What properties should I use for analysis elements?

Question

What properties should I use for analysis elements?

Answer

The answer depends on why you are using them and what you are trying to achieve.  Two common scenarios are discussed;

• For Wind Wall spanning.
• For ‘rigid’ elements for structural offsets/ connectors/ ‘battened’ members

General Recommendations

Whatever the reason for using analysis elements, some general principles to follow can be stated:

• Do not just accept the default property values as these are very small and entirely arbitrary.  We advise that engineers always make a rational choice of suitable values and enter these.

• Avoid using infinitesimally small or astronomically large values.  See below for more on the reason for this.  We recommend applying some form of ‘sanity-check’ to the values entered.

• If the members are not ‘real’ and you don’t want them to add any additional weight to the structure, then you can make elements weightless by creating your own material with zero density, see this article How do I define an element in the model with zero density?

Spanning for Wind Walls

If wind walls (usually representing some form of cladding/ glazing) do not span onto structural elements, what do they span onto?  Logically there must be something.  Glazing posts/ rails?  If so, we recommend it is sensible to use properties for elements representing these that are similar to those of the actual sections/material that will be used in practice.

Engineers may be concerned with ensuring that such members do not take any load away from the ‘real’ structural members.  How will this be achieved in practice?  Will it be?  These are worthwhile questions to consider.  If the answer is yes, then this can usually be achieved with appropriate member releases. It is not necessary to use either infinitesimally small or astronomically large property values.  For example:

• If you wish to ensure that a vertical member does not take significant axial load, then in general you can use the “Axial Load release” option for the top end of the member (note that this release can be applied to only one end of a member).  We do not recommend using infinitesimally small area values, either in an attempt to model tiny axial stiffness or ensure the element adds little weight to your structure.
  Image

  
• Note howe​ver, that this m​ethod cannot usually be used for flat slab structures.  If you are trying to achieve wind walls spanning to overhanging slab edges, or members framing to/supported by these, generally setting wind walls to "decompose to Rigid Diaphragms” - which does not require members to span to - is a better option.

• If you are using the Wind Wizard and decomposing wind walls to nodes, then a method you can adopt is to set the elements to be inactive (Properties “Active” checkbox set Off) as described in Inactive members.  Wind load decomposition will still ‘see’ them, but they will not be included in the analysis.  In this case it does not matter what property values are used, since the element is not included in the structure stiffness matrix.  So this is an exception to general principle 1. stated above.

   
  Image

  

• If you are decomposing loads to mem​bers, then it is always worth considering if this is really necessary.  The same general amount and distribution of load will be applied to the structure when decomposing to nodes. However, if it really is necessary, then any load decomposed to inactive analysis elements will be applied in the analysis (as reaction loads at their ends).

‘Rigid’ offsets or connectors

There are a number of scenarios where you may wish to join members with very stiff, generally short, ‘fictitious’ elements that you would like to idealize as ‘rigid’ or infinitely stiff.  For the purposes of this discussion, let us call these stiff links.

You might use these when modelling structural offsets - e.g. for an eccentric connection - or where two or more closely parallel members are connected together in some manner and acting as a  combined or ‘battened’ whole - e.g. a compound section column as, for example, discussed in the Tekla Structural Designer related forum post titled "Bi-Directional Portal Frames". Please sign in to the Tekla Discussion Forum for more information:

Following general principle 2. we caution against using astronomically large values in an attempt to make stiff links (close to) infinitely stiff because this can cause an analysis issue called ill-conditioning.  We discuss this in the article Why are there limits for 1D element length and 2D element quality? This also applies to the use of infinitesimally small values in order to model almost no stiffness (releases generally are a better solution) - it is the range of values in the stiffness matrix that can produce ill-conditioning.

Infinite stiffness is not necessary - it does not exist in nature as a rule.  There is an analytical form of such infinite stiffness termed a ‘rigid offset’, which is used in Tekla Structural Designer for concrete member automated structural offsets. See this topic for more information.

Rigid offsets are truly infinitely stiff, but don’t cause ill-conditioning because they are an explicit part of the 1D element stiffness matrix.  When comparison is made between results for models using stiff links with appropriate properties vs rigid offsets, it is generally found that there is no significant difference. There is nothing ‘magical’ about being able to specify infinitely stiff elements – or portions of elements – it is in general a modelling convenience.

Recommendations for modelling stiff links are:

• Following general principle 3, create a ‘rigid’ material with zero density. See also How do I define an element in the model with zero density? We recommend using similar properties to those of steel which are probably sufficiently large for most purposes we can conceive of. You can then control element stiffnesses further through the section properties.

• Set the Shear areas to zero. These are the "A parallel to minor/ major" values. This removes the shear component of displacement completely and causes no analysis issues.

   
  Image

  

   

  • Any reader with a question such as "What are shear area and stiffness?" should check out this Timoshenko beam theory Wikipedia article.

• Property values for section area A and inertias Ix (Torsional Constant), I major and I minor cannot be zero.  Thought should be given to what forces will be generated in the elements.  This will guide you towards what values need special attention.   Where large stiffness is required, values should be reasonably large, but not astronomically so (or ill-conditioning may result - see above).  It is useful to perform a ‘sanity check' by transforming your values into a square section size (or working to values from this) - e.g.
  In metric: an inertia of 10⁹ cm⁴ equates to a solid square section size of 3.31m! 
  In US customary: an inertia of 10⁹ in⁴ equates to a solid square section size of 331 in (27.6 ft)! 
  
  That would probably be adequately stiff for most circumstances.

  • You can always conduct a sensitivity analysis if you have doubts e.g. for inertia start with say 10⁶ cm⁴ or 10⁵ in⁴ (a solid square section size of ~590mm or 33 in), run a few successive analyses increasing or decreasing the property values by, say, a factor of 10 each time and compare results for convergence.