Professional Documents
Culture Documents
Volume 112
Sponsored by:
Volume 112
Volume 112
If you are new to the COSMOS Companion, a few comments on the program are warranted. The COSMOS
Companion series was developed in response to the request from many of our users for more detailed
information on specific and/or new functionality within the COSMOS products. Additionally, many users have
been asking for clarification of common design analysis questions to enable them to make more representative
analysis models and make better decisions with the data. Whats more, users have asked for this material to be
made available in a variety of formats so they can review it how and when they wish. To address this, each
COSMOS Companion topic has been pre-recorded and made available thru the COSMOS Companion
homepage as a downloadable or streaming video with audio, as static PDF slides for printing, or as a live
webcast enabling attendees to ask questions and engage in additional discussion. We are trying to provide
continuous learning on your schedule so you can be as effective and efficient as possible when using
COSMOS for design analysis and validation.
It is important to note that this material is not developed as an alternative to instructor led training. We still
believe that the best introduction to any of the COSMOS products is in a class led by your resellers certified
instructor. In this program, we are hoping to build on the lessons learned in your initial training. In fact, we
will make the assumption that you have basic knowledge of the interface and workflow from intro training or
equivalent experience. We will try not to repeat what was taught in those classes or can be found in the on-line
help but to augment that information.
Volume 112
Topics to be Covered
Introduction to Boundary Conditions (Loads & Restraints)
5 Steps for specifying Loads & Restraints
Two Approaches for identifying model boundaries
Example showing comparison of assembly layers
Example showing complicated assembly interactions
Comparison of different Load / Restraint methods
In this edition of the COSMOS Companion, well be discussing Tips for selecting Loads &
Restraints, also called Boundary Conditions. Well use examples to focus on determining
which parts should be included in the analysis to give the most realistic response without
bogging down the whole process. The extent of the modeled parts is the Boundary of your
model and you need to define Loads & Restraints to account for interactions (or conditions)
at this boundary; hence Boundary Conditions. Two approaches for establishing the model
boundary will be discussed with the pros & cons of each.
Volume 112
Boundary Conditions
Loads & Restraints
Represent the Interactions of the Parts You
Didnt Model with the Parts You Did!
Your Loads & Restraints MUST NOT Impose
Displacements, Stresses, Rigidity, or Other
Behavior That the Parts You Didnt Model
Wouldnt Have Imposed.
There are a lot of words here but the concept is extremely important. Once you are
comfortable with the fact that your choice of Loads and Restraints MUST cause behavior
that the interactions they represent would have caused, youre most of the way towards
making good BC choices every time. There may be situations where the BC choice doesnt
give you the response youre looking for and you dont know how to fix it. However, at
least youll have seen the problem and can call for help. It is when the problem isnt
noticed and the results are accepted blindly that FEA goes bad.
Volume 112
IN fact, Ill go as far as saying that improperly understood (& consequently mis-applied)
loads and restraints is the #1 cause of unrealistic results.
So how do you get a better handle on the load and restraint choices for your models.
Following these 5 steps should make the process easier.
In this session, well focus on establishing model extents, or boundaries, since the rest of the
process often falls right out of this exercise. Subsequent sessions will elaborate on these
other areas.
Volume 112
Sohow does one establish which parts or sub-assemblies need to be included in the model
and which dont? Well look at 2 approaches for this. The first, what I call The Sledge
Hammer approach is valid for nearly all cases but becomes extremely cumbersome as the
size, as in # of parts, of an assembly increases.
Essentially, you identify what part or parts you are interested in. If it is the entire assembly,
so be it. It is often 1 or 2 parts that are your primary concern and everything else just holds
them or loads them. Once youve isolated your parts, identify the interactions between the
parts you modeled and the ones you didnt. At those interactions, take your best guess at
loads and restraints that represent those interactions and solve the model. Save those
results, add on another layer of parts and repeat the process. When the results of interest
stop changing, youve identified the optimal model extents.
Volume 112
In this example, an engineer at a company with a long history of FEA questioned the BC
scheme that had been developed over the years for their castings that are used on a railcar
floor. In the past, this company made the decision that since they only made the casting and
werent responsible for the structural steel members that attached it to the rest of the floor,
they would simply constrain their part where it interfaced with the rest of the
systemshown in green on the model.
Interestingly, they never seemed to predict any problems with their designs. Good
engineering or bad FEA?
Volume 112
Using the sledge hammer approach, the part of interest, the casting, is isolated for a firstpass analysis. The loading was straightforward so well focus on the restraints or the
interactions with the rest of the railcar.
After analyzing the casting alone with restraints at all points of interaction with the rest of
the floor, the C-channel spine and the cross I-beams are added.
In the final iteration, the beams were shortened to see if simply moving the restraints off of
the casting made the difference.
Follow up iterations might include adding the outer beams on the floor, the additional ibeams that attach to the spine, and so on.
Volume 112
Comparing the first and 2nd studies, we can see that the model response is night & day
different. Adding in the i-beams and spine allow the casting to respond much more naturally
whereas having fixed much of the casting with restraints in the first case, there is almost no
responseas you might expect.
Volume 112
10
From a stress standpoint, we can see there is also a significant difference in stress
magnitudes and distribution when the beams were added. The model boundary clearly
needed to be expanded.
Case 3, with the shortened beams, might be pursued to reduce model size without impact
the validity of the results. As chosen, the beams were reduced too far and the response
looked more like the constrained casting than the full floor model. Design Scenarios could
be utilized to gradually reduce the beam length to see if there was a good intermediate
length. If you made many of these castings, the time would be well spent.
10
Volume 112
An Engineering Approach
11
The second approach is more appropriate when your assembly gets too large too fast for the
Sledge Hammer approach to be effective. In the more systematic, engineering approach,
we try to make intelligent decisions about interactions before leaping to the conclusion
additional parts need to be added.
One of the key components of this is to determine the relative stiffness between interacting
parts. When the part of interest is an order of magnitude stiffer or less stiff than the part it
interacts with, representing that interaction with a load or restraint becomes likely. Your
selection of a load or restraint is essentially telling COSMOS that you know exactly whats
going to happen and the software doesnt need to work it out. In many cases, the parts you
are concerned about have a similar stiffness to the parts they interact with. If this is the case,
you may still have to revert to the sledge hammer approach.
11
Volume 112
Clamp Handle
3
Interactions
1.
2.
3.
4.
Pin to Base
Rod / Loop to Handle
Operator Loading at Tip
Gravity
2
12
In this example, the handle on a clamp assembly is the primary part of interest. The
clamping action occurs because the Loop on the Hook is pulled over center engaging a
positive latching force in the system.
The interactions between the handle and the rest of the system are listed and will be
discussed further.
Gravity was included for completeness since everything with mass on this planet is
subjected to gravity. However, in this scenario we dont always know how the system is
going to be oriented and it is a reasonable assumption that the other loads are much more
significant in the overall scheme of things. Gravity will be neglected in this problem but
this decision should be noted in the project report.
12
Volume 112
Clamp Handle
Pin to Base
No
Possibly
Possibly
Possibly
13
In the first interaction, well assume that we can spec a pin that can handle these loads. The
Pin Connector in COSMOSWorks will give us those loads so this is one option. Another
option is to simply restrain the hole with a Hinge Restraint that allows local rotation about
the axis. By restraining the hole, we are assuming that the Base Plate is perfectly rigid.
However, it is reasonable to assume that the base plate and the handle are of similar
stiffness so you cant know that when the handle loads up, it wont deform the Base. It is
likely that the base will need to be included in the study.
Also, while a Pin Connector will give us accurate results for the shear load on the pin, well
lose some resolution on the stresses in the handle near the pin. If this is expected to be a
concern, a solid pin with No Penetration contact should be considered. In this case, it is
reasonable to say well use a Pin Connector for the first pass and if the loading in that area
is suspect, a second model focusing just on this area can be solved. This way, we avoid
over-complicating the initial study.
13
Volume 112
Clamp Handle
Rod to Handle
Probably
No
Rod - Handle
Pinned
Possibly
Loop - Rod
Fixed
Options
a)
b)
c)
d)
e)
f)
g)
h)
14
This third interaction is a little more complicated. On the surface, it is simply a pin in a
hole scenario. However, determining the magnitude and orientation of that load is very
difficult. Complicating matters is the fact that the system, at the extremes of its operation
where the handle is pulled over-center, is not fully restrained or stationary. Any restraint
that holds it in place runs the risk of redirecting the load path. Since the load in the system
is generated by the stretching of the Loop and the deformation of the other parts, you should
be fairly confident that a part is MUCH stiffer before removing it and replacing it with a
restraint.
One area of somewhat predictable response is the near-uniaxial stretching of the Loop.
Since the Hook acts in the center of the Loop, some bending is applied but a good first pass
assumption might be that it stretches uniaxially. Since we know that the actuation of the
clamp occurs when the system is pulled over center, one option for a more simple system
(vs. modeling all the parts) might be to calculate the expected stretch load and apply that
to the Pin or Loop at a position just before the bifurcation, or over-center, point of the
handle.
14
Volume 112
Clamp Handle
9.525mm
15
The physical stretch required to pull the handle over-center can be determined graphically
from the part geometry. For the center separation shown, the mechanism behaves as if the
handle needs the stretch 0.9 mm. If we assume 100% of this stretch is taken up by the
Loop, we can estimate a tensile force in the system. This will be conservative since we
know that there will be bending in the hooked part of the Hoop as well as in the Hook, Base
Plate, & Rod. If we apply this load to the model and stresses are acceptable, it may not
matter what the actual reduced value is.
15
Volume 112
Clamp Handle
What is worst-case?
What does this depend on?
Y Displacement
Extreme
Force
Absolutely
Interaction options:
a) Estimate single load reqd to operate
b) Apply an enforced displacement to Handle tip
c) Apply increasing load thru N/L or Design Scenario and track
response in system
Deflection at Rod
Stress in Loop
16
The final interaction on out list was the actuation, expected to be the operators hand. It is
reasonable to assume the steel handle is much more stiff than the hand so replacing the hand
with a load makes sense. However, the nature of these over-center systems is that they are
bi-stable, or they have a point where the force-displacement curve reverses. The plot shown
is indicative of this kind of behavior. In these systems, near the bifurcation point, a small
amount of applied load can cause a large increase in deflection. This is a difficult situation
to capture in FEA. An alternate method is to apply translation to the handle via an enforced
displacement. This controls the stability of the model and you can always request from
COSMOSWorks the reaction force required to achieve that displacement.
Along the lines of replacing the actuation with a restraint, applying the estimated Loop
stretch force to the system with the Handle tip restrained may provide some insight into how
much load is required to actuate the clamp.
16
Volume 112
Clamp Handle
F1
F1
X2
X1
F2
F2
17
The technique for swapping loads and restraints is fairly common. In this simplified
example, the stress solution in each case will be identical. The displacement solution will
differ due to the location of the restraint. In most cases, it is suggested that you choose a
restraint or contact condition that most represents the controlled displacement in the actual
system, if possible.
17
Volume 112
Clamp Handle
Restrained
Reaction = 1,845N
(Remember Symmetry)
Loaded
122,600 N
18
When the Rod & Loop are added to the model and the 122 KN load is applied to the Loop
where the Hook interacts, a reasonable looking response is calculated. Stress levels are very
high though, well beyond 500 ksi. If this was a true operating condition, we would have
some problems with the design. Another piece of data that casts doubt on this load
magnitude is the reaction force at the restraint applied to the tip of the Handle. One plane of
symmetry was used in the model and the reaction force was calculated as 1,845 Newtons.
For a full system then, it would require 3,700 N (830#) to actuate the clamp. This is also
excessive for a hand operated system. Intuitively, it would seem like the elasticity of the
other parts will come into play to reduce the load carried by the Loop and the system in
general.
However, it is important to reiterate that every time you apply a restraint, you are assuming
a part is infinitely rigid in at least one direction. This infinite stiffness can impact the results
of interest if you arent aware of it.
18
Volume 112
Clamp Handle
Single Part Analysis
1,845N * 2 = 3,690N
Hinge Restraint
Restrains Radial & Axial Displacement
19
19
Volume 112
Clamp Handle
Relative Part Stiffness
Previous Calculation showed the
Loop needed 122,620N (X2 =
245,240N) to stretch 0.9mm
How does this compare to the
stiffness of the interacting parts?
20
One way to do this is by analyzing each part separately with a consistent load and
comparing their deflection. This will provide a relative stiffness we can use to decide if one
or more parts can be replaced with a restraint.
After completing this study, you can see that all the parts are of similar stiffness and the
Hook & Rod are both substantially less stiff. Consequently, it is reasonable to expect a full
system analysis is required to accurately capture the overall response.
20
Volume 112
Clamp Handle
Enforced Displacement
Pin Connector
No Penetration Contact
Symmetry Restraints on
All Cut Faces
No Penetration Contact
Bonded Contact
Hinge Restraints
2006 SolidWorks Corp. Confidential.
FESanity
Check Series
Solution Details
Static Study
FFE Plus
Soft Springs Enabled
21
Large Displacement Enabled
This model was set up as shown. A single plane of symmetry was used and a Pin Connector
attached the Handle to the Base. Otherwise the true stiffness of each component will be
accounted for in the study.
Soft springs were used to stabilize the model due to the fact that contact conditions
restrain the Loop / Rod combination. The model is essentially stable but the soft springs
ensure stability in the early load steps as contact forces are being established.
Large Displacement was enabled because the handle is being rotated thru a finite angle.
Without Large Displacement, the No Penetration contact conditions may add fictitious
stiffness to their respective joints as the contact surfaces re-orient. This effect is explained
in more detail in the COSMOS Companion unit on Contact.
21
Volume 112
Clamp Handle
Exploded View of Complete Assy Model
22
Final result is at full displacement of Handle. No intermediate results are provided using a
static study. However, we estimated that the force-displacement response had a bifurcation
point so there is no good way of knowing where we are on that curve without multiple
studies using Design Scenarios or a full nonlinear study, requiring COSMOSWorks
Advanced Professional.
22
Volume 112
Enforced Displacement
No Penetration Contact
Hinge Restraints
Bonded Contact
23
For speed, a partial assembly was solved using a full nonlinear solution. The Hook & Plate
are assumed rigid but the most flexible component in the model, the Pin, is included. This
will give us a good feel as to the sensitivity of the system near the bifurcation point. In this
model, the back of the Loop was allowed to pivot and an enforced displacement was applied
to the Handle tip, as in the Static full assembly solution.
Note that the stress distribution is similar to the other studies suggesting a simpler set-up
might be valid for a trend study. However, due to the redundancy in the system, determining
absolute magnitude stresses will require a more complete assembly model.
23
Volume 112
Outside
2006 SolidWorks Corp. Confidential.
FESanity
Check Series
Inside
24
Graphing stress in the Loop, we see that the response is truly nonlinear but not as sensitive
to final position towards the end of the solution as we might have expected.
24
Volume 112
0
100
200
300
400
500
600
700
40000
-10
35000
-15
30000
Loop Tension (N)
0
-5
-20
-25
-30
25000
20000
15000
-35
10000
-40
5000
-45
0
0
-50
Force (N)
100
200
300
400
500
600
700
25
These plots from the nonlinear analysis tell us a more complete story. Most importantly,
they confirm the expected bi-stable response of this system as the Loop goes over-center.
In the graph on the left, we see that the position of the Handle tip to applied load is very
sensitive at the bifurcation point. However, the graph on the right tells us that once we hit
that bifurcation point, the load in the Loop remains fairly stable. It would be worthwhile to
know if the Handle performance is more sensitive to the Applied Load or the reaction load
at the Loop. If the results of interest are primarily sensitive to the cantilevered applied load,
a simpler solution might work out.
25
Volume 112
Clamp Handle
Linear Results
It would be tempting here to believe that the Handle Only solution is equivalent
to the Complete Assy solution!!
Dont fall for this!!!
2006 SolidWorks Corp. Confidential.
FESanity
Check Series
26
Comparing the result from some of the Static Studies we did, we see that when the
assumption was made that the Loop will carry the entire stretch as the system moves overcenter, (Plot A), the Handle shows the worst response of the three. In (B), when the entire
assembly is modeled so that the over-center deformation can be distributed over several
parts, the Loop does not reach the previously estimated tension and the overall system stress
is greatly reduced. Once we established in the nonlinear analysis that the higher loads in the
system, the Loop tension, is not sensitive to the Handle position after the bifurcation point,
this response is probably the most representative.
Interestingly enough, the results with the highest level of confidence are nearly identical to
the results for the single part analysis (C) which I would feel the least confident about. A
common mistake is to believe that since they look so similar that the simple model is
equally correct. However, weve established that the system response is controlled by
multiple part stiffnesses as evidenced by the difference in (A) & (B). The model in Plot C
cannot account for changes in these stiffnesses so the similarity in this case is a fluke. It
would not be safe to use this model for prediction of response to changes in the system.
26
Volume 112
27
For a model of this complexity, there is no escaping modeling the entire system, at least
once. Armed with that data, one might be able to find a faster solution using the Sledge
Hammer approach backwards.
This model truly highlighted how important an understanding of the actual system and partpart interactions are. When the parts are all of similar stiffness and contribute to the over all
response, especially in this case where the system is deformation controlled due to the overcenter mechanism, youll need to consider expanding the model boundaries to include more
parts.
In this case, foregoing the applied load estimates and simply actuating the device with an
applied displacement got us over many hurdles. The model might not have solved near the
bifurcation point otherwise.
Finally, if Ive said it once, Ive said it a hundred times, in FEA, the wrong answers often
look like the right answers. You have to know what is happening with your parts to be able
to distinguish correct from coincidence.
27
Volume 112
Presentation Summary
In this COSMOS Companion unit, we reviewed:
Five steps for choosing meaningful Loads & Restraints
The first step, Establishing Model Boundaries in detail
A simple approach for straightforward assemblies
Keep adding parts until the results stop changing!
28
This concludes our first discussion on Loads & Restraints in COSMOSWorks. These
concepts are applicable to anyone utilizing Design Analysis.
We discussed the 5 steps for specifying loads and restraints; (1) Establish Boundaries; (2)
Determine if interactions at that boundary should be loads or restraints; (3) Determine the
Distribution of the interaction; (4) Determine the magnitude of the interaction; (5) and
Evaluate the response at these interactions to gain confidence that the part can really move
like that.
We touched on all of these points but focused on establishing model boundaries.
Subsequent COSMOS Companion programs will expand on these other steps.
Finally, reviewed two examples where choice of Loads & Restraints are not immediately
clear. The needs of the
28
Volume 112
Conclusion
For more information
Contact your local reseller for more in-depth training or
support on the load & restraint options in COSMOSWorks or
to discuss modeling techniques
For more on the 5 Step Program, plan to attend upcoming
COSMOS Companion programs
Review the on-line help for a more detailed description of the
features discussed
Attend, or better yet, present at a local COSMOS or
SolidWorks user group.
See http://www.swugn.org/ for a user group near you
29
Id like to thank you for taking the time to join in this edition of the COSMOS Companion. I hope you
will look at your models in a different light and put more thought into your load & restraint choices.
I encourage you to talk thru your model setup and boundary condition options with the support team at
your local reseller and take advantage of their experience in using COSMOSWorks. If you have time,
you should also read thru the on-line help topics on shell elements and shell modeling for more detail on
some of these concepts.
If this information was helpful, keep tuning into further editions of the COSMOS Companion. While
focused discussions on loads and restraints will be scheduled in the future, many of these concepts are
addressed in context of the other topics.
Finally, I hope you have a chance to get involved in a local COSMOS user group. This is one of the best
vehicles for sharing and learning from the experience of others who face the same challenges as you. You
can locate a local COSMOS group on the SolidWorks User Group network website shown. If there arent
any COSMOS groups near you, get involved in your local SolidWorks groups and introduce some
COSMOS related topics to foster some discussion on design analysis and validation.
Thanks again for your time and interest and I look forward to seeing you next time on the COSMOS
Companion.
29