MechanicsOfCompliantMaterials1 7

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 49

MECHANICS OF COMPLIANT MATERIALS

From Toothpaste to Titanium

Arun R. Srinivasa
Department of Mechanical Engineering
Texas A & M University

August 2021

2
Introduction
This collection of notes is for my class on the mechanics of compliant materials. Too often, this kind
of class is taught (if at all) AFTER going through continuum mechanics and perhaps elasticity. By
the time we get to the modeling of interesting materials, students are exhausted and have lost their
initial interest (and in many cases, they are now only PhD students left). On the other hand, UG
courses on strength of materials (a misnomer if ever there was one) is restricted to the linear response
of steel, concrete and perhaps, wood or other composite beams. With no mention of viscoelastic
response, plasticity etc., in any serious way. The aim of this set of notes is to introduce students
to the wide range of compliant materials and how to model them. Rather than show them these
models in a general 3-D continuum context or as UMATs in ABAQUS/ANSYS, I will show them
how compliant trusses can be modeled. This leverages students pre-existing knowledge of trusses
and frames and covers 90% of the applications. We will not shy away from nonlinear response using
the built in capabilities of MATLAB/PYTHON to do ”quick and dirty” solutions to ODES, linear
algebra and optimization.
I was inspired by the Youtuber 3Blue1Brown to rethink how linear algebra is taught and the
superb book on information theory by Mckay. My aim was to focus on “meaning” and insight and
not necessarily on proofs and the howtos. I have internalized, I hope, the core lessons from these
and other masters of explanation : (1) concrete is better than abstract, (2) start with context and
not definition, and (3) give full details and don’t skip steps. This set of notes, however inadequate,
is my tribute to these and other great teachers from whom I have had the privilege of learning.
The notes are meant to be a live book with links to videos on YouTube and to web pages. I
will add to them periodically either to expand the scope, or find more and better visualizations
and to add more exercises and python scripts/programs and to link to worked solutions. My aim is
to provide a “hybrid” learning environment where there are written materials, video explanations,
exercises and programming all in a single place that can adapt to changing technology. Hence, rather
than embed videos or programs, I provide links so that if I find a better explanation, I don’t have
to redo everything.

MODELING PHILOSOPHY–History of concept of knowledge


If you are interested in how the present approach to modeling came about and the current debate
going on between the Physics based modeling people with AI (data-sicnece approaches). Plase be
assuered that this has been going on from the beginning of humankind
Please watch these two videos by me on how modeling was thought about since ancient times
History of Knowledge-1–a non-experts view
COomen approaches leadning to scientific method

How to learn from these notes: YOU CANNOT BUILD MUS-


CLES BY WATCHING AN EXERCISE VIDEO

This is a 5-step process for self learning :


1. Download and print the material, and use it as a workbook.

2. Read it, highlight it, and make notes in the margin.

1
3. Watch the videos WITH YOUR NOTEBOOK SO THAT YOU CAN TAKE NOTES FROM
THE VIDEO. Pause the videos frequently, and take NOTES. This is how you will get the
maximum out of the videos. Simply watching from one end to the other is not useful.
4. Then, practice the Feynman technique: Pretend to teach a concept you want to learn about
to a student in the sixth grade. Identify gaps in your explanation. Go back to the source
material to understand it better. Organize, and simplify.
5. Do the exercises at the end USING A PENCIL. The idea is to get 100% right;
and not to just try it and move on. DO NOT MOVE ON UNTIL YOU GET
EVERYTHING RIGHT! THIS IS THE MOST IMPORTANT ITEM.

Useful tip: Don’t memorize formulae in symbolic forms. Math formulae are like text messages—
code for something in English. You don’t talk in text messages, do you? So, don’t memorize the
text message; rewrite in English, and then memorize.
For example, the centripetal component of the acceleration vector: Don’t remember it as v 2 /r—
you will forget what v and r mean; So, you won’t know how to use it: Remember it as “square of
the speed divided by the radius of the turn”. Now, you will remember it.

2
1 Introduction: Math supplies the logic, you supply the
meaning

The key to all of science and engineering and the like is the following:
1. We observe or notice something concrete. We want to make predictions or decisions
about some real life situation. Predictions require recognition of some kind of a pattern: we
notice that if a heavier object smashes against a lighter object, then, usually, the heavier object
“wins” i.e., the two objects seem to go in the same direction as the heavier object, most of the
time.
2. We hypothesize that there might be a pattern: Here is where the difference between
you as a student and as a researcher comes in. As a student, you are sometimes too eager to
match a pattern (Pattern Based Learning). So, you might generalize that if one problem has
been solved in a certain way, then all problems will have the same steps. This is too literal.
In reality, patterns are more subtle. You cannot directly use the pattern; You have to check it
out first.
3. We abstract or IDEALIZE it: We are in the interface world: We cannot directly
reason with the real world. We have to simplify it, leaving aside the unimportant aspects.
This requires judgement as to what is unimportant. We may circle back, and re-examine
the items that we thought were unimportant. This world is the world of drawings (including
things, like Solidworks) and graphs etc. Even here while we are not in the real world anymore,
we know what the drawings mean. This is an important point. Notice that up to now we have
not talked about any math.

4. Now we reason about it by stripping the meaning but leaving the logic: The problem
is that languages etc., have cultural context, and ”yes” in one culture may mean “maybe”.
So, we want to reason in a firm way that everyone follows. Enter the Math world: We
systematically convert all our images into symbols and strip them of all meaning., leaving only
the logical operations behind.
Let me repeat that. MATH IS ONLY ABOUT LOGIC—IT DOES NOT CARE WHAT IT
MEANS. Don’t look at math for MEANING; it only deals with LOGIC. YOU have to supply
the meaning.
5. From the results of math, we interpret it by putting back the meaning: Remember:
Math supplies the LOGIC, you supply the MEANING. That is why you have to “understand”
the results. If you cannot write what something is in English (not in equations), you did not
understand it.

1.1 Video Activity:

Now, watch this on the role of math in mechanics and this one by Richard Feynman on how
mathematicians and physicists are different. Use the space below for notes from the videos.

3
1.2 EXERCISE: General ideas about abstraction:
1. What are the 3 worlds through which you have to move in engineering?

2. What is the key reason for working in the symbolic world?

3. What kind of knowledge is needed for Design, and Diagnostics?

4
2 Introduction to compliant materials

By reading and doing the exercises in this section, you should be able to

1. Describe what a compliant material means,


2. Describe the potential uses for compliant materials/structures,
3. Describe the advantages and challenges in using compliant materials, and
4. Describe the meaning of terms such as “stiff, hard, tough, strong, malleable, ductile” that are
used to describe different responses.

2.1 Q: What is the difference between material,structure, mechanism?


All objects 1 have two prominent characteristics: at any given time they have some sort of shape or
geometry and they respond to external influences by changing this geometry (which is usually called
“mechanical behavior” or “response”). The response of the object is a combination of properties
that depend upon its shape and external stimulus and those that don’t.
To help us gain insight into how these objects respond, we separate those aspects of the behavior
that depend upon macroscopic geometry and those that don’t. The shape dependent properties
are called “structural properties” and those that are independent of shape are called “material
properties”.
For example, the stiffness of a spring depends upon its geometry (like, how many coils, what is
the diameter, pitch etc., and cross sectional shape of the wire) and also upon properties that are
independent of its shape (like, Shear Modulus).
Typically, we cannot envisage any “material” divorced from its shape. In reality, a “material” is
the name given to a collection or set of properties that are independent of the shape of the object.
For example, when we say “steel-material” (as opposed to steel-bars, steel-plates or steel-ingots),
we refer to all the common, shape independent properties of steel bars, plates , ingots, etc. All of
them are made of iron atoms with specific concentrations of other materials; they have a particular
crystal structure, Young’s modulus etc.
On the other hand, the name STRUCTURE is usually reserved for a very specific kind of body,
one whose purpose is to transmit forces and moments while retaining its shape. Structures are
primarily used to position objects precisely relative to each other. Typically, structures are either
in the form of a TRUSS or a FRAME.
A MECHANISM is a body that moves in a specific controlled manner under the influence of the
external stimuli. Typically, a mechanism has a finite (as few as one) degrees of freedom and is used
to transmit or convert motion from one kind into another.

2.2 Q: What is meant by a compliant body?


All objects, no matter how rigid, will deform a little when subject to forces. A compliant body is
one that changes its shape significantly (this is where the observer comes in) when external loads
are applied to it. Whether a body is compliant or not, depends upon a combination of its current
geometry as well as geometry independent properties.
1 What do we mean by “object”? At the highest and most abstract level, it is a finite region of space that we

somehow perceive to be different than the rest of space

5
For example, rubber is compliant no matter which shape you make it; but while a steel block is
not compliant, a steel spring is.

2.3 Q: what are the potential uses of compliant bodies?


2.4 Video Activity:

Watch this video on Compliant Mechanisms and this video by Disney on how they are using
this for their animatronic creatures.

As can be seen from these examples, the potential for compliant mechanisms is that we can make
them in ONE GO with NO Assembly and no moving parts. This is called Design for NO Assembly.
This is particularly the case for micro or nano-machines. Note that for any object, the surface
area is proportional to the square of its size while the volume is proportional to the cube. So if you
shrink something, surface area will dominate and friction( and stickiness due to Vander Waals forces
between atoms) will increase and the mechanism will simply “seize-up” and become unusable. For
example, two s in contact will actually weld themselves together. Hence, it is better to avoid sliding
and use compliant mechanisms at the microscale.
Also, you must have seen that all industrial robots are separated from human beings and are in
cages. This is because they are incredibly stiff and if they hit you, you can die. But the emerging
area of “cobotics”, the idea is to make robots “compliant” so that they will not injure you. Also,
gripping tasks require delicate operations–think about how we are able to grip paper, eggs, screws
and make delicate operations with them. An elephant’s trunk can pick up a tree or a single peanut
with precision. This is not a capability that robots have currently.
When you consider weight savings: if you want transportable objects (think emergency housing
or mars habitats etc.), then you want things that will occupy very little space, weight very little and
can be deployed (like inflatable raft) to full size when needed.
Finally, environmental considerations also play a major role. If we are able to harness natural
materials (especially polymers) to make products as efficiently as we use artificial materials, we can
enjoy a sustainable and high quality of life.

2.5 Q: what are the advantages and challenges of using Compliant ma-
terials?
Advantages: (a) No need for assembly-easy to recycle (no parts), (b) easy to manufacture, (c) more
efficient at a microscale, (d) lightweight, (e) ...
Disadvantages: (a) loss of precision of movement, (b) large loads are difficult to sustain, (c)
complex to design—nonlinear mechanics is needed, (d) unexpected DOFs/buckling may occur, (e)
Only some materials and structures are suitable, (f) .... (Add your own items here)

3 Material Behavior: the foundations of modeling

By reading and doing the exercises in this section, you should be able to

1. List response characteristics of solid-like materials, fluid-like materials visco-elastic materials


and plastic materials,

6
2. Describe what solid-like and fluid-like behavior mean,
3. Describe qualitatively the kinds of experiments run to determining response functions,
4. Describe the three elements of modeling, and

5. Describe the meaning of a constitutive response function.

3.1 Q: What material characteristics are utilized for building compliant


mechanisms and structures(CMS)?
The key to building CMS is to match the response characteristics with needs. For example, if you
want a compliant system with just one degree of freedom like a gripper, then you might want to
keep most of the mechanism being rigid or stiff and have compliant joints (usually flexible). On
the other hand, if you want something that will conform to the shape of a terrain, then you want
something that is very soft–almost fluid-like so that it can take the shape of its surroundings but
contain in it a stretchy material like a balloon so that it can recover its shape.

3.2 Q: What is the difference between a solid and a fluid?


At an intuitive level, a solid is a body that has a definite size and shape whereas a fluid will take the
shape of its container. A more formal approach is that solid is a body that can sustain shear stress
(i.e., it will deform but not “flow” ) whereas a fluid will flow no matter what level of shear stress is
applied.

3.3 Q: Why do we focus on shear stress and not normal stress?


When we consider bodies deforming, we know that there are two major “modes”: It can change its
volume (which is called dilation or compression), or it can change its shape without changing volume
(this is called ”distortion”). Most solids and fluids can withstand even very high compressive loads,
so it cannot be used as a reliable means for distinction. But solids and fluids behave very differently
under loads that cause distortions–shear stresses. That is why we focus on these.

3.4 Q: How can we apply shear forces to objects in real life?


Any kind of action that changes the shape of a body without changing its size can be considered as
some sort of shearing action. The typical way to apply shear forces is by
• “twisting” or torsion,
• stirring (which is called azimuthal shear),
• or by sliding between a pair of plates (called simple shear),

• pulling the center of a tube like a telescopic antenna (called Telescopic shear),
• squeezing in one direction while allowing it to flow in another (called pure shear).
Can you think of other ways to change the shape without changing the size of objects?

7
Figure 1: Dilation versus distortion of a cube: When we dilate something, we increase its volume but EVERY angle
is fixed. In the figure, not only the corner angles but also every angle in any figure drawn on it is preserved. On the
other hand, when we distort it, we preserve volumes of all drawings but we change at least one angle somewhere. In
the figure, all corner angles are still 900 but the angles of the red figure on the side of the cube are different.

8
Figure 2: Common ways in which you can distort a body. We use one or more of these to create compliant systems.
For example, torsion bars used in car suspensions is a compliant system that utilizes torsion! . A typical shock
absorber uses telescopic shear in its fluid bath. A door closer uses azimuthal shear. etc.

9
3.5 Q: Does a body have to retain its shape or recover fully for it to be
considered a “solid”?
No, we perceive something as a solid if you apply a distorting load, and it stops moving in the time
span of interest and within the spatial resolution we have. Even the ground that we live in may
move slowly over millions of years so that to a plate tectonics expert, it would be a fluid. Also, if
we have a very high resolution microscope, we may see that what appears to be not moving may
“creep” slowly. So, there is a characteristic length resolution and time scale to all our considerations.

3.6 Q: What is the operational difference between an elastic solid and


a viscous fluid?
As a starting point to considering time scales, we will define the two extremes:
• If a body deforms to its final form instantly when force is applied AND recovers fully and
instantly when it is removed, then it is an elastic solid.

• If a body flows instantly upon application of shear forces and stops flowing instants when the
load is removed (so that it does not recover at all) it is called a viscous fluid.

3.7 Q: Are these the only possibilities?


NO, there are a vast number of intermediate responses between these two extremes:
• If a body flows for a while and then stops deforming when a shear force is applied and recovers
fully after some time when the force is removed, it is called a VISCO-ELASTIC solid.
• If a body continues to flow endlessly when a force is applied and takes some time to stop flowing
(not instantaneous stoppage) when the force is removed , it is called a VISCO-ELASTIC fluid.
• If a body deforms to its final form instantly when force is applied and recovers instantly but
only partially so that shearing may cause a persistent instantaneous shape change then the
body is called a PLASTIC solid.
• If a body flows for a while and then stops deforming when a shear force is applied and recovers
partially after some time when the force is removed, it is called a VISCO-PLASTIC solid.

3.8 EXERCISE: Intuition for materials


1. Can you see why something is designated as solids and why others are designated as fluids in
the above list?
2. Why is there no “viscoplastic fluid”?
3. Can you explain why the squeezing deformation( item below simple shear in the figure) is
called “pure shear”? Where is the shear?

4. Can you identify ways in which you can use these concepts to test a material and classify it
as one of the items above?

10
4 Material Behavior: Divide and Conquer

By reading and doing the exercises in this section, you should be able to

• Describe the purpose of studying the physics of compliant materials,


• Why the compliant mechanisms are harder to simulate compared to their stiff counterparts,
• List the three principal features that are responsible for the response of a compliant mechanism,
• Name the geometric elements that we combine to make actual bodies as well as their degrees
of freedom, and
• Identify the difference between the point based elemental shapes and their ”thick” counterparts.

4.1 Q: What is the main goal for studying compliant materials and
systems?
Our main goal is to design, fabricate, and use compliant mechanisms and structures for practical
purposes.
• For purposes of DESIGN: We should know what is the actual need and how it maps to a specific
kind of material/structural response (i.e., do I need a stiff or compliant elastic response?; do I
need a visco-elastic solid or fluid or combination?; do I need a plastic material or structure?).
How to define the specific kind of material and geometry? – Should I use rubber or Polyethylene
or steel?; What should be the shape of the mechanism?
• For purposes of FABRICATION: What kind of fabrication process should I use–3d printing,
laser cutting, machining, etc.). What will its effect be on the final response?
• For purposes of USE: How exactly will it behave?; Will it meet my expectations?; How durable
will it be (will it lose function or break under repeated use)?; How do I dispose it off after
life?; How easy is to re-engineer, recycle or reuse?

4.2 Q: These seem to be the same questions that we need to answer for
ordinary design also. How are compliant systems different?
For ordinary systems such a rigid link robots, the material and geometrical considerations are not
linked. You can draw something in SolidWorks and test its function before choosing the material.
All the motion comes from the rotation or sliding at the joints. We can “buy parts and then assemble
them”. This is quite different from compliant mechanisms. Here, the geometry and material are
interlinked – the functionality comes from the combination of material response and geometry, not
separately. How you make it will influence how it works, which will influence how you design it
which will further influence how you make it...
You get this circular dependence, right?

11
4.3 Video Activity:

To understand this difference between conventional mechanisms and structures and these
compliant mechanisms, watch my TEDx Talk: “Can Objects be Smart?”, it starts out a bit
slow so be patient.

Designing compliant systems is a HOLISTIC process, you have to model and simulate the entire
thing and then figure out whether it will work. Thus, unlike conventional systems you have to
understand the simultaneous interplay between material response and geometrical shape change to
predict the behavior of these materials.

4.4 Q: How do we find a suitable mechanism or compliant system for


a given application?
For a conventional rigid bar+joint mechanism, it is possible to start from the motion we want and
obtain a mechanism that will provide this motion. It is much harder to do this with compliant
systems. The closest we can come to it is that (a) we start with a rigid-link mechanism and then
replace the joints with flexible (i.e., ”bendy”) joints. Even this creates a lot of problems. A body
with distributed compliance is yet another level of complexity and here, there is still no simple way
to go from what we need to a mechanism that will do it.

4.5 Video Activity:

To get an idea of how these are done, See this video on designing mechanisms from Disney
The challenges of extending to flexible joints is described in this video

4.6 Q: WHY is it so hard to go from rigid link mechanisms to flexible


mechanisms?
A key advantage of standard rigid links and jointed mechanisms is that they are ”kinematically
determinate”. This is the counterpart of statically determinate systems and means that, at least
in principle, the way it moves is independent of the forces on it. That means that from a design
perspective, we can just focus on motion and find a mechanism that will work. This will determine
its geometry. We then focus on forces and make it strong or stiff enough to withstand the forces on
it
This is what makes rigid link robots so precise. But it also makes it completely unadaptable and
dangerous.
Compliant mechanisms by nature will have some “give”. So, you cannot design it simply based
on movement alone. How it moves will be influenced by the forces on it. This means that it is
sensitive to the environment (which is exactly what we need). But this makes it incredibly hard to
design—we have to consider all kinds of external forces and find ways to compensate. This is the
challenge. So, modeling is very complicated.

12
4.7 Q: How do we go about modeling these flexible mechanisms?
Now is the time for the story of the engineer, the mathematician and the coconut tree: A mathe-
matician and an engineer are stranded in a deserted island. There is only a single coconut tree to
feed them. When they get hungry, the mathematician tells the engineer: why don’t you show me
how to get a coconut. The engineer climbs up the tree, grabs the nearest coconut and comes down.
They enjoy it. After a while, they get hungry again. This time, it is the mathematician who climbs
up the tree. But the nearest coconut is gone. So they grab the next one, tie it to the spot of the
nearest one, and climb down. The engineer is flabbergasted and says, “Hey what are you doing, why
didn’t you get the coconut?”. the mathematician calmly replies:
“I have reduced it to a problem that we know how to solve.”.
Applying this to the compliant mechanism problem, we observe that the overall response of
the mechanisms comes from a combination of three different features: (a) the geometry of the
mechanism, (b) the material response, and (c) the influence of the external world in the form of
loads and constraints. Our plan is to carry out a divide and conquer approach systematically:
1. Geometry simplification: We divide the given geometry into chunks and keep doing it until we
get a chunk (or elemental shape) of a fixed geometry (line segment, triangle, tetrahedron). The
chosen shape must be such that any other shape can be viewed as just a combination of this
basic elemental shape connected together. The “elemental” shapes or elements allow us to not
worry about the overall geometry of the body. The geometry “emerges” from the connectivity
of the shapes.

2. We then look at the material properties and do the same thing: We divide the response into a
combination of elemental responses so that the overall response ”emerges” from the interaction
of these elementary responses.
3. Finally, we look at the external world and drastically simplify it—we consider only extremely
local influences and treat it either as applied contact loads or constraints, or we consider it as
action at a distance(non-contact loads) like gravity. So, the only external influences will be
due to something touching it or gravity or electromagnetic effects.
By considering the body to be made up of such “elemental” pieces and responses, we can drasti-
cally simplify the analysis and actually get somewhere. It is exactly like constructing a novel out of
words2 (elemental pieces) which are strung together to form sentences which then form paragraphs
and then whole books.
As with books, it is necessary but not sufficient to have a good vocabulary. You must know
which words to choose, how to make elegant and precise sentences, and how to convey the overall
message in your book. This is what modeling is about. We learn the words and learn to make
meaningful sentences. In your undergraduate courses, we had a very limited vocabulary and our
sentences sounded like ”See, Spot. Run” . We are about to greatly expand our vocabulary and form
complex sentences in this course. That is where we are headed in the remaining chapters.
2 The words themselves are formed of alphabets but unlike words, alphabets have no “meaning”. The words are
like our material while the alphabets are like electrons, quarks etc., – too elemental for our purpose.

13
4.8 Q: Wait a minute, I didn’t quite understand how we get these ”el-
emental shapes and models”. This is new to me although I have a
vague idea. Can you show me by means of some examples?
The idea of using a “combination of elements” for describing bodies is as old as science, and even
predates it. Remember “the fifth element?” It is not some sci-fi stuff, that is how we actually do
things3 , but we slip it to you so insidiously that you don’t know it at all.
For example, we are taught about particles, rigid bodies, cables, beams as if they are actual
bodies, not idealized elements from which bodies are made. There is no such thing as a ”particle” or
a ”pont mass” but it is a convenient element from which other elements can be built; same with an
idealized cable or an axially loaded member (or strut or link). For example, if we want to analyze a
bridge such as the one below, we can idealize it as a truss which is made up of axially loaded links.
The link is a simple straight bar with simple geometry, and it can be connected together to make a
most complex truss with very complicated shape. The shape of the bridge “emerges” from how the
simple bars are connected together.
In the same way, we break up a frame into individual beams that can bend and twist. By welding
them together, we can make extremely complex shapes.
If you had done FEA using SolidWorks or ANSYS or ABAQUS or whatever, it will dis-aggregate a
body into a finite number of triangular or tetrahedral pieces. The actual shape of the bod ”emerges”
from the connection between the pieces. That is why it is called a FINITE ELEMENT method.
The major advantage is that if we know how to model something with a simple known shape,
then all other shapes can be then be modeled by assembly of this basic shape.

4.9 Q: By basic shapes, do you mean like in geometry - point, line,


plane etc?
Not quite. There are two major differences:
• Physical objects have properties beyond what geometrical objects do – they also have physical
properties so there are more options than what is present in simple geometry. (On a side note,
most of geometry was invented to describe physical objects.)

• Unlike geometrical shapes like, lines and planes which are of infinite extent, we will always
have finite sized bodies.

4.10 Q: What are these basic shapes and how do we use them?
There is quite a list. I have divided them into two groups depending upon starting point (literally
:-)) )
1. The first and by far the most common element is the particle or, more precisely, point-mass.
Notice the latter name is more descriptive and shows both the geometrical attribute (location)
and the physical attribute(mass). This is essentially the points of geometry endowed with an
additional physical property “mass”.
We can create a sort of a “body” by using a collection of a finite number of points with some
kind of interaction forces. If we do that and then use F = ma for all the particles, we will get
what is called “Molecular Dynamics” or Simply MD. If we also include drag and more complex
3 although we will have none of the exciting adventures that Bruce Willis did

14
forces, then we will get Smooth Particle Hydrodynamics (SPD). This alone will be able to give
you all kinds of beautiful CGI animations. Just watch this one on sloshing water in a tank
and this opensource SPD to get an idea: (it is mesmerizing to watch) All of these awesome

Figure 3: The building block of a lot of models is a point mass this has 3 DOFS. You can simulate pretty complex
things by putting together a lot of particles as can be seen from a sloshing tank simulation using SPD. See sloshing
water in a tank to get an idea of how realistic it looks.

looking videos are just F = ma applied to millions of particles. The complex shape changes
are called “emergent behavior” since they are not explicitly built into the forces but emerge
due to the complex interactions. Notice that in these simulations, the particles are separated
from each other—there are gaps; We can cut the body into two without touching the particles.
If we take the limit and consider an INFINITE number of particles infinitesimally close to each
other (more precisely an uncountable number of particles that cannot be separated, i.e., no
gaps) we will get a ’continuum. You were introduced to this in your fluid mechanics class.
We, of course, are not going to learn much about continua in this class, but the next class in
this sequence is continuum mechanics.
2. We will move up the ladder and introduce the physics equivalent of a line segment. This is
called a “link” or a “truss element” . In some cases, when it is subjected to only compression,
it will be called a “strut”. We will use ”truss element” to describe this. Geometrically, we
can draw it by specifying the location of two points. So, a truss element is visualized like a
dumbbell with point masses at both ends.
When you connect a finite number of these links into a 2-D or a 3-D network, you will get a
truss. If you want to see an example of its use, here is one Tensgrity structure Design which
shows how to convert a body into a truss etc You are most familiar with this from your MEEN
225 (statics dynamics, strength of materials) classes where we discussed trusses. However, in
MEEN 225, we generally treated the link as though the distance is fixed. When you went to
your strength of materials class, you revisited this and they probably introduced you to axially
loaded members. In other words, SOM introduced you to an elastic truss element.

15
Figure 4: A truss element is the physical equivalent of a line element.It is describe by 2 point masses and when it is
connected together you can get some pretty complex shapes

We will vastly generalize this to consider other types of trusses. This is extremely useful because
it allows us to illustrate every aspect of behavior of compliant mechanisms in a familiar format.
Most of this course will be devoted to such truss elements.
3. We will now move up the ladder further and introduce the physics equivalent of a “2-D plane
segment” which is the 2-D counterpart of a “line segment”. This is called a TRIANGULAR
MEMBRANE ELEMENT (or TRIANGULAR ELEMENT OR CST (Constant Strain Tri-
angle). (i.e., triangular region). This, and its 3-D counterpart, called tetrahedron, are the
workhorses of all simulations. You can make extremely complex shapes by just stacking them
together. TO describe the geometry of the triangle, you need THREE position vectors (one
for each corner).

Figure 5: A triangular membrane element is the workhorse of all 2-D simulations. It has 6-DOF in 2-D and 9 DOFS
in 3-D (the position vectors of the corners). You can create extremely complex shapes by simply assembling these
triangles. You can decompose any body into a set of triangles with any degree of closeness that you want.

16
4. By now, you should have recognized the pattern—next up is the tetrahedral element (the
workhorse of 3-D FEA). It requires FOUR position vectors to describe its geometry. BY
stacking them together, you can pretty much make any shape you want in 3D . See this
amazing video on the simulation of a beating heart by using 17000 tetrahedral to get an idea
of how powerful this divide and conquer approach really is.

Figure 6: Tetrahedral Elements are made of 4 points. they have a total of 12 degrees of freedom and can be assembled
into staggeringly complex shapes such as the heart shown here.

4.11 Q: Is that it? – particle, link, triangular element, tetrahedron?


No, we have just scratched the surface. There is another group of models that begins with sort of
..er...“thick” particles. The problem is you want to simulate rigid bodies. But until Euler (and then
Hamilton and others) came along, it was very difficult to efficiently define what the heck a rigid
body was. What do we mean by “efficiently”? Well, a conventional definition of a rigid body is
that it is a collection of particles wherein the distance between any two particles is FIXED. This is a
lovely definition but utterly impractical. Think of a game developer who wants to simulate a brick
as a rigid body— How many particles should you use? Where should you put them? how do impose
rigidity constraint? Before Euler, people struggled with the notion of rigid bodies and they used
something called the “principle of rigidification” (don’t ask...).
But when we did statics and dynamics, we didn’t actually use this definition! Thanks to Euler,
we found out that if we start from the other end( not with bodies but with particles), we get a better
picture. Euler’s idea (not quite his; but for the purpose of this narrative, I’ve attributed it to Euler)
was that
A particle or point mass has 3 degrees of freedom, (3 translations)+ 1 physical quantity
called mass (or more precisely translational inertia) which shows up in the laws of motion.
Similarly, a RIGID body has 6 degrees of freedom, (3 translations and 3 rotations). It
requires 6 numbers: 3 for position vector to show where it is and 3 ”Euler angles” to show
its orientation or “pose”. Physically, it has translational inertia (Mass) and rotational
Inertia (Moment of Inertia). How we get these values is immaterial for now. This is a
minimalist description of a rigid body.
With this approach, an entire Fighter plane can be represented by just 6 numbers (3 to locate its
center of mass and 3 angles “Roll”, “Pitch” and “Yaw”)! Check out this website and you can drag
an aircraft around and see how the angles change.

17
Figure 7: A rigid body is represented by a point with 3 arrows sticking out of it to represent its orientation. It has 6
DOFS, the location of the point (3) and 3 angles (such as Roll Pitch Yaw). It is an incredibly compact representation
since an entire airplane can be represented by just these 6 DOFS if we treat it as rigid

We should imagine it like a coordinate frame— a point with 3 unit vectors sticking out of it
representing its orientation. This is how it is represented in all game engines. We can think of it as
a sort of “thick” point( The actual generic name for particles with additional geometric attributes
is “cosserat point”, but almost no one knows it. My advisor did a lot of work on this. So, I am
familiar).

4.12 Q: Okay, fine! I know what is a rigid body now. I can see how
the intuitive notion of position and orientation are quantified, but
are there more things to it? Does anyone use them? Or, is it just
some weird carnival curiosity?
Not only are there other kinds of useful shapes connected to rigid bodies, but you have also used
them widely for all kinds of things already. We just didn’t tell you how to think about them; We
just showed you how to use them. Objects made out of a rigid body are called ”BEAMS” (yes, that
one!). Connected collection of beams is called a ”FRAME”. Then, there are are plates and even
3-D versions.

4.13 Q: How is a beam related to a rigid body?


In the same way a link is made up of two points, a beam-element or more generally a ROD element
is made up of two “rigid bodies”, position and orientations of the two ends. (That is why when you
do beam problems you have to specify displacement and SLOPE (“pose”) and the two ends of the
beam.
So a beam or rod element is quantified by specifying the position and orientation of its two ends
(total 12 numbers (2 position vectors (total six) and two sets of Euler angles for orientation. Watch
this youtube video Discrete elastic rods to see a beautiful animation of this. The animations of
plectonemes (look them up) are really interesting.

18
Figure 8: A beam or rod element is a flexible link that is described by the location and orientation of the two end
points. It has 12 DOFS (6 dofs for each end). Even if it starts out straight, the loads will cause it to curve (bending)
and twist(torsion). If you connect beams together, you will get a FRAME. You have seen this in your earlier mechanics
classes, but you didn’t analyze this. If you took my MEEN 468(and actually did the computation for the project and
not just delegate it:-)) ), you would have carried out a frame analysis.

By the way, EULER was the first to describe these, and he called the ”thick links as ”ELAS-
TICA”. If you google it, you will learn a lot about this. It took us just about 300 years to figure
out how to simulate the equations for rods described by Euler.
These ideas are used for everything from DNA coiling (yes, DNA is modeled as an elastic rod1)
to Hair simulation in movies, like AVATAR and other game engines.
If we connect beams together, you will get a FRAME (this is the building block of almost all
civil structures and most mechanisms and robotic simulations).

4.14 Q: Are there more items?


Yes, you bet there are. Just like bars or beams are “thick links” , we can consider plates as “thick
memberanes” and they can sustain bending.
Just like the Triangular MEMBRANE element, in the next up from links, we have TRIANGU-
LAR PLATE or SHELL elements (THREE orientable points at the three corners which are used for
modeling everything from aircraft bodies to egg shells. While a lot is known about them for elastic
plates, not much is known about how to model viscoelastic plates etc. This is current research. Then
we have what are called Tetrahedral cosserat bodies: these have 4 orientable points at four corners.
These are used for modeling discrete grains of sand and other granular media with microstructure.
I am vague about it because you have now reached the frontiers of what is known and this is a
research topic.
We have now exhausted all the commonly known geometry types.4
4 We also have quadrilateral and cuboidal elements but they are not the basic building blocks. They have additional

DOFS. If you take a FEA course, you will learn about them also.

19
Figure 9: Beam elements or rigid body elements are routinely used in everything from robotics (to represent tool
orientations) to Flexible Mechanisms and even hair in Game Engines.

Figure 10: A triangular plate element has 18 DOFS( six at each corner). It is widely used in structural mechanics
and aerospace for simulating shells. It is also widely used in the sheet metal forming industry to predict what shapes
you will get with sheet forming.

5 Material Behavior: Model Development-Variables, Dual-


ity, and Lumped Parameter Models

By reading and doing the exercises in this section, you should be able to
• Describe a mathematical model,
• Describe what duality means in physical models, and identify dual pairs in link geometries,
and
• Describe what lumped parameters mean.

20
5.1 Q: What do you mean by a “model” for a system?
You are most possibly used to the notion of a model as in a “model airplane” i.e., a miniature
representation of the actual object. Similarly, you might have been exposed to SOLIDWORKS
models, which are virtual geometrical representations of real(physical) objects.
We are now going to define yet another type of model— A mathematical Model.

A mathematical model is a symbolic representation of the behavior of a physical system. It contains


variables that describe the state of the system and mathematical relationships between the current
(and perhaps, even past) states.

5.2 Q: What do you mean by state? I have been hearing about this
from my chemistry classes all the way through thermodynamics;
but I only have a vague idea. So, what is the definition?
What we mean by a state is any quantifiable property of the system that we are interested in.
Notice that the state is a subjective matter: it depends upon what we, the observers, are inter-
ested in. This is why, you need to know the purpose of your model before you use it. This is what
defines the possible states. Without purpose, we won’t know what will interest us. Also, the same
system may have different state variables based on what we are interested in.
The way we have described “state” is broader than how the thermodynamics community defines it.
There are additional restrictions used by them, including the idea of internal state. Our use is closer
to how the the controls system folks use the notion of the state of a system. We will refine it further
as we get deeper into modeling.
Next, also note that it should be quantifiable. So, for example, “color” is not a state; “wavelength
of light reflected by the system” could be a state variable. So, we need to have a way of quantifying
it. It is this step that allows us to develop a mathematical model.

5.3 Q: Can you give me an example of what you mean?


Consider the list of elements that we did in chapter 04. Presumably, as a first step, we are interested
in the geometry of the elements. So, we need variables that will represent the geometry. Here is a
table of the elements and their “geometrical state variables”.

21
Geometrical State Variables for the elements
Element Geometric state de- State Variables
Name scription
Point-mass Location < x(t), y(t), z(t) >

Link End-point locations < x1 (t), y1 (t), z1 (t) >, < x2 (t), y2 (t), z2 (t) >

CST element Corner locations < xi (t), yi (t), zi (t) >, i = 1, 2, 3

Tetrahedron Corner locations < xi (t), yi (t), zi (t) >, i = 1, 2, 3, 4

Rigid Body CM location and < x(t), y(t), z(t) >, (θ, φ, ψ)
orientation

Beam/Rod End-point locations < xi (t), yi (t), zi (t) >, (θi , φi , ψi ), i = 1, 2


and orientations

Plate/Shell Corner locations < xi (t), yi (t), zi (t) >, (θi , φi , ψi ), i = 1, 2, 3


and orientations

5.4 Q: How come you didn’t include velocity and acceleration as state
variables in the above list?
I should, and sometimes, it is done but it is easy to see that ”technically” they are included in the
idea of position etc., as a function of time. So, velocity and acceleration can be “derived” from them.

5.5 Q: Why did you put angled brackets for position and curved brackets
for the others?
We reserve angled brackets for vectors and curved brackets for state variables (like rotation angles)
that form a group but not a vector. Just to be clear, the state of a system can be described by any
combination of scalars, vectors, matrices etc, as long as it can be quantified. They are collectively
called the degrees of freedom of the system (DOFS) and could include things like temperature etc.

5.6 Q: Is there only one way to describe the state of a system? I thought
that the length of the link or truss element was a state variable.
The same geometrical system can be explained in many different ways. For example, you could have
used polar coordinates to describe the geometrical state of a particle or some other coordinates. For
a link or a truss element , we can use location of mid point and the length and 2 orientation angles
of the element to describe it. But the number of state variables i.e., the number of DOFs will be the
same no matter how you describe it.

22
5.7 Q: Why did you not include forces, torques etc., in the state vari-
ables?
Remember the definition of state? If we are only interested in the geometry of the element (as we
do in SOLIDWORKS), then only geometrical variables are of interest. If we are interested in the
physics of motion i.e., mechanical state of the system, then I have to include forces, torques etc. If
we intend to consider thermal effects, then I will have to include temperature, entropy, etc., in the
description.

5.8 Q: So, what are the mechanical state variables of the elements that
we considered so far?
In addition to the geometrical state variables that we have listed in Table 5.3, we have to list the
following.
Geometrical State Variables for the elements
Element Load-State De- State variables
Name scription
Point-mass Force < F x(t), F y(t), F z(t) >

Link End-point forces < F xi (t), F yi (t), F zi (t) >, i = 1, 2

CST Ele- Corner forces < F xi (t), F yi (t), F zi (t) >, i = 1, 2, 3


ment

Tetrahedron Corner forces < F xi (t), F yi (t), F zi (t) >, i = 1, 2, 3, 4

Rigid Body CM forces and Mo- < F x(t), F y(t), F z(t) >, < M x(t), M y(t), M z(t) >
ments

Beam/Rod End-point forces < F xi (t), F yi (t), F zi (t) >, <


and Moments M xi (t), M yi (t), M zi (t) >, i = 1, 2

Plate/Shell Corner forces and < F xi (t), F yi (t), F zi (t) >, <
Moments M xi (t), M yi (t), M zi (t) >, i = 1, 2, 3

5.9 Q: Why is the list very similar to the list of geometrical variables?
Is this an accident?
No, it is not accidental that (a) the list of load states has the same form as the list of geometrical
states, and (b) there seems to be some correspondence between the geometrical state variables and
force state variables. This is actually a very deep aspect of physics and is called duality. We will
describe it in more detail shortly.

23
5.10 Q: But, I noticed that while orientation variables were NOT vec-
tors, the moments < M xi , M yi , M zi > are vectors. Why is there a
difference?
Good catch! The reason is again something deep about physics modeling. The loads cause the
geometry to change so the duality that I referred to is the relationship between the forces and the
change in geometry! For position, which is given by a vector, the change in position is given by
velocity which is also a vector. Orientation variables are not vectors, but changes in orientation are
given by the angular velocity VECTOR! So, the idea of duality is related to changes, either with
respect to time (which is called “rate” ) or with respect to space (which is called “gradient”)!

5.11 Q: Whoa, whoa, this requires some getting used to, and I am not
completely sure that I understand. So, can you tell me what du-
ality is again?
I am glad you asked. Here is what duality means:

For most mathematical models of physical systems, the variables that describe the state of the
system always come in pairs, these pair of variables are such that their product will have the physical
dimensions of energy or more precisely, the product of one variable with the increment or change in
the other variable in the pair will have the dimensions of energy rate or energy gradient or power
This is an abstract definition. To root it in more common terms, consider electrical circuits. The
variables of interest are voltage and current, and their product is power( Jules/sec). So, voltage and
current form a conjugate or duality pair. In the same way, if we consider forces and positions, we
can see that force times displacement per unit time is power. For pipe flow problems, we see that
the variables are flow rate and pressure, and their product is power yet again!
Similarly, for spinning systems and beams etc., the variables are moment and change in angle,
and their product is energy per unit time (if the change in angle is per unit time = angular velocity),
or energy per unit length if the change in angle is measured per unit length (i.e., it is the curvature).
In every case, if we identify one variable, you should also look for its conjugate variable by asking
which variable will give me energy dimensions.

5.12 Q: That’s interesting! Are there any guidelines for how to look for
conjugate variables?
Typically, in terms of our experience, we think about duality in terms of what is the “cause” and
what is the “effect”. So, you can start with that and designate the variable you have as either
“cause” or “effect”.
Ask yourself if we think of the variable that have as the “cause” (or “effect”), then you need
to find out the “effect” (or “cause”). For example, if you have a bar and you have identified the
temperature of the ends and want to identify its conjugate, you can start by asking “what is the
effect of the temperature at the two ends being at different values?” You might arrive at “heat will
flow”. So, we have now identified the conjugate variable–heat flux.
Try your hand at finding conjugate pairs with some other variables.
Another alternative is to think in terms of “control variables” and “response variables”. For
example, you can ask “If I control the temperature at the ends of the bar, what will be the response
of the bar, what will happen in the bar?” This is the control systems approach. When you do tension

24
test, we do a similar thing — we say , “we will control the displacement and measure the response
(force).”
The flow analogy: Finally, there is one more way to think about it based on the electrical circuit
or pipe flow analogy: Identify the variable that you have as a physical quantity that either “flows
through the system” (like current or mass or heat)—the “through variable” or that we measure
”across the system” (like voltage)—-the “across variable”.
If what you have is the through variable, then its conjugate is the across variable (and vice versa).
Of the three strategies, the last one is the most “physical”. My recommendation is to use
one or the other of these strategies to identify the variables and use the other two to check your
understanding. This will be part of your next homework.
Also there is no simple mapping between “cause”/“effect” and “through”/“across” variables.
Sometimes the across variable is the cause and the through variable is the effect (Voltage difference
is the cause and current is the effect).
In other cases, the roles will be reversed. For example, force is the cause and the change in
displacement is the effect.
By the way, what is cause and what is effect is sort of subjective. Only conjugacy or duality matters,
whether we call something cause or effect is part of the meaning of the variable, the underlying
mathematics doesn’t care.
So, to not confuse these things, we will use “control variable” and “response variable” as our
generic names for the duality pairs.
Mathematically also, this duality plays a very important role. There is a similar sort of “duality”
as observed above between row and column vectors in linear algebra; their matrix multiplication
will always give you a scalar.
(In the language of tensors, they are called “contra-variant” and “co-variant vectors” —-don’t
ask why the names.)
So, this idea of duality is central to our whole approach to modeling.

5.13 Q: How does duality help in modeling systems?


It tells you what to observe. For example, for truss elements, it will tell you that you need to
observe forces and not something else. From an experimental point of view, you can figure out what
to control and what to observe.
We will develop relationships between the control and its corresponding response variables.

5.14 Q: Can you show us an example of how this model works?


At last, after all this description and discussion, we are ready to look at some actual responses. We
will focus only on truss elements. As a quick recap, the geometrical state variables are the position
vectors r 1 , r 2 at the end points.
Duality will tell us that the conjugate variables are the forces f 1 , f 2 at the ends. The position
vectors are the across variables and the forces are the through variables. To make matters a bit
simpler, we can use the displacements u1 , u2 as the geometrical state variables if we keep the
starting or reference state fixed.
Now the key is to find a way to relate f i to ui . If we think of carrying out a test to do it, we
have to find two vectors f 1 and f 2 given the displacement vectors at the end. This is a hard task,
but we have a trick up our sleeve that will make it considerably easier.
We are going to make the truss element into a lumped mass system

25
Figure 11: A truss element AB with end A at r 1 and end B at r 2 will move to ab. The displacements of the two ends
are u1 and u2 ), The dual variables are the forces f 1 and f 2 .

5.15 Q: What is a lumped mass system. Why do we care about it?


The problem with physical bodies is that wherever there is material, there is also mass. For example,
if I want to consider the elastic response of a rubber truss element, I have to contend with the fact
that its mass is distributed along its entire length. This will make the applications of Newton’s Laws
very messy. Among other things, it will now have rotational moment of inertia and will make my
life miserable.
So, we will take a drastic step: : divide and conquer or if you remember your 401 separate
your functions!. This will be our motto throughout our modeling.
When the going gets tough, find a way to separate different aspects and then assemble them to get
what you want. Divide and conquer; separate and assemble. We have already done this for the
geometry, we will now do this for the response, too.
In our truss element model, we want to separate the inertial aspects (that has to obey Newton’s
Laws) from the rests of the response. So we will lump the mass and move it to the two ends which
are called the NODES. Then the region between the two ends is massless and so we don’t have to
worry about the inertial effects and we can do everything like statics. This is a huge simplification
and, although there is a price to be paid in terms of loss of fidelity, the inaccuracy is so small that
the bang for the buck is fantastic.
As an added bonus, when you put these truss elements together, all the mass will end up at
the nodes, and when you have to apply Newtons laws, you simply do method of joints and write
equations of motion for each joint (like you did in your 225 class)–how simple is that!
In almost all the models that we consider, we will always lump the mass as shown.

26
Figure 12: A real truss element will have a distributed mass as shown on the left. We are replacing this with an
idealized body where the masses are lumped at the two ends and the middle portion is massless. So, we can use
concepts of statics for the middle portion! When the elements are assembled as shown in the “bridge truss”, we can
see that all the mass is lumped at the joints. This way, we can still use method of joints in a pretty straightforward
way.

5.16 Q: Now that you have lumped the mass, what’s next?
Well, the forces f 1 and f 2 are now applied to the region between the ends; and since it is massless,
we can use equilibrium for a two force member (channelling your engineering mechanics class) and
conclude from linear momentum balance that

f 1 = −f 2 (5.1)

In other words, the forces between the bars have to be equal and opposite! By balance of angular
momentum, we can also conclude that the forces have to be only axial; there cannot be any transverse
forces! Wow, just from lumping the system, we gained A LOT! It is worth repeating:
For truss elements, if we lump the mass at the nodes, it becomes a two force member. Then, we can
conclude that (a) the nodal forces are equal and opposite, and (b) they are along the line joining
the two ends in other words (see figure 13). The forces must be of the form

f A = f eba , f B = f eab , (5.2)

where eij is the unit vector starting at i and pointing towards j.


This is a big deal! Instead of having to find two force vectors, we have now reduced it to finding
a relationship for a single scalar, namely f .

5.17 Q: I already sort of knew this from my engineering mechanics


class when we did two force members and trusses. So, what is the
big deal?
Well, I sort of slipped one over you somewhat sneakily. There are several things that we did at once,
so let me outline the steps more carefully:
1. We first removed the end masses and just drew a FBD of the bar (see figure 13).
2. Then, we considered what happens to the remaining massless bar when forces are applied to
it. NOTICE that the direction of the force is NOT along the original direction AB
in figure 11 but along the final direction ab. This is because the balance of linear and
angular momentum should be applied to the loaded state ab where the force is acting, not the

27
Figure 13: FBD of our idealized lumped parameter system starts out as AB in between the lumped masses (shown
as dotted lines since they have been removed) at the two ends. we then consider the forces transmitted from the end
masses to the bar in the middle: Since the bar is massless, it will move until (a) the forces align with the bar and are
equal and opposite. The forces and the unit vectors eab and eba which indicate the direction of the bar are shown.
The lumped masses will be analyzed separately using Newton’s laws. This idealization approach allows us to separate
application of Newton’s second law and the structural response characteristics. Note that unlike the Engineering
Mechanics class, where the forces were acting on the state AB (and forces were assumed to have negligible effect on
the geometry of the bar), the bar length and orientation changes A LOT here and the forces are aligned in the current
state ab, and not the starting state AB.

unloaded state AB. This will cause all kinds of issues when we try to do solutions but that is
several weeks ahead. Let’s enjoy the simplification and we will deal with the headaches later
(of course, we have a trick for that too:-)) ).
3. We can then see that we need to consider the structural response of bar ONLY under equal
and opposite forces. There is no need to worry about vibrations of the bar or rotational inertia
or other pesky aspects that come from the distributed mass.

5.18 Q: Wait a minute, what do we do with the end masses that we


sort of “threw away”?
We will consider all the forces from all the bars at each joint and then just use method of joints. The
beauty of this entire separate and conquer idea is shown in figure 14 below. One way to imagine
the resulting shape change is as follows : “the truss moves in a way that is determined by the
competition between how the end masses want to move (by Newton’s laws) and the resistance to
the motion from the structural response of the truss.
In your UG classes, the trusses are so stiff that their resistance wins the battle and the truss
hardly moves. For compliant systems, the two sides are sort of “evenly balanced”, and we may have
to consider both. By the way, this process will work with all the other elements as well.
For triangular elements, we will lump at the triangle corners and the triangle itself will have
simple response functions. For beams, the ends will be treated as rigid bodies, etc.
We are now ready to look deeper into the modeling of the elements. This is the topic of the next
chapter.

28
Figure 14: the lumpy skeleton model: Our approach to modeling is revealed in this single graphic: (a) We model
the geometry using a collection of connected truss elements. (b) We then simplify the truss physically by moving the
mass of the truss to the joints. (c) This allows us to focus only on the structural response of the truss bars without
worrying about satisfying Newton’s laws in the bar (they are automatically met when we put equal and opposite
forces. The result is that we get something that we learnt in our Engineering Mechanics classes. (d) We use Newton’s
Laws by leveraging the method of joints.

6 Structural : Elastic Truss Elements

By reading and doing the exercises in this section, you should be able to:
• Define the meaning of the elastic response of a truss,
• Describe the typical shape for rubbery and leathery behavior of elastic materials, and
• Find constitutive relations and strain energy of an elastic truss element.

6.1 Q: We talked in quite a bit of generality about models and separation


etc. Can you get down to some actual models and show me how to
model these?
I know that we have been describing a lot of “philosophical” issues but the aim is to get you to do
your own models and know what idealizations are actually involved in an intuitive way. My aim is to
get you to model the response of compliant mechanisms and structures( not just a cookbook recipe
follower). This is the difference between a junior required class (I show you how to do something,
not necessarily how it works) and a senior/grad tech elective. So, bear with me. We will start by
modeling something that we sort of know, which is an “elastic element”.

6.2 Q: What is an elastic material-model? I only know what I learnt


in my materials science classes and strength of materials— “stress
is proportional to strain up to yield”. Is that what you mean?
Actually, NO! We must distinguish between a “model” which is a cartoon (we have converted all
objects into “lumpy skeletons” already) and it will exaggerate certain features at the expense of
others. In our case, we will ignore yielding, fracture, etc., and focus only on “ideal elasticity”.

29
A truss element is considered “elastic” if
1. the length of the truss element is determined entirely by the current axial load
and not load history (no load memory),
2. its length changes instantly when it is axially loaded,

3. the free length is recovered instantly when the axial load is removed (linked to item 1), and
4. when we apply and remove the load, the net work done by the external axial load is exactly
zero.
The central idea is that (1) the axial load and length are the conjugate variables (since their
product will give you energy or power), and (2)there is a unique relationship between them. In other
words,

faxial = g(length),or
length = h(faxial ). (6.3)

Notice that item 4 is not something that we might have thought about, but it is a key aspect of
elasticity. Without it, you could create a perpetual motion machine— cycle the truss element back
and forth and produce endless work. This is because, the response has NO memory. (For a plastic
or viscoelastic material, no matter which way you cycle, it will consume work and convert it into
heat; this is because it has memory of past states. This is not obvious but we will discuss it later).
This aspect of the response of elastic materials is called “Green Elasticity” due to a remarkable
mathematician by the name of George Green. You should look him up–he was a miller and had
only one year of formal education but he appears to have been a mathematical genius (Remember
Green’s Theorem from vector calculus— yep, that is the guy!).

6.3 Q: But I thought that elasticity for a spring meant f=-kx. How are
these related?
This brings me to a very important point:
Never memorize a concept through some formula: It will completely hide the meaning. Formulae
are like the text messages or emojis of the math world i.e., without context, you won’t be able
to interpret them. Always memorize concepts and ideas by writing them out in words
not in symbols. If you cannot write your concept or idea in words, you didn’t really
understand it, and all you can do is simply use it without knowing what it is.
This is because math only focuses on logic, not meaning. It says “if you know what ‘f’, ‘x’, and
‘k’ are, then for a linear spring, they are related by f = −kx”. But what if the spring was along the
y axis or along the radius or something else? Which way does the spring point?
For example, you might know centripetal acceleration as v 2 /r but what is v? What is r? What
do they mean? If you write it down in words, then you will see that centripetal acceleration is “the
square of the speed divided by the radius of the turn”. Now, it has a meaning!

6.4 Q: So, how should I remember the elastic response for a truss ele-
ment?
We first define the “free length” of a truss element.

30
The free length of a truss element is the length of the bar when no load is applied and the truss has
stopped moving. For an elastic or viscoelastic truss element, the free length is fixed. For plastic or
viscoplastic elements, the free length depends upon its past history.
Now we can define an elastic truss element:
An elastic truss member is one for which the axial pulling force is a function of the difference between
the current length and the free length. If the relationship is linear, then we have a linear-elastic
truss element.
Notice how specific I get: Firstly, I didn’t just say force, I actually said, “axial pulling force” (so
we get the direction that we prefer to use–pointing away from the truss element at either end.
Secondly, note that I said “function of”, not proportional to. Finally, I didn’t say “strain” or
“x” but actually “current length minus free length”. This way you can operationalize the definition
and interpret it.
A steel bar could be considered as a linear elastic material but a rubber band has a nonlinear
relationship between forces and lengths; and both are elastic.
We can write this in symbolic form as
fa = g(l − lf ), where fa is axial force, l is current length, and lf is free length (6.4)
Here, g is called the elastic response function. We can think g for “Green-elastic” as a tribute to
George Green.

6.5 Q: What does this elastic response function g look like for rubber
and steel?
If we plot axial force versus change in length, we will commonly get one of the 3 types of responses
1. Type 1: Linear response: The entire response can be characterized by ONE parameter— the
slope of the line which is called stiffness (not modulus) since it is a structural property
2. Type 2: S shaped curve: This typically has an initial stiff response (with high slope) followed
by a shallow nearly-flat portion and then another much stiffer response. This is called rubbery
response. You have experienced it when blowing up a rubber balloon: very difficult to blow
up initially (stiff response) followed by a region where it is very easy to blow up and then
finally a very stiff response. The last part you can feel with a rubber band–it gets very stiff
before it fails. The result is a classic S shaped curve. Materials that have this kind of elastic
response will be sort of stiff (and not very flexible) and not floppy like cloth and they will feel
...er ...rubbery :-)))
3. Type 3: J type response. There is an initial very low modulus region followed by a very
abrupt stiffening. This is classic cloth like or tissue like “leathery” response. This will make
the material sort of floppy and not very stretchy. It will be very flexible as a sheet but will
not stretch like a rubber-band. Your skin response looks like that.
A word of caution: Real rubber and tissue are NOT elastic as we show, they actually have more
hysteresis and energy loss. This is why dampers and suspensions use rubber and vibration dampers–
the MODEL is simple to use.

6.6 Q: What is the first step for creating response functions for elastic
truss elements?
It is time for some math now. First, we will introduce the variables :

31
Figure 15: The three types of commonly observed elastic responses. The linear response is typically observed in very
stiff materials like, steel below yield, or brittle materials.

Figure 16: Actual response of real compliant systems. Note that polyurethane and silicone have a slight hysteresis,
but human fingertip has a substantial hysteresis. Our elastic model simply ignores this for the time being. Note that
we cannot model hysteresis using any elastic model no matter how complex; we will need some history dependence
or lag in that elastic model. But elastic models are very useful nevertheless because they are MUCH simpler.

• fA is the axial pulling or tensile force on the truss element.


• L(t) is its current length at time t,
• Lf is the “free-length”—the length when the force fA is zero.

32
• Af is the free-cross sectional area. While our model is a thin line (technical called an ”elastica”
by Euler), the real truss element will have a finite cross sectional area. This will also change
when it is extended, but its area when it is fully unloaded is fixed and is called Af .

fA , L(t), Lf , and Af are the minimal list of fundamental variables that we will use for all truss
element models—not just elastic truss elements. All Other variables will be written in terms of these
fundamental variables. We may extend this list to consider other variables also but this is the bare
minimum.

6.7 Q: We have many quantities now, which variables form conjugate


pairs?
fA and L(t) are conjugate pairs, the product has units of energy. More precisely,

dL
fa = mechanical power input to the Truss element (6.5)
dt
This is just mathematical but the usual idea is that “Force times velocity = Power”. The way to
remember it is “axial force × rate of change of length = Power supplied”.
This is a very important quantity. We can think of this quantity as the “earnings” or “rate of
pay”. Different materials do different things with the energy that they “earn”. Their true nature is
“revealed” when you look at what they do.

6.8 Q: I notice that whenever you talk about conjugate pairs you always
say “product must be energy” but you emphasize “power”. Why is
that?
In all of engineering as well as chemistry (as opposed to pure physics), we are interested in moving
things (motion or chemical reactions) and we want to find out how fast something will go. This
depends upon POWER SUPPLY. As we go, you will see that both energy and power matter but
somehow, power turns out to be more useful.

6.9 Q: Now that we have decided on the variables, what is the next
step?
The next step is to introduce a set of useful derived variables (again, for all trusses, and not just
for elastic truss elements). In principle, we can just work with the list of variables that have been
introduced but we need some more for this to work more....elegantly .
So, here is the list of derived variables that we will use:
• The ratio L(t)/Lf := λ is called the “stretch ratio”. This is very important, and λ will always
be reserved for “stretch ratio”.
• ∆L = L(t) − Lf is called the “extension” (and measured by an Extensiometer, like in your
MEEN 360 tension testing lab)
• E := λ − 1 = ∆L/Lf is called the engineering strain and is zero when there is no load.
• e := ln(λ) is called the logarithmic strain or “true strain” and it also goes to zero when there
is no load.

33
• σR = fa /Af is the Engineering stress (force per unit free cross sectional area). Again, σR and
λ are our conjugate pairs. If you multiply them, you will get a quantity with dimensions of
energy per unit volume. As usual (you guessed it...), we will talk also about power.


σR = power per unit free volume (6.6)
dt
The way to remember this is “engineering axial stress × rate of change of stretch is the
mechanical power per unit free volume.”

6.10 Q: Why is it force times length and not force times distance moved
by the force as in my Engineering Mechanics classes?
See this brief video on the work done on the ends of a rubber band.

6.11 Q: Now what is the next step?


We can now make the response of a Rubbery and leathery material precise:
For a linear elastic material, the graph of the engineering axial stress versus the stretch is a straight
line; For a rubbery material, the graph of the engineering axial stress versus the stretch is a S-shaped
curve; For a leathery material, it is a J-shaped curve (see figure 15). In all these cases, it is customary
to use the quantity λ (easier to compute) rather than engineering or true strain for modeling.
For a linear elastic material, this curve can be written in an easy way: it is a straight line of the
form
Y Af
σR = Y (λ − 1) ⇒ fA = (L(t) − Lf ) (6.7)
LF
The first version talks in terms of stress and engineering strain, and the slope is the Young’s Modulus
(since we have already used up E, we have to go with Y for “Young’s Modulus”). We can also write
it in terms of the axial forces and length.
We can see that the former version does not involve the starting geometry of the Truss element,
but the latter one involves Af and Lf . So, the former will be called a material property (geometry
independent) but the latter will be called a “structural response” (geometry-dependent). If we know
that the material is linear elastic (or stiff), we can characterize them by using a single number: the
Young’s modulus Y . If we know its value, we can then write the relationship (6.7).
There is NO consensus on what characterizes rubbery and leathery materials! We still report
“Young’s modulus for rubber” but what exactly is it? And, what about the S-shaped response or
J-shaped response of skin? It is a ROYAL MESS.
So, let us try something simple: let us focus on the S-shaped curves. There are two broad ways
to do it: The first one is to simply do a curve fit, like what you perhaps did for one of your lab
classes like MEEN 404 (design of experiments). Pick some polynomial in λ with unknown constants:
N
X
σR (λ) = an λn = a0 + a1 λ1 + a2 λ2 + · · · (6.8)
n=0

and, use excel or whatever to do a regression fit. This is straightforward, but such an approach has
real problems for our elastic materials. First, there are some basic requirements that have to be
met. They are:

34
1. when L is equal to the free length, there should be no force (definition of free length). This
is the same as saying λ = 1 (current length is 1 times free length). So, we should have
σr (λ = 1) = 0.
2. σR should be negative for λ < 1(compression), and σR > 0 if λ > 1 (tension).

3. It should be impossible to make λ = 0. We shouldn’t be able to fully crush the material


(current length is 0 times the free length). So, σR (λ = 0) = −∞.
4. The first three are required. This one is in the highly recommended category: generally, we
expect that σR should be a monotonically increasing function of λ.
Item (3) above precludes the use of simple polynomials like (6.8). Why?–put λ = 0 in (6.8) and
you will see that the only way to meet item (3) is to put a0 = ∞, thus, giving you absurd answers.

6.12 Q: So, what to do? How do we get rubber elasticity curves that
satisfy the proprieties listed above?
The most common way to get out of this difficulty is to use “Divide( separate) and conquer” again.
We take care of λ < 1 (compressive) response and λ > 1 (tensile) response separately and then mix
them together (see Fig 17). For λ > 1, we can simply use powers of λ like λα . This will curve up

Figure 17: Mixing and Matching different pieces for overall response

if α > 1, like what we want for leathery materials. If α < 1, then it will curve down (like what we
want for rubbery materials). It, however, goes to 1 for
lambda = 1 and zero for λ = 0 and. So, we need to fix that.
For λ < 1, we want things to dive down rapidly towards −∞. We can do this by choosing
negative powers of λ. For example if we pick f (λ) = −1/λ2 it will go to negative infinity as λ tends
to 0. it will go to zero as λ tends to +∞. So we will use functions of the form 1/λα = λ−α to
account for the compression.
By the way, there is no necessity for α to be integers.
So, we should try functions of the form

35
N
X
σR (λ) = mi (λαi − λ−βi ), (6.9)
n=1

where mi and αi and βi are adjustable parameters.

6.13 Q: How can we be sure that they meet the criteria?


As long as (a) the positive and negative powers occur in pairs, we can be sure that σR = 0 when
λ = 1 (since the positive and negative powers will cancel out). Also, each pair of terms will give you
either an s-shaped or a flattened curve.
A specific form of (6.9) is very popular and is called the OGDEN model and is quite commonly used
for rubbery materials. For our truss element case, their model is of the form :
N
X
σR = mi (λαi −1 − λ−αi /2−1 ). (6.10)
i

Here, the adjustable parameters are mi and αi .


We now keep some terms (say 2 or 3 terms) and try to match the data by doing nonlinear
regression.
A specific form of the Ogden model with 3 terms is shown below:
Ultimately, engineering judgement must be exercised to determine the appropriate number of
constants for a particular application (balancing, for instance, the need to minimize the error to
acceptable levels against the danger of over-fitting).

6.14 Q: For a given truss element, how do we find the parameters?


Welcome to nonlinear regression or “curve fitting”. The basic idea is the following: We are given
a list of data points (xi , yi ) obtained from experiments. We have some model function with some
adjustable parameters pi which is written as f (x; pi ). For example, the Ogden model will be written
as f (λ; mi , αi ). The terms before the semicolon is the independent variable and that after the
semicolon are the parameters. They are collectively listed as a list or “vector” p for simplicity.
For any set of parameters, pi , the error between the actual values and the measured values is given
by
X  yi − f (xi , p) 2
Error(p) = , (6.11)
i
si
where si is the “standard deviation” of the i-th measured data. It tells you how “uncertain” you
are about the value that you just measured. Another way to think about it is that 1/si are the
“weights” and they tell you how important that data point is. If you put si to be a very small value
relative to the others, then that data point is very important. If all the data are equally important,
just put si = 1 (the default in many programs) Observe that the Error depends upon the values of
the parameter list p but not on the x or y. If your parameters are a “bad fit”, the error will be high.
If the parameters are a “good fit”, then the error will be small.
Our best parameters are the values of the “list-vector” p that minimizes the error. So, this is an
example of the minimization problem.

36
Figure 18: An example of the response curve for the Ogden Model

6.15 Q: How do we do the minimization problem?


Luckily, the curve fitting problem is so common and so important that programs like MATLAB or
python have built in functions that can do it.
I urge you to look at this video on how to use EXCEL for doing nonlinear curve fitting. This is
a super useful one since it even shows how to fit an S-shaped curve.
A similar video (with a lot of details on what the residuals etc., look like and how to know if you
have enough terms) is found in this video on how to use curve-fit with python.
If you are a MATLAB person, the explanations are not as good or as detailed as the one for
python but here is a short video on how to use MATLAB for curve fitting (nonlinear regression)
For any of them, you have to do the following four steps:
1. Make a list or vector of [x]values and [y] values : these are the values from the experiment–
the “ground truth”. How good your function is, is determined by how well these are matched.

37
2. Create the matching function f (x; p). How this should be written is described by MATLAB
or python.
3. Choose the si : this is called standard deviation or uncertainty in the data points.
4. Choose the starting guess for the parameters p.

5. Call the curve fitting or nonlinear regression function with these values. Typically, the curve
fitting function will look like:
bestparameters = curvef it (f unction, [Xlist ], [Ylist ], [Stddevlist ], [startingparameters ])
In Python, this is done through the curvef it module: here is

BONUS MATERIAL: YOU CAN SKIP IF NOT INTERESTED


6.16 Q: How does this minimization process work?
In your math class, they will tell you that to minimize a function f(x,y) find the gradient and set it
equal to zero. This will give you two equations for two unknowns. Then, solve for x and y, i.e., you
have to do:
∂f (x, y) ∂f (x, y)
= 0, =0 (6.12)
∂x ∂y
then check the second derivative to ensure that it is positive to get a local minimum. If we have a
function of N variables f (x1 , x2 ...), then find its n partial derivatives and set it equal to zero and
find the minimum by solving n equation for n unknowns.
This is correct but not useful.
The reason is in most cases, the solution to (6.12) is too hard. There is a simpler way. If you
are on a hillside and want to reach a valley, what would you do? (a) find a down hill direction (b)
take as many steps downhill as possible in that direction (c) find a new downhill direction. If all
directions are uphill, then you have reached the bottom–you have nowhere to go but up. See figure
20.
This works pretty well in all minimization problems. The key questions we have to decide are
are (a) where to start, (b) how to select a downhill direction, and (c) how far to march?

6.17 Q: How do we know where to start?


The starting point is really quite important. If your function is “nice” i.e., bowl shaped (technically
“convex”) , you can start anywhere, and you will reach the minimum; but if your function is
“bumpy”, then where you end up depends upon where you start (see figure 21). You require
judgement (i.e., know approximately where you are going). In some cases (and curve fitting is one
of them), if you find any minimum that gives you a curve with acceptable error, you can stop. If it
doesn’t give you an acceptable curve, there could be two problems: (a) your starting point was bad,
or (b) your fitting function was not chosen well. For this, you need to test your minimum by starting
at different random starting points. Part (b) requires a GREAT deal of judgement and exploration.

38
[thb!]

Figure 19: An annotated python program for basic curve fitting

39
Figure 20: Gradient Descent: In 1-D in multiple dimensions you have to find a minimum direction

Figure 21: For convex functions, it doesn’t matter where you start, you will always end up at the bottom, For bumpy
functions, you might end up in different places depending on where you start

40
6.18 Q: How do we find a downhill direction?
An easy way to find a downhill direction is to sample a bunch of different points and find a point
that is downhill . A very clever way to do this was invented by Nelder and Mead in 1965. It is
the most popular way to solve and is implemented in MATLAB as fminshearch(). To get you an
idea of its impact, this paper has been cited 34139 times! Just that one paper is more cited than
the combined citations of all but a handful faculty in our department (only Prof. Reddy’s Finite
Elements book is cited more than that).
Also, they were not mathematicians: They belonged to the National Vegetable Research Station
in Warwick!!! It is more important to be creative than rigorous!
The idea is sample:

• Pick the minimum number of data points to sample: in 1-D, we need two points to find which
way to go. In 2-D, we need at least 3 points; they will form a triangle. In 3-D, pick 4 points;
they will form a tetrahedron... and so on. This is called a simplex.
• Among the points, select the HIGHEST (yes, highest not lowest).

• Do a mirror image of the point to the opposite side–go in the direction opposite to the highest
point, you will get a new simplex;
• Repeat. The simplex will “tumble” downhill.
This video is a really cool animation of how the Nelder-Mead algorithm works.

6.19 Q: Is this the only way? Are there other ways?


Yes, there are 1000s of other ways, the next most popular way is called gradient descent. In 1-D,
we can find a downhill direction by finding the slope of a function (see figure 22). Similarly, the
gradient of a function is a convenient way to find the direction of steepest descent—we just go in
the direction negative to the gradient, since this will cause the function to decrease in value. One
way to see this in 2-D is by doing a “contour plot”. The contours represent likes of equal height so
the gradient is perpendicular to the contour lines. With this geometry, we can descend along the
gradient like this: See this website to get an idea/animation of how gradient descent works. This

Figure 22: Gradient descent in 1 D and 2 D. In 1-D, the slope of the function tells you which way is downhill. In 2-D,
the gradient is the vector that is perpendicular to the contour lines and points uphill. We go downhill, and how far
we go is controlled by δt which is called the “learning rate”.

41
also explains how the algorithm called “momentum” works.
This is just to provide you with insight into how minimization works. The main thing is to
think that all minimization problems work by mostly going downhill from some starting point and
sometimes uphill (as is done with momentum) so it doesn’t get stuck in a ditch (technically a ditch
is a local minimum and you cannot get out of one unless you build up some momentum.

42
7 Strain Energy-Energetic and Entropic Elasticity

By reading and doing the exercises in this section, you should be able to:
• summarize the notations and definitions for describing truss element response,
• describe strain energy and the difference internal energy and strain energy,
• describe what happens to the work done on a truss element
• describe qualitatively what occurs in a compliant material like rubber and how it differs from
steel,
• describe what is meant by energetic and entropic response for truss elements, and
• describe some consequences of the difference between rubber and steel response

7.1 Q: Can you summarize the basic quantities and their definitions in
one place so that I can kinda see what is going on?
OK, here goes
Quantities and Definitions
Variable Name Notation Comments
Current Length L length across ends of bar at time t – usually treated
as independent (control) variable

Free length Lf Length when no force is acting, also called gauge


length. Related to geometry of element and must be
known up front

Stretch Ratio λ most important parameter – all strains are defined


through this.

Axial Force Fa Force applied to the ends of the bar

Free Area Af CS area of the bar for the truss element: Another ge-
ometrical parameter that should be known up front

Engineering Stress σR = Fa /Af Force per unit free area. Is the conjugate to the
stretch ratio

True Strain e = ln λ useful function of λ that goes from −∞ to ∞ rather


that 0 to ∞ .

True stress σ = λσR It is the conjugate to the true strain in 1-D.

Strain Energy den- W (λ) Is the maximum recoverable work you can get by
sity unloading the truss. ∂W /∂λ = σR . If you write it
in terms of e, then∂W /∂e = σ.

43
7.2 Q: What is the difference between the strain energy and the internal
energy of a material?
When you deform an elastic truss element, you do work on it. If it is perfectly elastic (with no
internal damping), then you can get it back when you (carefully) unload it. Of course, you can get
more or less work by also heating and cooling a body but then we are talking about carnot cycles
etc. So, let us consider only isothermal conditions here.
The maximum amount of work recoverable from a body when unloaded under isothermal conditions
is called the strain energy of the body: Its other more technical name is Isothermal work function
or Helmholtz potential or “Helmholtz free energy”. Read about him, he also discovered how color
vision works among many other things.

7.3 Q: why is it not the same as the internal energy of the body?
The point about strain energy (and free energy, in general, is that it is not necessarily stored in the
body whereas internal energy is STORED in the body.
Think of it this way: Internal energy is Cash on Hand. Free energy is your net worth including
your investments in the outside world etc.

7.4 Q: Do you mean to say that the work that is done is not actually
stored in the body?
So, what happens is this: When you do work on a body, the body is earning some “apy”. If you are
a material like steel, you are very conservative, you don’t trust banks etc. You don’t spend it, and
so you keep it under a mattress in your house. When it is needed, whether slowly or quickly, you
can immediately get it back.
Also, you are relatively insensitive to the external environment—no matter what the state of the
economy is, you will be fine.
For steel, the mattress is its bond energy—when you deform it ,the bonds stretch and work is
“stored”. When you unload, it will give it back to you.
You could be more of a “gambler” than this. If you have higher risk tolerance (Gases and
Polymers do), you don’t simply put the money in your mattress.... You put it in a bank or other
assets. You don’t “store it”.
So, even though you earn money, your cash in hand does not increase. Your internal energy is
unaffected, you are not storing cash.
But when you want to retrieve it, you have a problem— it takes a while to “liquidate” your
assets. You can only retrieve all of it if you have sufficient time. This is how gases work.
TO get a better idea, Consider a gas in a thermally conducting cylinder. The Internal energy
(cash on hand) of a gas is just the KE of its molecules. When you compress it slooooowly, the
molecules of gas may get just a little faster (i.e., its KE and temperature increase slightly)—the gas
earned a bit of cash. But the temperature difference means that the extra energy diffuses out — the
gas has invested its money. Its internal energy does not change. When you want the work back, the
reverse happens, the gas liquidates its assets and gives it back to you but only if you do it slowly. If
you try to get work out fast, the gas doesn’t have time to liquidate its assets,it has to give it from
cash on hand, so its KE decreases and the gas becomes cold.
Polymers with their long chain molecules also behave the same way:. The internal energy is
mostly the Jiggling of the chains. When you stretch the rubber band, the chain uncoils and jiggles
less —the bonds don’t stretch at all. Just like a gas, the polymer will also “invest” all the work done

44
on it under isothermal conditions by simply uncoiling and recovers it upon heating. In fact, steel
expands when you heat but a rubber band shrinks as you heat it. This is because heating will cause
the rubber band to jiggle more and so, the end to end distance of the polymer will come down.

7.5 Q: Phew, I did not know that. But how does it matter to me?
Well the behavior explains how the S shaped curve comes about. Rubber bands have long chain
molecules that are sticky (due to the van der Waals forces between the chains). So when you
start pulling them, they first unstick, then uncoil (during which the stress is practically the same)
and finally, when fully stretched the bonds or cross links will begin to stretch (causing increasing
stiffness). For tissue, the initial stickiness is not present due to the presence of water which keeps
the strands apart.

7.6 Q: What is meant by energetic and entropic elasticity


From Thermodynamics, the free energy of a truss element under isothermal conditions is given by
a function of the stretch λ and the temperature T as 5

W = U (λ, T ) − T S(λ, T ), (7.13)

where U is the internal energy, T is the temperature and S is the entropy. This is called the Helmholtz
free energy. Roughly, U is the cash on hand which depends upon the bond energy and the KE of
the molecules (represented by the temperature T) −T S is the ”other assets invested outside. Note
at any temperature, the lower the entropy S the more the assets are invested outside. We can think
of T S as the ”net liabilities” with assests being negative liablities. ”So think of the Free Energy as
Cash on Hand - liabilities”
We are interested in mechanical processes so we will keep the temperature T as constant and
change only the stretch λ. When we do that, we can see from (7.13) that both U (cash on hand)
and S we change.
We can consider two extreme cases6 depending upon where λ appears, ie whether the internal
energy or the entropy depends upon stretch.
Case 1: U is only a function λ and S is only a function of T

Wenergy = U (λ) − T S(T ) (7.14)

In this case, changing the stretch will cause the internal energy (cash on hand) to change.
Everything is stored. This is roughly what happens in stee. Everything is due to changes in the
bond energy.
CASE 2: The other extreme case is when U is only a function of T and S depends upon λ.

Wentropy = U (T ) − T S(λ) (7.15)

Now changing λ will cause only the entropy to change and no change in energy occurs, this is called
entropic elasticity. This is the response of gases and rubbery materials.
Most materials are somewhere in between.
5 We wont go into why or how I got this form, but trust me, this is what it looks like
6 please beware that this is only a ”cartoon” to get you a feel for what happens. Real materials will have more
complications

45
7.7 Q: Where do viscous fluids fit into the picture
Viscous fluids are the extreme example of a profligate person. Whenever they earn, they “spend”
it; no storage, no investment, nothing: they party all the time. No wonder their lives are turbulent.
:-)))
For these purely viscous truss elements, since they do not store or invest the supplied power under
isothermal conditions (you cannot change the bond energy or configurational entropy by doing work
on the system), theif free energy is only a function of the temperature, ie

Wviscous = U (T ) − T (S(T ) (7.16)

So changing λ will have no effect on the Internal energy or entropy.


A standard way to model viscous filaments is by setting the true stress to be a function of the
rate of true strain
σ = f (ė) (7.17)
A material is called a Linear viscous fluid or Newtonian fluid if (in 1-D) the true stress is given
by
σ = µė (7.18)
µ is called the viscosity.
Here is fun example of a ”viscous truss” (a viscous version of the Eiffel tower) slowly collapsing.

7.8 Q: are all viscous fluids linear viscous?


Just like polymeric solids which have non-linear elastic response, most industrial fluids have a non-
linear viscous response. If we plot the magintude of the true stress versus magnitude of the strain
rate, we will get the following curves
The cuplike response is that of a shear thickening fluid. Under constant stress they will flow
easily initially but become much more viscous and not flow well as the stress increases— This is the
response of OObleck.
The curve which bends down is called a shear thinning fluid—extremely common. Most industrial
liquids and slurries –from face cream to fracking fluids–will have this form
They are very sluggish at low stresses but will flow easily at high stresses. That is why paint
and face cream are easy to apply (flow when you force it) but stay on the surface when you you stop
shearing it.
The extreme case is called a Bingham-Plastic–They are almost solid like until the force exceeds
a critical value–then they will ”flow”.
This is very desirable for 3-D printable materials, when it flows out of the nozzle the stress is
high so it should flow easily, Once you deposit it, the stress will be low and you dont want it to flow.
SO we try do design polymers that can do this well.

7.9 Q: You said that for viscous fluids, the work done is not stored,
then what happens to it?
Pushing the cash analogy a bit. If you spend all your money the moment you make it, you are living
paycheck to paycheck, so you can move about only if you are always getting money, you cannot live
off your savings. Viscous fluids can move about only as long as the power supplied is positive, ie
only as long as
σ ė = f (ė)ė > 0 (7.19)

46
All of this power supplied is converted into heat” i.e., it will increase the entropy of the surroundings
and will be lost–dissipated.
FOr a linear viscous fluid, the above condition (?? is satisfied because

σ ė = µ(ė)2 > 0 (7.20)

so it wont flow unless you supply power to it. That is why a spring will bounce whereas a viscous
fluid will not.

7.10 Q: Are elastic materials and viscous fluids the only possiblities?
Just as with humans you rarely if ever come accross a 100% hoarder (elastic) or a 100% wastrel
(vscous fluids). Most humans and mateirals hoard some, invest some and spend some. These are
viscoelastic materials.
We will study more about them and other viscoelastic materials in the next chapter.

7.11 EXTRA MATERIAL: SKIP IF NOT INTEREsTED


7.12 Q: I keep hearing about entropy and I have some vague idea about
it but what exactly is it?
Entropy means ”degree of uncertainty about the state of the molecules (also called microstate) when
the state of the body is known”. For example in the truss element, just because you know the total
length of the truss, its temperature and the force on the ends, does not mean you know what the
molecules are actually doing– you have an idea that the molecules are confined to the bar but you
cannot know for sure where each molecule is exactly—this is what we mean by Uncertainty and we
quantify it by entropy
Roughly there are three sources of uncertainty. To understand it, consider a box of air. If I reach
in and try to grap a molecule, I will be uncertain about
• where the molecule is exactly (we know that it is in the box, but where is it?—this is called
configurational entropy. this depends upon the size and shape of the container and can be
changed by changing shape (mechanical processes). THis is the focuse of many nonlinear
(polymeric) mechanics classes.
• how fast is it going.– this is called Thermal entropy . THis depends upon the Temeperature of
the body and can be changed by changing its temperature. THis is what is the primary focus
of thermal science classes
• What kind of molecule is it? is it Nitrogen, Oxygen.... THis is called mixing entropy and can
be changed by changing the concentration of materials. This is the focus of chemistry
If you want more details on this, watch this video On Entropy, Heat Engines and Hurricanes

47

You might also like