Some of the connection problems discussed above result in mechanisms (structures that have moving parts), but there are other ways to create mechanisms in a VisualAnalysis model. Some of these are obvious once you "see" the problem. Others are more subtle mathematical instabilities that would not happen in a "real" structure, but cause problems in the matrix analysis.
PROBLEM: 3D Instability
When trying to analyze models in 3D it is common to get instability because of either (a) forgetting about the out-of-plane direction, (b) forgetting about rotations, or (c) having simply too many end-releases and/or not enough support. You can debug your model in this fashion:
- Start with just one or two load cases for speed
- Remove ALL the member end releases (shift+click a member), except perhaps simple beams.
- Don't ever put end-releases on columns.
- Consider the rotating columns problem/solution below.
- Is your model stable for analysis under both dead and lateral load? reasonable displacements?
- If so, add back end releases carefully, only add releases where you need them: where you get moments or forces you don't expect. (If you don't have weak axis bending, you don't need weak axis releases!) Return to step 5.
PROBLEM: Rotating Columns
When modeling a braced frame in 3D you will often have columns that are pinned at the base, supporting simple beams, and possibly bracing members. This is how the structure is built, but your model is a mechanism. Here is a picture of an isolated column in such a model, and a Model View snapshot showing the framing detail.
Typically you will model this with end releases on all the beam members because you do not want them to carry moments through the simple clip angle or shear tab connections.
When you try to analyze the model, VisualAnalysis tells you that node N7 (at the top of the column) is not supported in the RY direction.
This can be a confusing message! What does node N7 have to do with anything (Often there are members framing into this node as well, in all four directions!) It helps if you figure out that "RY" means rotation about the global Y-axis, as defined in the Help file, search for "Notation".
The answer is that the entire column is free to rotate about its own axis. The pin support at the base provides no rotational restraint, and if all the beams have weak-axis moment releases, then they provide no resistance to this rotation either. The error is in modeling the connection to the foundation, which will not rotate. Fix the column base with a nodal support in the RY direction.
PROBLEM: Spinning Nodes
Here is a similar problem, also involving incorrectly modeling the base support condition. We have a model of a swing set, which is a pure truss model, although modeled as a frame in VisualAnalysis to allow the continuous top chord to carry the swings. The main support members are pinned at both ends because the designer marked the connection type as "truss member" under the Modify tab of Project Manager. The base nodes are pin supported.
When you try to analyze the model, the software complains about one of the base nodes not being supported in the "RY" direction.
Although this is the same message that we received in the previous problem, you may try to solve the problem differently. Perhaps these aluminum posts are just sunk into the sand and there is nothing to prevent them from rotating at the base. You could fix the posts at the top because they are definitely not going to rotate there, but this creates a moment connection that you do not want!
The solution in this case is to remove the end releases at the base of the supports. These releases are separating the member from the node in a rotational sense, and because the node itself is not restricted, it can spin. We have a mathematical 'singularity' in the stiffness matrix that effectively prevents analysis. Once you remove the end releases at the base, the node is tied to the support member and is mathematically stable.
PROBLEM: Four-Node Truss Panels
Perhaps one of the most embarrassing problems you will face in structural modeling is when you create a situation that you know is unstable from your first structural analysis course in college. Everybody knows that a truss is made out of triangles and that a four-node truss panel is unstable. And yet, it is easy to create a 3D space truss like this:
The problem is that we often think about structural analysis in 2D rather than 3D!
If you analyze this 3D model, you will find that it works just fine! At least it appears so at first. For gravity loads the model is stable and you get reasonable results. However, for a laterally applied wind load (hitting the bridge sideways) you will find some tremendously large displacements!
The solution is to look at the structure from the end. Now you can see the four-node truss panel. The solution depends on the structure you intend to build. Most real structures are not pure trusses, but will carry moments. You may also wish to provide some knee braces or similar supports.
PROBLEM: One-Way Members are Removed
Another mechanism problem that you can run into in VisualAnalysis is not really a problem with your modeling, but in the non-linear analysis of tension-only or compression-only members or spring supports.
When you analyze a model with one-way elements, the analysis is iterated. Elements are checked after each analysis to determine whether they need to be removed or put back. This continues until the solution converges—that is, no elements need to be removed from or replaced in the model. Unfortunately, you do not always get convergence and you have to resort to manual solutions or modeling changes.
Sometimes however, you can end up with a different type of problem. Because elements are removed from the model automatically, and somewhat randomly, your model can become unstable. The following simple structure demonstrates the problem. The diagonal brace is defined as tension only, but the load will put it in compression so it will be removed during analysis.
When you try to analyze you will receive a message: "Analysis has stopped due to an instability in the model..."
The trick is to remember that you have these one-way elements. Mark all of your one-way elements to behave normally (perhaps in a copy of your project file) and see if the analysis will run in a linear fashion. If so, you are closer to solving it. The solution though, may be involved and is placed under the heading "The Art of Modeling"; in other words, IES Technical Support is not going to help you solve this one!