Skip to main content

Geometric non-linearity:fundmantal question-shallow truss problem

Submitted by kajalschopra on

Hi,

I've been looking at he chapter 2 of Crisfield's book on "Non-linear Finite Element Analysis".

Basically, my aim is to write a finite element program to carry out non-linear`analysis (geometric non-linearity) of structure comprised

of shallow truss elements.

The chapter 2 of Crisfield's book gives a set of sub-routines in Fortran as well as- to start with it derives the finite element equations

for a shallow truss element using the principle of virtual work.

I haave got some fundmantal doubts in the derivation of finite element equations which I state below.

I shall be grateful if someone good kindly help me regarding the same:

To start with-in deriving the finiteb element equations using the principle of virtual work we proceed as follows:

1) First and foremost, the shallow truss element is a 2 noded element with 2 degrees of freedom at each node.

2) We write the shape functions- which express the displacements inside the nodes to nodal displacements

3) The shallow truss element is an isoparametric element and the same shape functions which are used to express the displacements are also

used to express the geometry.

4) We get the strain epsolon

5) We get the increase in strain del_epsilon due to increase in displacements del_w and del_u (where 'w' and 'u' are verical and

horizontal displacements respectively)

6)We get the virtuial strain

7) We write the virtual work equation

8)We express the tangent stiffness matrix which is the ratio of small change in internal force to small change in  virtual dislacement

My questions are:

A)Please correct me if any of my steps are incorrect above



B)We get the tangent stiffness matrix as above- but what are we finding when we solve F = k u ('k' being tangent stiffness matrix) in each incremental step (say we are using incremental procedure) ---? Are we solving for the virtual displacements- we are actually solving for the real displacements- right? But the above steps pertain to a virtual displacement?



C)Though I underatsnd the above steps, I do not follow how the above steps (hence principle of virtual work) leads us to solving the displacements for each incremental step using incremental procedure for solving the geometric non-lienar problem.

Please help-i shall be obliged.

The weak form of Galerkin weighted residual method is identical to the principle of virtual displacements. Only the weighting function needs to be interpreted as the virtual displacement and everything falls in place.

Nonlinear finite element analysis actually involves several linear analyses successively and the results are algebraically accumulated. The principle of virtual displacement is actually a statement of equilibrium of foces (net virtual work done by a system of forces in equilibrium should be zero). So we actually seek incremental solution (in terms of real displacements) which will satisfy equilibrium (which will be the case when there are no residual (or, out of balance) forces). 

Hope this helps. Best wishes,

Tue, 05/26/2009 - 19:00 Permalink

Thank you very much for the reponse.

 

Yes, I realised that principle of virtual displacement is actually a statement of equilibrium of forces.

 

If the system is in equilibrium, it comes out as:

 

Total work done = 0

 

That is:

 

Internal (virtual) work + External (virtual) work = 0

 

That is:

 

 

Internal (virtual) work done = integral over volume of (virtual strain x real stresses) ------------- 1

External (virtual) work done = Real forces x virtual displacement--------------------------------  2

 

Summation of 1 and 2  = 0

 

When we do this summation:

 

We get an equation of form:

 

F = k u

 

(whether linear/ non linear)

 

My question is:

 

We are solving above for virtual displacements NOT real displacements? Right?

 

Please help-are we solving for real displacement above?

Tue, 05/26/2009 - 21:12 Permalink

Virtual work = Work done by real forces in moving through virtual displacements 

= (Real Force) * virtual displacement

Therefore, the equation of virtual work that you get is:

F * du = (K*u) * du (where, du is the virtual displacement and u is real displacement, and Ku is the real internal force and F is the real external force)

or, F = Ku

which you solve to obtain the real displacement increments.

Tue, 05/26/2009 - 21:56 Permalink

I remember solving problems in structural analysis, by unit load method- consider a truss problem wherein we need to determine a displacement at a joint in a truss.

 

Here, we apply a unit load on the node in the direction of the displacement to be determined (this is a non-real / unrelated system-virtual)-System I

 

Next, we have a real system consisting of real external loads- System II

 

considering 2 systems above:

 

virtual force (the 1 Kn force - in virtual system-system I) x real displacement at the joint- system II = Ext virtual work

 

Internal virtual work =  internal force in system II in each member due to 1KN load x real axial displacement in all members is system II

 

Solving, we get the real displacement at the joint.

 

This is also principle of virtual work? 

The principle of virtual work can otherwise be also stated as:

 

"The product of a zero force and a non-zero displacement is still zero - even if force and displacement are unrelated phenomenon"

 

This poduct of zero force and correponding non zero displacement cannot be real work as force and displacement are unrelated-hence the name virtual work.

Tue, 05/26/2009 - 22:02 Permalink

Hi,

Aren't you talking about Reciprocal theorem in the above comment, while earlier discussions were on virtual work principle?  I don't think they are related, since in principle of virtual work, we consider virtual displacements, without any regard to the forces, which caused them (virtual displacements should only satisfy the displacement boundary conditions), while in reciprocal theorem, we consider real displacements due to assumed forces.  In other words, the virtual displacements used in principle of virtual work are arbitrary (except for the acceptability condition mentioned), while those in reciprocal theorem are "real" displacements due to "virtual" forces. Please correct me, if I am wrong in making above statements.

Also, I have few related doubts:

1. The virtual work method was called as a special case of weighted residual method in the earlier  discussion.  Isn't it more proper to call it a sepcial case of variational principle?  Aren't we making potential energy stationary in this case?

2. Isn't it necessary that the virtual displacements be small?  If it is so, such a restriction should be due to a stationarity requirement, and can not appear in case of a weighted residual method.  Also the acceptability requirement on virtual displacements imply the same, in my view.

From book by Zienkiewics and Morgan (Finite Elements and Approximation), I understood that a wide variety of methods can be thought of as special cases of WRM, but I think it is only a mathematical possibility.

Thanks for any clarfications on the above items...

Jayadeep

Wed, 05/27/2009 - 05:09 Permalink