throw-it-over-the-wall type of project. And the more you accomplish, often the more you have new questions. It may take two weeks or nine months. But either way, the PM needs to be constantly talking to everyone involved.
And yes, here at PADT, we have actually modeled a train car going off the rails. The project, though, stayed on track because we kept a close watch and stayed focus on the specifications. Things did surprise us, and we had to change some of the model when we got the first results, but we planned for that, communicated with the customer, and kept our changes to what was needed to answer the customer’s questions.
Our cars have incredible crash safety, very few planes fail because of bird ingestion, and we create amazing components out of formed sheet metal because of this type of non-linear dynamic simulation, and in most cases, Ansys LS-Dyna. Proper project management that recognizes the challenges and differences for this type of project can make a massive variety of products even better.