[With the introduction of in-CAD simulation tools, more and more design engineers are analyzing their own models. In this article, our FEA expert introduces the techniques the pros use to prepare models for analysis, so CAD designers can get more reliable results when they run simulations.]
When you develop a design, and build a CAD model, one of the objectives is to ensure that it provides a very tight definition of the design intent of the component or assembly.
A challenge with Finite Element Analysis (FEA), is that it now introduces a whole raft of new requirements which will affect the geometry. In this article I review some of these requirements, and hope to provide a basis for planning how to migrate the CAD model into a form ready for analysis
There are a range of requirements that FEA exerts over the geometry. They can be broken down into five main areas:
The actions that need to be taken on the CAD geometry are common across all five areas. These actions include:
Idealization is a process where we take the 3D definition of the component, and change it to a more abstract representation. This can include using 2D thin shell elements, to represent plate and shell type structures, and using 1D beam elements to represent beam-like sections of the structure.
The FEA model can also include the use of spider elements and lumped mass elements, to represent components that do not need to be modeled in detail. Springs and connectors can be used to simulate joints, and so on.
There are many motivations for doing this, including avoiding the cost and complexity associated with large numbers of 3D solid elements, improving the accuracy of the structural response, and simplifying the form of stress output so that it can be more easily interpreted.
At its simplest level, the process can include de-featuring CAD details which are not needed in the FE analysis. Holes and external radii are typical examples of this. The general rule is that features that are not influencing the strength, or stiffness, can be ignored
One of the most important objectives when creating an FEA mesh is to put small, regular shaped elements in regions where we anticipate high stresses. The accuracy of the stresses in the FEA method totally depends on the size and quality of the elements. Small elements throughout an FEA model, can have a very big penalty in terms of the solution time and the amount of computer memory required. So, there is a trade-off between cost and accuracy.
There are dedicated meshing controls in the FEA preprocessor tools, which will work on existing geometry. However, if we leave it at that, we are greatly limiting our opportunities to control the mesh. Imprinting surfaces, splitting bodies or other techniques can help guide the flow of the mesh in a very powerful way. It takes some experience working within the CAD environment to develop efficient techniques. Frankly, it is an art rather than a science. Then again much of CAD model creation is rather like that! I would really recommend this as a technique for designers to start to experiment with.
A good analogy is that of a wine glass. The geometric definition of the wineglass will give continuous axisymmetric surfaces throughout. However, imagine carrying out an FEA analysis, with wine in the glass and fingers supporting the bowl or the stem. How do we introduce the support regions where our fingertips touch? We will need to imprint our fingertips onto the glass. There may be three or four scenarios to consider – a delicately balanced support, through to a gorilla -like grip! We could end up with whole series of imprinted pads on the continuous surfaces.
All of this new geometry is superfluous to the original design intent. But it constitutes a very important part of what I call the “FEA friendly” CAD model.
Loading is a similar story. If we consider the bearing load applied to the hole in a lug, then it is conventionally applied over 180°, if in tension, and around 120° if in compression. There may also be a range of loading angles that we want to consider. Often, the hole has just one continuous inner surface. We need to imprint the boundaries of all loading distributions and can end up with multiple surfaces.
Anticipating the locations where we want to investigate results, and preparing geometry to improve the accuracy, is one of the most neglected areas in FEA. We need to evaluate the peak stress values at fillets, cutouts, and other important stress raisers. We also want to investigate the stress gradient around the peak value. Have we got enough elements to represent the gradient? In a typical mesh, solid tetrahedral elements can appear well distributed on the surface of the component. However, it is very difficult to control the quality of elements through depth. As a result, elements are often scattered in a very chaotic way through the depth.
It pays to manipulate the geometry so that we have faces and surfaces in these regions. Imagine a simple hole in a plate, which is loaded axially. The peak stress occurs at the 90° and 270° positions, relative to the loading direction. If we slice the geometry across the diameter here, it forces the elements to align with that cut face. This means that the elements are a better shape, with nodes and element faces embedded in the cut face of the geometry. The stress values do not have to be interpolated across elements spanning the critical plane. XY plots of stress versus distance will be far more accurate.
Again, it will take some practice to build up experience in how to anticipate these areas and prepare them by manipulating CAD geometry. But that’s a great way to transition into a real understanding of the capabilities of the FEA tool
All these techniques require additional work, beyond the initial creation of the CAD model. There is a natural tendency to think of these as being methods which are only of interest to the dedicated analyst. However, if the ambition is to get the most out of FEA, then I think CAD users are in a great position to adopt these approaches. Manipulating CAD geometry is second nature to them!
Want to learn more about how to make simulation work for you and your team? Download the infographic to learn the Top 5 Best Practices for empowering design engineers. And get started realizing the potential of simulation: fantastic products.
Tony has worked on FEA for over 40 years. He started his career in the UK aerospace and defense industry. His project work spanned dynamics, fatigue and fracture, nonlinear, and many other areas of FEA. He has since worked closely with clients and software development and helped lay the foundations for the user interface in several CAD embedded FEA projects.
Since 2007, Tony has run FETraining, which provides FEA consultancy, training, and mentoring. Contact Tony at firstname.lastname@example.org