FEA Hand Calculation
FEA Hand Calculation
This makes the problem simple, and we can easily follow the
math in such a case! Let’s take a look!
The above shows you the basic relation between stress and
strain. I also explained how all of the parameters can be
established in the case of our rod (Note that Young’s
Modulus is a material constant).
Now, let’s get back for a second to our rod. We would love it
to be in equilibrium right? This means that the forces on
both ends should have equal values, but work in opposite
directions. In such a case…
You can notice that I’ve decided to make the F1 force
“negative”. One of them had to be (for them to work in the
opposing directions), and F1 just “draw the shorter straw”. If
I would do it the other way around it would still work at the
end : )
It’s easy to notice, that if you would know the forces (or
deformations) you could relatively easily solve the above.
It’s just a set of 2 “linked” equations. But if there would be
plenty of those (and there will be!) it’s much more
convenient to solve those in a matrix form. Well… maybe not
convenient for you, but definitely more convenient for your
computer!
For now, it’s only a matrix for a single element (rod in our
case), but if we would have more elements in our example
we would “bash them” together into one big “global”
stiffness matrix.
Keeping It Real
So far it’s simple I hope… normally I would test if the
previous equations “work” but the above is not a “real case”.
We just assumed that we will apply equal forces at both
ends of the element to keep it stable. Obviously, according
to our assumptions, this would work. Let’s try to do
something more “real”:
As you can see, this would be the simplest static task you
can have. There is a 2m long rod loaded on one end and
supported on another end. The difference here is, that it’s
actually a “solvable” task, so we can try to use FEA to solve
it, and gain some confidence that it actually works!
All I did below, is simply input the “real numbers” from the
example into the equation from the previous part. All the
logic still applies, nothing fancy happened. You just need to
remember that in a place where support is, deformations are
equal to zero (hence u1=0) but there is a force applied
(reaction force)… you just don’t know how big the reaction
force is (this is why F1 is written as… F1 – we don’t know
the value yet):
Facebook
LinkedIn
Important note!
You know… use the software when you try to solve real
stuff! This is just for demonstration/fun!
What is cool about each FEA model is, that the nodes are
connecting elements together. As you can see above,
element A and element B “share” the node (2). This means
that both elements A and B “add” something to the
“situation” in the node (2). When we will build a global
stiffness matrix it’s made from “blocks”. Since we already
know that matrixes of each element correspond to specific
DOF, we can denote them as:
On a side note:
It’s good to understand that corresponding DOF number is
not a node number. Here it’s like that, but in “real” FEA
problems it’s not the case! We are solving a very specyfic
problem where each node has only 1 DOF (along the
element). Normally beams have 6 or 7 DOF per node, which
means that first 6 or 7 DOF’s are still at node 1!
Now, we have u3 and u4… so we can solve the first row for
u2. At this stage I think it’s clear why I wanted to have zeros
in the bottom left corner of the matrix – such a state makes
solving really easy:
Yes! We actually did it. I must say that I was writing this
post step by step doing the math as I went. I was wondering
if this will play out right. After all, I’ve only made a few
Internal Forces
Great! Above we’ve managed to solve displacements and
reaction forces of the system. We already know a few useful
things. But obviously not all of them. Now, let’s wonder how
to calculate the internal forces in each element – those can
come in handy in design right?
The twist is… we don’t know yet how the elements deform! I
mean we know how the beginning and the end of each
element deforms, and we are using linear elements, so it’s
painfully obvious. But our solver doesn’t know how the
elements deform… and we need to do something about this
if we want to use it!
The above looks nice and dandy, but let’s do a quick test if it
is any good. Element “B” seems to be a good candidate. It
starts with node 2 (u2 = -0.355mm) and ends with node 3 (u3
= -0.71mm). It’s pretty obvious that in the middle of the
element the deformation should be -0.533mm, Well, let’s see
if this is also obvious to our solver:
A Cruel Twist!
I must admit that I was surprised that I pulled this off. I
never did FEA by hand, and it took me several hours to find
materials in books to make this work. But I did it, and I
definitely learned something new. I think that this post
contains more equations than all other posts I’ve written in
the last 3 years O.o’
But… this is still engineering for crying out loud! The fact
that you know how to move a matrix around does not mean
that you know how to design stuff! In fact, I would say that
those are two completely different things!
I know that at Uni they mostly (if not only) talk about this
stuff in a way I did in this post… but this doesn’t mean that
this is the only thing that exists. Far from it… it’s not even
the beginning. After all, I’ve been successfully using FEA in
design for a decade now… and I learned how to solve those
freaking equations a few hours ago…