# Understanding Analysis

VisualAnalysis uses a Finite Element Analysis method (FEA). This numerical procedure is beyond the scope of this help file, but the following offer practical guidelines for applying the method. For information about the implementation (theory and assumptions) be sure to look at the references in the various analysis topics.

- Get Ready For Analysis
- Understand Analysis Limitations
- Analyze Large Projects
- Solve Lack of Support Errors
- Solve Mechanism Errors
- Solve Member Connection Errors
- Investigate Static Imbalance
- Investigate Large Displacements
- Investigate Yielding, Cracking, Crushing, Buckling
- Re-Analyze after Changes
- Remove Analysis Results

## Get Ready For Analysis

The following checklist outlines many common problems that you should check
to ensure a successful analysis. Many of these issues are addressed
automatically in
*VisualAnalysis* through the **Analyze | Check Model for Errors** menu
command. This command checks many of the discussed items below and provides the
results in a report. These issues need to be corrected before an analysis is
performed.

- Geometry is properly defined using members, plates, and springs. Your model does not contain a mechanism.
- Enough nodes are properly constrained using supports or spring supports to prevent the model from having rigid body displacement or rotation in any direction.
- Any Equation and Factored Load Combinations include at least one Service Load Case with a non-zero factor. Included cases must not all be empty.
- If you need to perform a 2
^{nd}order (P-Delta) analysis, your model must not contain one-way members or one-way spring supports. - If you intend to perform a mode-shape or response spectrum dynamic analysis, your model should not contain any one-way members or one-way spring supports.

## Understand Analysis Limitations

With the exception of P-Delta analysis results,
*VisualAnalysis* assumes that the deformation of the
structure is not large enough to severely affect
equilibrium. Most often it is large rotations of members and plates that can
bend that can cause results in the finite elements to deviate from their
real-world behavior. Currently, *VisualAnalysis* does not implement any
finite elements that can handle large rotations, therefore you should always
question their results if deflections appear large.

One indicator of problems will be that the
Statics Check tabulated on the **Results** tab
of the Project Manager (or in a report) will show
a difference between the Applied Loads and Reaction forces and moments. This is
a key indicator that equilibrium of the structure has not been met and that your
results are questionable. The reaction results are based on the deformed shape
of the structure whereas the Applied results are based on the undeformed shape.
When these deviate (which should never happen in reality) deflections are large
enough to generate false results.

When P-Delta results are shown, a process in
which stiffness nonlinearity resulting from large axial forces and stresses is
taken into account. *VisualAnalysis *utilizes a one iteration approach as
described below which has the effect of accounting for bending magnification of
structures resulting from lateral motions. This process does account for some of
the large rotation effects, but is not complete, therefore even P-Delta results
can be different from real-world behavior especially when large deflections are
predicted.

Stresses are assumed to be small so that the classical first derivative
definitions apply. In addition, rotations and displacements are also assumed to
be small. **
If any of these assumptions are violated for the actual structure response, the
use of VisualAnalysis as a prediction tool must be questioned.**

Presently, the only material model utilized by *VisualAnalysis* is an
isotropic model. Material properties are assumed the same in all directions. If
materials are orthotropic or anisotropic, *VisualAnalysis* may be unable to
accurately predict response.

## Analyze Large Projects

The only practical limit on model size is the amount of memory you have on your computer. The amount of memory available is of two forms: physical memory (RAM), and virtual memory (secondary storage, like a hard drive).

The largest size problem depends upon many things including:

- The speed of your machine (CPU, cache, hard drive, etc.)
- The structure type chosen (number of degrees of freedom at each node).
- The number of nodes and elements.
- The number of load cases and loads.
- Performance vs. Accuracy settings. See below.
- The number of other applications and processes running on your system.
- The number of windows you have open.
- The phase of the moon. (Just kidding, although solar flares can and do affect the accuracy of your CPU.)

In other words, it is nearly impossible to estimate the largest problem you can solve. Trial and error is the best means to get a handle on a practical limit.

In many situations you will not run out of memory, but rather just slow down to a crawl. In these cases, try a faster machine with more memory, or subdivide you model and only analyze, say for a multi-wing structure, a wing at a time . Seriously question whether you need all those load cases you've defined, or all the plates you subdivided. Do you really need an iterative analysis or frame instability when you know these effects are negligible? There will probably come a day where our word processors will demand entry level computers which surpass all structural analysis needs, but we are not quite there yet.

## Solve Lack of Support Errors

*VisualAnalysis* traps the common support problems, such as no support
in the Y-direction. The analysis will stop and a message indicating the
deficiency. In most cases this is just an oversight on your part and you will
simply define your supports and go on.

There are situations in 3D modeling where you think you have enough supports, and indeed you might have enough support to restrain against the direction of loads. For example if you have a rectangular mat footing, loaded vertically only, you might think that vertical supports are all you need. But if you have modeled this as a Space Frame then you also need to support against horizontal translation in two directions and rotations about all three axes. Even though your model may not be loaded in one of these directions, you still need to constrain it. In situations like this you should just pick arbitrary supports in a way that will constrain the problem mathematically without affecting your results.

## Solve Mechanism Errors

You might receive a message during an analysis
regarding a negative or zero diagonal in the stiffness matrix. This is a
technical way to say to you that your structure is unstable or nearly unstable.
Consider a center column line that is unsupported and all adjacent beams are pin
connected using member moment releases. The center column line can move up and
down unopposed. Running this model with just self-weight fails and *
VisualAnalysis* reports a "…zero on the diagonal of the stiffness matrix.

One procedure that we at IES have followed in our pursuits to solve mechanism errors is as follows. Begin in a section of structure that you believe to be the culprit and start applying nodal supports and taking out member releases until you get the analysis to proceed. Look at the resulting deflected shape and verify its correctness. Many times this alone will point out the violating members, plates, and nodes. Once you've got it analyzing, slowly take the supports back off and put releases back in until failure. Again this should give you some insight into the problem.

**Got Instability?** Try removing **all** member end releases and then slowly add them **only** where you really need them. Column members rarely should have end-releases. Also fix column support nodes against rotation about the column axis.

## Solve Member Connection Errors

A common mistake with members is to over-release them, leaving them partially
connected. *VisualAnalysis* does not allow the obvious things such as
releasing the torsional restraint at both member ends, which would allow the
member to spin about its own x-axis. However, you can easily over-release a
member or a "group" of members. You need to think like a builder: "Would it stand up if I
really built it?" You also need to think like a mathematician – things that will
not rotate in the real structure might in your "pure" model! You need to think in 3D too: The classic
example in VA is a column pinned at its base and able to rotate around its own
axis because all the connected beams are released.

## Investigate Static Imbalance (Statics Checks)

A Statics Check is made for each load case analyzed. You can see the check in a Report View, or view the check in the

tab of .The total applied load or moment in each coordinate direction is calculated. These real loads are applied at their load points on the deflected shape of the structure. These values are effected by large deflections. A comparison is made against the sum of all support reactions or moments. If the structure is in equilibrium, the values are equal and opposite in magnitude.

If there is an imbalance, *VisualAnalysis* will give a message to warn you about
the condition. The tolerance on the imbalance is arbitrarily set. If you get this warning message, you will need to **use your best
judgment** after inspecting your model and results carefully.

### What to Look For?

Just because there is a warning does NOT mean there is a problem. With a numerical solution there can be round-off errors. You should check to see if displacements and stress-levels in your model are "reasonable" and "expected". If the only imbalance is a "moment" error, keep in mind that moments are highly sensative to the location about which they are summed and may have less merit than a force imbalance.

A large force imbalance, either in percentage or in absolute value is a problem that cannot be ignored. Errors are usually caused by large displacements or rotations. These can be global or very localized problems.

### What to Do?

If you have performed a 1st order analysis, you might see if a 2nd order (P-Delta) analysis shows a lesser problem. If it gets worse or doesn't change you may need to stiffen your structure or reduce your loads.

## Investigate Large Displacements

It is possible to analyze without error or warning messages and yet have
results that are meaningless. *VisualAnalysis* will make a check and
provide a warning if it detects "large displacements". However, there is no
fixed definition of what is large.

*VisualAnalysis* tries to safeguard against this. However, in structures
where loads have been misapplied, or geometric and material properties have been
incorrectly entered, large displacements can arise. You may get large
displacements if you do not use reasonable preliminary sizes for members and
plates.

## Investigate Yielding, Cracking, Crushing, Buckling

Your analysis may "succeed" in *VisualAnalysis*, but your results may
not be realistic! The software may report (in ignorance) that the stress in your
steel member is well beyond the material's yield stress. The software may
provide some checks against abnormally large stresses that would cause yielding,
cracking or crushing of materials, but you should verify that your forces and
stresses are within reasonable limits. *VisualAnalysis *does not (indeed
cannot) check against code-specified allowable or ultimate forces, buckling
limits, bearing stresses, or other criteria defined in building codes.

This type of checking is mostly beyond the scope of *VisualAnalysis*.

## Re-Analyze after Changes

When you make changes to your model after an
analysis, certain results will become invalid. *VisualAnalysis* tries to be
intelligent when it "throws out" your results. For example, if you modify
loading information only the affected load case results are invalidated. On the
other hand, if you modify the structure geometry, all results are invalidated.
By default, you will be prompted to confirm the changes and remove the results,
but you can turn this warning message off using **Edit | Preferences**.

## Remove Analysis Results

Remove all analysis results manually with **Analyze | Toss All Results**.
This feature is useful if you know you have bad results and don't want anyone to
use them. Also, you may want to recover memory on your system.