Alastair Robertson
Femap Marketing Manager Siemens PLM Software Plano, Tex.

Part of the complexity comes from trying to accurately simulate the variety of joints and fasteners, such as spot welds, gluing or bonding, and bolts. Each behaves differently. Interactions range from simple contact to contact with friction. The demand for accuracy has been driving the need for fast and reliable ways to simulate interactions between components.

The first step to a useful simulation is model setup. This involves identifying all components that touch each other and accurately defining the contact conditions between parts. Automatic assembly detection is available in pre and postprocessor software such as Femap. The capability determines which components are in contact, and displays all contact instances either by contact-face to contact-face, or part-to-part. Connection behaviors can be set for either simple contact or treating them as if glued, and by referencing an appropriate property definition. Users can spot contact using a contact-detection feature and later in model setup by easily changing existing property references. In addition, users can manually define new contact segments. Contact conditions can be applied to models before meshing, or to finite-element faces of solid or shell elements.

To easily spot model-contact segments and regions, up-to-date software transparently highlights them. This lets users see how complex models fit together and their components interact.

Assembly simulations are getting closer to the real world by letting users apply linear contact, glued connections, spot welds, and bolted joints. Because each behaves differently, a few tips can help select the right conditions for accurate simulations.

Linear contact can still be accurate. A rigorous solution to surface-contact problems typically required a nonlinear approach to account for contact and other potentially nonlinear behaviors, such as large deformations and nonlinear material properties. However, if it is possible to stay within certain limitations (small deflections and linear-material behavior), then selecting Linear Contact provides a faster solution that doesn’t compromise accuracy.

The surface-to-surface, linear-contact capability in NX Nastran for example, runs a linear-static-analysis solution, and iteratively searches for element faces that may come into contact under the applied loading conditions. Users can also simulate finite sliding and apply friction when needed.

Glued connections simulate bonded or glued components. They join parts with differing or noncontiguous meshes, as long as the parts share a connecting surface. A Glued Connection refines the meshes of the joined surfaces so the load transfers across the surface as evenly as possible.

Results from this method show smoother contours than previous methods, particularly across the joining interface. This is typically where more-traditional approaches to noncontiguous mesh connection, such as with rigid-body elements or multipoint constraints, yield poor results. Glued connections also work well in models that require high accuracy. The capability eliminates the need to keep this kind of connection away from areas of interest and regions with high stress changes.

But which correctly represents the real-life situation: linear or glued contacts? That is an engineering decision. Selecting glued over linear contact changes model behavior, affects load paths, and hence results within the model.

To illustrate, a simple lug and pin have been modeled two ways: one uses linear contact and the other, a glued connection. The load transfers between pin and lug in the linearcontact example only at the bearing surface. A gap opens up on the opposite side of contact. The glued model, however, lets the pin pull and push against the lug giving a completely different result and stress distribution. The point is that the glued connection can be used to represent parts that are actually glued together, or the connection can be used as a tool to join differently meshed components where it is intended that they be a single part and without having to remesh to get a contiguous model.

In addition to different ways to model contact, there are also different ways to connect components using elements such as fasteners, spot welds, and bolted joints.

Spot welds and fasteners are typically modeled with a connector element that joins two surface patches, elements, or points. The connected meshes need not be coincident, which allows more freedom when assembling components. Typically the connector- element’s stiffness determines the weld diameter, length (if not a spot weld), and material property.

When components use bolted connections, the torque that tightens the bolts translates into axial-bolt preload. It is often useful to analyze bolted structures with preload already set. But you may also be interested in finding stresses due to preload alone.

Traditionally, bolt preloads were modeled in FEA with an equivalent thermal loading by letting the bolt contract at some low temperature. However, the method is approximate and solving models that contain multiple bolts require many iterations. More up-to-date software, such as NX Nastran, automates and simplifies this multi-solution process. Bolts are represented as beam elements with preloads applied at the ends.

The glued contact produces a max stress of 50,500 psi. The two stress plots show how different selections greatly change results. In general, most component connection instances require some kind of contact, and linear contact is an effective and efficient way of modeling it.