Predictive Engineering White Paper On Small Connection Elements-Mpc and Cbush Rev-1
Predictive Engineering White Paper On Small Connection Elements-Mpc and Cbush Rev-1
TABLE OF CONTENTS
1.
Summary ..................................................................................................................................................................... 6
1.1
1.2
2.
Introduction ................................................................................................................................................................ 7
3.
3.2
3.3
3.4
4.
INTRODUCTION ................................................................................................................................................................................... 21
4.1
4.2
Page: 2 | 40
4.3
5.
Appendix ................................................................................................................................................................... 36
5.1
List of Figures
Figure 1: How Femap creates a MPC and some terminology from the NX Nastran User Guide. ......................................................................... 8
Figure 2: Basic FEA terminology for the way MPCs are handled. ......................................................................................................................... 9
Figure 3: A dependent DOF of the RBE2 (node 2) is given a SPC in the Z-direction (circled on the graphic above). The resulting
error message calls out the problem using the NX Nastran terminology of GRID POINT 2 COMPONENT 3 ILLEGALLY............... 10
Figure 4: A six DOF RBE2 was used in Figure 3. A functional RBE3 is shown on the right. Note the Independent DOFs required for
stability in the RBE3 formulation. If one does not check the RX box for the RBE3 element, NX Nastran will provide an
error message. ..................................................................................................................................................................................... 11
Figure 5: A column is pulled with a force of 1,000 lbf. A RBE3 is used to uniformly distribute the load. RBE2s are used to
simulate a bolted connection to the base structure. .......................................................................................................................... 13
Figure 6: A full-on simple example of how a RBE2 works with all six DOF enabled and then reduced to just one DOF. ................................... 14
Figure 7: The connection R-element is switched from a RBE2 to a RBE3. Only translation DOFs are used for the independent
DOFs on the RBE3. If all six DOF are enforced (left), the result is reasonable whereas with only three translations DOF,
the result is not plausible for a bolted connection. ............................................................................................................................. 15
Figure 8: With the RBE3, it is impossible to force a rotational stiffness into the structure. As per the formulation, only the load is
transferred via the MPC regardless on how the dependent node is defined. Please note how the independent nodes
are defined for stability. ....................................................................................................................................................................... 16
All Rights Reserved 2013
Page: 3 | 40
Figure 9: When transferring load (or weight via a mass element), there is nothing like a RBE3 to do a nice job. Note that the
spatial relationship of the force is captured by the R-element (i.e., RZ moment = 20). ..................................................................... 18
Figure 10: Result of using a rigid RBE2 for thermal analysis. Note: The only applied load is that of a uniform temperature. ........................ 19
Figure 11: How to setup the RBE2 to handle thermal loading. ........................................................................................................................... 20
Figure 12: The CBUSH element setup within Femap v11. The column labeled Structural refers to Structural Damping. ................................ 22
Figure 13: If the user does not select an Orientation CSys (highlighted), then the X, Y and Z directions align with the DOFs as 1, 2
and 3. The rotational DOF follow this same convention. ................................................................................................................... 24
Figure 14: Femap setup for the CBUSH element. Even though the CBUSH property card was set to use Cartesian CSys, the
element definition still requires an Element Orientation Vector (viz., just like a beam element)...................................................... 26
Figure 15: Comparison between RBE2 and CBUSH connection modeling. It is best said that there is no free lunch since the CBUSH
element does not quite provide rigid type results, unless one uses high stiffness values, which could then lead to
numerical problems. ............................................................................................................................................................................ 27
Figure 16: Another comparison between a RBE2 and CBUSH connection element. The CBUSH element does not create a rigid
mechanism. .......................................................................................................................................................................................... 28
Figure 17: The obvious setup for a CBUSH element attached to a solid element where the load is vertical (Y-direction)................................ 29
Figure 18: The CBUSH nodes are coincident and the load case is a 100 lbf vector load in 1,1,1 direction. The CBUSH element is
stable. ................................................................................................................................................................................................... 30
Figure 19: The solid element only has translational DOF and if the CBUSH tries to attach to nothing, NX Nastran will respond
appropriately. ....................................................................................................................................................................................... 31
Figure 20: With the end node of the CBUSH constrained in RX, RY and RZ, the element is stable. ................................................................... 32
Figure 21: With a stable arrangement, the CBUSH can transfer moment loads onto solid elements; however, the stiffness of the
CBUSH element is likewise transferred unlike the RBE3. .................................................................................................................... 33
Figure 22: The CBUSH elements with rotational constraints coupled with the RBE2 elements in a cylindrical coordinate system
provides a smooth torsion transition along the tube. ......................................................................................................................... 35
Page: 4 | 40
Figure 23: An internal pressure is applied to the cylinder. The RBE3 provides constraint without rigidity. The dependent node is
SPCed in all six DOF at the center. This is impossible without transferring these SPCs to the Um DOF. .......................................... 37
Figure 24: Setting up the RBE3 with the Um option to provide restraint to the internally pressurized cylinder. .............................................. 38
Figure 25: Another approach to solving the internal pressurized cylinder with an RBE2 connection. ............................................................... 39
Figure 26: Using the CBUSH Element to obtain the same result as shown for the RBE3 and RBE2 connections. .............................................. 40
Page: 5 | 40
1.
SUMMARY
1.1
This white paper assumes that the reader has the basics of FEA down pat and an inkling of how R-elements work. The objective is to
describe in detail how to use R-connections and CBUSH elements correctly and with confidence. If you make it through this note, youll
most likely know more about these little connections than 99% of your peers.
Well cover the basics of MPC terminology which is the foundation of the RBE2 and RBE3 connections. A keen understanding will be
provided on how to think in terms of independent and dependent nodes. Itll be obvious after this discussion that it is not logical to apply
SPCs to dependent nodes or to connect other dependent nodes between different R-elements. Best practices will be covered and some
recommendations given.
The thermal CTE capability of the RBE2 connection will also be covered for completeness.
Lastly, the CBUSH element will be introduced and applications given on how to use this replacement for the CELAS element. The downfalls
on using this element will be discussed and also why this element is a useful as a companion to the RBE2 and RBE3 elements.
1.2
ANALYST RECOMMENDATIONS
R-elements (RBEs) are not elements but multi-point-constraints (MPCs) that just happen to look like elements graphically.
Never apply SPCs too dependent nodes.
Try not to daisy chain RBEs.
R-elements work at the constraint level and offer a powerful idealization tool that ensures that the stiffness matrix will not become
ill-conditioned as is possible with spring elements.
R-elements are linear. If the analysis is expected to be nonlinear use Sol 601 where the R-element formulation is switched
automatically to a spring-type formulation.
Thermal CTE capability is possible with R-elements. The element is switched from Rigid to Lagrange.
CBUSH elements are the go-to-replacement for the CELAS element.
CBUSH is the multi-function, six DOF spring element that correctly models the spring behavior between non-coincident nodes.
All Rights Reserved 2013
Page: 6 | 40
2.
INTRODUCTION
This white paper will walk you through the theory underlying multi-point constraints (RBE2 & RBE3), CBUSH, CGAP and CBEAM elements
and how to use these elements to create efficient and elegant FEA models. Our goal is to provide a working knowledge of this technology
and we encourage you to build your own FE models to verify the concepts presented in this paper. As we tell many of our support clients,
there is nothing like structured organic learning.
3.
The tomes that have been written about R-type elements number in the dozens. In their usage they have friends and foes. For one highly
experienced Nastran engineer that I know, the use of RBE3 elements is generally prohibited among their crew simply due to their ability to
suck the life (i.e., load) out of a standard analysis run and such death-inducing ability is often reason enough to just say no to their usage.
To get a handle on how R-elements work, it is best to think of them as multi-point constraints (MPC). Which then leads to the question as
to what exactly is a MPC? Taking a look at the NX Nastran Element Library Reference, one is presented with some dense logic but at its
core, it defines a MPC as a linear equation among chosen displacements that is equal to zero. Heres the mathematical form of a MPC:
Page: 7 | 40
Figure 1: How Femap creates a MPC and some terminology from the NX Nastran User Guide.
It is nigh near impossible to reinforce this concept of independent / dependent DOFs without a little matrix math. Along with this matrix
math, well introduce the jargon that NX Nastran uses for error messages when the user applies a SPC to an R-element dependent DOF.
Figure 2 shows a sequence of equations that should help to illustrate the MPC relationship to the FEA process.
[ ]{ }
Page: 8 | 40
[ ]{ }
{ }
Where us is the set of independent DOFs and um is the set of dependent DOFs.
Using these relationships, the dependent DOFs are defined as:
]{ }
Where A* is a special transform of the MPC constants (see NX Nastran theory manual).
Figure 2: Basic FEA terminology for the way MPCs are handled.
With the dependent DOFs removed, it seemed a bit obvious that NX Nastran would bark at any user that had applied a SPC to any
dependent DOF of an R-element. An example of such an error message is given in Figure 3.
Page: 9 | 40
Figure 3: A dependent DOF of the RBE2 (node 2) is given a SPC in the Z-direction (circled on the graphic above). The resulting error
message calls out the problem using the NX Nastran terminology of GRID POINT 2 COMPONENT 3 ILLEGALLY.
One may note that the model happily accepts the forces applied to the two dependent DOFs of the RBE2. This is logical given that forces
have nothing to do with the {u} column matrix and one can apply forces to either independent and/or dependent DOFs.
Although we havent formally defined the RBE2 and RBE3 element, one would most likely not be reading this white paper unless you had
some inkling of their function. But just to be formal, heres a brief description:
RBE2: A MPC that defaults with six DOF for its independent node and the user can then choose what DOF to enforce upon the
dependent nodes. The MPC creates a rigid mechanism between the independent and dependent nodes.
RBE3: A MPC that defines the constraint behavior of the independent nodes as a function of its dependent nodes. The element has
no default DOF and the user must choose the MPCs constraints carefully to prevent a free body mechanism. The MPC interpolates
the overall constraint behavior of the dependent nodes upon the independent node. This formulation imparts no rigidly between
All Rights Reserved 2013
Page: 10 | 40
the independent and dependent nodes. One key application is the smearing of a force from the independent node to the
dependent nodes.
Now, lets take the given RBE2 above and morph it into a RBE3. Figure 4 shows the setup for this transformation. Since RBE3s are meant
for interpolating between the dependent / independent nodes, NX Nastran gives you the option to scale this dependency with a Factor.
Within NX Nastran they call this a local weight factor and the default Factor is 1. This implies that the dependent node is equally influenced
by the independent nodes. If one is really clever and has done their reading, in some rare cases, it makes sense to change this factor but
honestly I have not stumbled across any good reasons. The other odd looking option is UM DOF in the dependent node section. This is a
rather rare and unique option but it boils down to letting the user apply a SPC to the dependent DOF on the RBE3 element. Although this is
typically a no/no it does this trick by taking nodes from the independent set and trying them back into the MPC to allow the dependent
node to be constrained. If this topic is of interest, take a look at the Appendix.
Figure 4: A six DOF RBE2 was used in Figure 3. A functional RBE3 is shown on the right. Note the Independent DOFs required for stability
in the RBE3 formulation. If one does not check the RX box for the RBE3 element, NX Nastran will provide an error message.
Page: 11 | 40
In the exact formulation shown in Figure 4 the model solves with no problem. As with the RBE2 element, applying forces to either the
dependent/independent DOFs is allowable. If one has been paying attention to this discussion, one may ask what about applying SPCs to
RBE3 dependent DOF? Well, that is a question that is best left to the inquisitive mind by building and running a small RBE3 model; dont
want to spoon feed all the good stuff.
3.2
If one grasps the concept of independent and independent DOFs in their usage, every common problem can be avoided.
With the above in mind, here are three rules for using RBE elements or R-elements safely:
o Never have a dependent node defined in another R-element.
o Never have a dependent node defined in another R-element.
o Never have a dependent node defined in another R-element.
RBE3 elements can be brutal. Think of them as small little free bodies floating in space. They need to have sufficient DOF defined to
be stable but no more. An over constrained RBE3 is illogical and NX Nastran will do unexpected illogical things to the solution. A
couple of rule-of-thumbs are provided:
o Only use translation DOF for the independent DOF. If rotation DOFs are absolutely necessary, test it as if your life depended
upon it.
o Do not hook a RBE3 into a RBE2 unless you really know what you are doing.
o RBE3s are awesome elements and do a great job of distributing load without enforcing rigidly. They can be extremely useful
if used with a bit of understanding.
3.3
3.3.1
The usage of the RBE2 is fairly trivial for someone experience in the ways of NX Nastran. That is, if you dont apply SPCs to the dependent
DOFs or tie in other RBE2s or RBE3s, youll have smooth sailing. The key concept to a RBE2 is that it is a rigid mechanism that is defined
All Rights Reserved 2013
Page: 12 | 40
via a linear relationship among its selected nodal DOF. Since it is defined as a constraint relationship, it is formulated once at the start of
the solution and then used henceforth forward. If a nonlinear solution is performed (i.e., SOL 106), dont expect a RBE2 (likewise a RBE3) to
know the difference between a linear and nonlinear deflection. It is a little caveat that can sneak up and bite you if you are running a lot of
SOL 106 analyses. As an analysts note, if you run the NX Advanced Nonlinear Solution (SOL 601), it will convert your RBE2/RBE3 elements
into another non-MPC formulation.
To start out this application section, a joint connection is modeled using a beam element and two RBE2 elements. Figure 5 shows the
results for this configuration with the RBE2 set to constrain the dependent nodes at all six DOFs and then just the translation DOFs.
Figure 5: A column is pulled with a force of 1,000 lbf. A RBE3 is used to uniformly distribute the load. RBE2s are used to simulate a bolted
connection to the base structure.
Page: 13 | 40
What I like about this example model shown in Figure 5 is that both solutions look realistic. The only difference is how rigid one makes the
RBE2 by the use of additional DOF applied to the dependent DOFs. In this example it makes sense that as one completely constrains the
dependent DOF (left) the stresses would be higher than that on the less constrained RBE2 (right).
A good homework exercise for RBE2 elements is to create a small model and then play around with the dependent DOFs of the MPC. A
prior seminar was done on connections in 2011 (see Predictive Engineering Resources web page under Seminars) where the model shown
in Figure 6 was used to explore this concept.
Figure 6: A full-on simple example of how a RBE2 works with all six DOF enabled and then reduced to just one DOF.
At this stage, the R-element model shown in Figure 6 should be self-evident. One of the nice things about this figure is the use of the freebody-diagram (FBD) to verify or illuminate the behavior of this MPC (remember that RBEs are actually not elements).
3.3.2
The next example shown in Figure 7 is that for the RBE3 where the same switch is used between the dependent node. As given in our
guidance, only translational DOF are used for the independent nodes thus leaving only the dependent node to play around with. The
All Rights Reserved 2013
Page: 14 | 40
results shown in Figure 7 are stark showing the huge differences that can happen with an interpolation element that does not impart
stiffness but only attempts to balance the relative displacement movement between the independent to dependent nodes.
As in the prior example with RBE2 connections, a simple example is presented in Figure 8 to help explain how the MPC nicknamed RBE3
works. The difference between these R-element formulations is immediately apparent since the RBE3 cannot provide any stiffness between
the independent and dependent nodes, the beam structure is not stiffened and only forces are transferred. This is an extremely useful
feature for the modeling of many features where the analyst does not want to impart stiffness into the structure.
Figure 7: The connection R-element is switched from a RBE2 to a RBE3. Only translation DOFs are used for the independent DOFs on the
RBE3. If all six DOF are enforced (left), the result is reasonable whereas with only three translations DOF, the result is not plausible for a
bolted connection.
Page: 15 | 40
Figure 8: With the RBE3, it is impossible to force a rotational stiffness into the structure. As per the formulation, only the load is
transferred via the MPC regardless on how the dependent node is defined. Please note how the independent nodes are defined for
stability.
Page: 16 | 40
3.3.3
The last example that is shown in Figure 9 is that for a mass element attached onto some brick elements. The model is loaded with a body
load and the objective is to transfer the weight of the given object smoothly onto the structure with no artificially induced stiffness. If one
tries to do this with a RBE2, one induces a rigid mechanism onto the structure and the force transfer could be considered artificial. When
the R-element is switched over to a RBE3 element, the stress results show a nice smooth pattern.
To verify the behavior of the R-element, FBDs were included. The mass element with a weight of 10 lbf was placed at 2 from the center of
the tube. Given the 1 g loading, it is heartening to see that a RZ moment of 20 lbf-in was calculated for both element types.
One little detail to note is that since we are attaching to solid elements, the respective dependent (RBE2) or independent (RBE3) nodes can
only represent translation DOF since solid elements only have translation elements. It is a rather small detail since NX Nastran
automatically ignores the users request for rotation DOF, but if someone else reviews your model and notices such modeling detail they
will wonder if you know what they know or just think that you really dont know what you are doing.
Page: 17 | 40
Figure 9: When transferring load (or weight via a mass element), there is nothing like a RBE3 to do a nice job. Note that the spatial
relationship of the force is captured by the R-element (i.e., RZ moment = 20).
Page: 18 | 40
3.3.4
There is another facet to using MPCs that should be mentioned and that has to do with thermal-stress work. A common experience is for
analysts to create bolted connections using RBE2 elements and then need to add in the effects of thermal stress. Lets take the prior model
from Figure 5 and throw a thermal load onto the top column structure of 100 degrees. Well use a play CTE value of 1e-5. Figure 10 shows
what happens when thermal considerations are ignored. Since the rigid link is rigid with no knowledge about a CTE value, it stays rigid and
restricts the expansion of the surrounding material.
Figure 10: Result of using a rigid RBE2 for thermal analysis. Note: The only applied load is that of a uniform temperature.
Since this is a somewhat common practice, Nastran allows the option of adding a CTE to the RBE2 connection. However, when this is done
the element formulation is no longer a MPC but one that adds stiffness terms to the [K] matrix. The formulation switch is called LAGRAN in
Nastranese since it uses a technique called Lagrange Multiplier Method. The short story is that Nastran handles this switch efficiently
and rarely does the Lagrange multiplier method produce an ill-conditioned stiffness matrix. Figure 11 shows the results for this switch. One
RBE2 element is left in its standard formulation (which is denoted as LINEAR within the NX Nastran Element Library Reference).
Page: 19 | 40
Page: 20 | 40
3.4
RBE2 transfer stiffness as a rigid body among the independent/dependent node set.
The analyst can control the DOFs where stiffness is transfer but at the end of the day stiffness is added.
A RBE2 is always numerically stable (no free body mechanisms) since the independent node starts out with a full set of DOFs.
RBE2 and RBE3 elements both correctly capture the spatial relationships between the independent/dependent nodes.
RBE3 elements can create free body mechanisms if the selected nodes do not create a stable mechanism.
The force interpolation capability of the RBE3 element is often worth the hassle of dealing with its special restrictions.
MPCs operate at the constraint level and do not create stiffness terms within the [K] matrix. Although this makes these connections
only suitable for analysis work that is strongly linear, they also do not perturb the matrix with high stiffness terms as spring elements
might do to obtain equivalent behavior.
If thermal CTE is desired be aware that this switches the R-element to a Lagrange formulation and stiffness terms are added to the
[K]. Although this method is efficient, the R-element is no longer a MPC and it is possible to ill-condition the stiffness matrix [K].
4.
INTRODUCTION
The CBUSH element is now the default spring element for Femap users as shown in Figure 12.
Page: 21 | 40
Figure 12: The CBUSH element setup within Femap v11. The column labeled Structural refers to Structural Damping.
Historically and numerically there are problems noted with the use of the CELAS formulations when the two nodes that define the spring
element are not coincident. This is pointed out in the NX Nastran Users Guide:
Zero-dimensional Elements
When you use CELASi elements to represent concentrated springs between two components
of translation, the directions of the two components must be coaxial. Even small deviations in
direction can induce a significant moment to your model that does not exist in your physical
structure. When you use a CELASi element, the locations of the two end points should be
coincident to avoid this type of problem. If the two end points arent coincident, consider using a
CROD or CBUSH element instead.
Page: 22 | 40
Given this uncertainty and a preference for not having coincident nodes within a model where merge operations might occur (although
Femaps smart merge capability would never merge two nodes connected with an element), we have standardized on only using the CBUSH
for our springey needs. For anecdotal evidence, this principal engineer once used a nested mess of CELAS elements on a large model and
had this exact same effect occur. The sad part was that this error was pointed out to me by the clients chief engineer. I felt pretty small. A
rather complicated example of how to screw up your model with CELAS elements is given in the NX Nastran Element Library if you feel so
compulsive.
The CBUSH element should be treated as you would treat a beam element with six DOFs. Once this adjustment has been made, their
rational use is not such a big deal. Figure 13 shows a schematic from the NX Nastran Quick Reference Guide. As with beam elements, the
user needs to define the orientation vector of the element to fix the Y-direction (DOF 2) of the element. For this reason, our standard
practice, if possible, is to define the CBUSH using the standard global Cartesian CSys and avoid this bit of modeling.
Another consideration about CBUSH elements is that they can be used in all solution sequences and, in particular, they are correctly
translated for SOL 601 (NX Advanced Nonlinear) to handle large deformations. As one may remember, MPCs (i.e., R-elements) are defined
within the constraint matrix and are strictly for linear mechanics. This is an advantage of the CBUSH element is its ability to morph correctly
for the SOL 601 analysis. If the user is interested in running the standard NX Nastran nonlinear analysis (SOL 106) and large deflections are
foreseen, then NX Nastran has the CBUSH1D formulation. I would rather not dwell on this odd-ball formulation since if you really need
large deflection; one is better served with the CBUSH element and SOL 601. If this topic peaks your interest, take a look at the NX Nastran
Basic Nonlinear Analysis Users Guide and the Advanced Nonlinear Theory and Modeling Guide.
Page: 23 | 40
Figure 13: If the user does not select an Orientation CSys (highlighted), then the X, Y and Z directions align with the DOFs as 1, 2 and 3. The
rotational DOF follow this same convention.
4.1
The CBUSH formulation is stable and proven with no surprises. It is the go forward spring element for all seasons.
It is the infinitely variable stiffness beam element. The user can control any DOF.
Nonlinear capability built into the element along with advanced dynamic features.
The only downside is that it has six DOFs and the user has to make some mental adjustments when using these elements.
The user should remember that they are adding stiffness terms into the stiffness matrix [K] and that if one chooses large CBUSH
values it could cause unexpected numerical difficulties. Note: MPCs are based on constraints and do not affect the [K].
All Rights Reserved 2013
Page: 24 | 40
4.2
4.2.1
To not get too far ahead of ourselves, Figure 14 shows how the CBUSH element is defined within Femap. Even though we had set the
CBUSH property card to use a global Cartesian CSys, it still requires the user to define an Element Orientation Vector. The first example will
start out with mimicking the behavior of RBE2 element shown in Figure 6 by removing the current RBE2 and replacing it with a very stiff
CBUSH element as visually described in Figure 15. If one sets the stiffness the 6th DOF within the CBUSH card, the resulting reaction
moments (RZ) are somewhat like the prior RBE2 setup. The difference is that the element is free to pivot at the center whereas the RBE2 is
a rigid mechanism. This result is given in Figure 16.
Page: 25 | 40
Figure 14: Femap setup for the CBUSH element. Even though the CBUSH property card was set to use Cartesian CSys, the element
definition still requires an Element Orientation Vector (viz., just like a beam element).
Page: 26 | 40
Figure 15: Comparison between RBE2 and CBUSH connection modeling. It is best said that there is no free lunch since the CBUSH element
does not quite provide rigid type results, unless one uses high stiffness values, which could then lead to numerical problems.
Page: 27 | 40
Figure 16: Another comparison between a RBE2 and CBUSH connection element. The CBUSH element does not create a rigid mechanism.
Page: 28 | 40
4.2.2
The biggest stumbling block with the CBUSH element is that it requires a little thinking in its application. To illustrate this point, lets take a
solid element and pull on it with a CBUSH element. The CBUSH is vertically aligned with the top corner node of a brick element. The load is
then aligned to pull in this direction.
Figure 17: The obvious setup for a CBUSH element attached to a solid element where the load is vertical (Y-direction).
This is the only configuration that works. If we move the top node of the CBUSH element just the tiniest distance out-of-alignment, the
solution fails. For example, we take the top node of the CBUSH and move it 0.001 in the Z-direction, the run will fail with an error message
as RUN TERMINATED DUE TO EXCESSIVE PIVOT RATIOS IN MATRIX KLL with the R1 DOF listed as the problem. This makes sense if one
works through the DOF logic. That is, the solid element only supports translation DOF and no rotational stiffness is available for the CBUSH
element to grab onto and furthermore our CBUSH element is setup to not have any rotational stiffness. With the element offset in the Zdirection, the element is unstable in the RX (R1) DOF. Lets show another example where we make the CBUSH element coincident with the
All Rights Reserved 2013
Page: 29 | 40
corner node and apply a force in all three directions. In this example, the CBUSH element will have an initial stiffness in the 1, 2 and 3
directions. Figure 18 shows the results for this test case.
Figure 18: The CBUSH nodes are coincident and the load case is a 100 lbf vector load in 1,1,1 direction. The CBUSH element is stable.
Page: 30 | 40
What is interesting and logical, that if the CBUSH nodes are coincident, the element does not contain any rotational free-body-mechanisms.
More importantly, this example encourages the reader to focus on how the FEA method works. The first step is the formulation of the
stiffness matrix followed by the application of constraints (SPCs and MPCs). The stiffness matrix is then decomposed for the application of
forces. If the stiffness matrix is non-singular (i.e., excessive pivot ratios in the matrix KLL), the run will abort. Loads have nothing to do with
the analysis running or not running. Given this understanding, it makes sense that if one removes the stiffnesses in the DOF 1 and 3 within
the CBUSH element, the solution would proceed as if nothing is unusual. But youre not out of the woods quite yet. If the CBUSH element
is defined to have a rotational stiffness, then the solution will fail predictably (see Figure 19).
Figure 19: The solid element only has translational DOF and if the CBUSH tries to attach to nothing, NX Nastran will respond appropriately.
For the sake of conversation, lets say that you needed to apply a rotational load through this one CBUSH element? It would be impossible
without adding a plate element or some beams to the top surface of the brick element to provide the additional rotational DOF. Saying
that, there are some other tricks one can do. Figure 20 shows a setup where the CBUSH element is in the position as shown in Figure 17
(not coincident) but with its end node constrained in all rotational DOF (RX, RY and RZ). Since the CBUSH element has stiffness in the
rotation DOFs (4, 5 and 6); the element is stable and NX Nastran solves appropriately.
Another example is that where a group of CBUSH elements can likewise provide stability for the application of a moment load. Figure 21
shows this arrangement and demonstrates that the utility of the CBUSH for a wide range of load application.
Page: 31 | 40
Figure 20: With the end node of the CBUSH constrained in RX, RY and RZ, the element is stable.
Page: 32 | 40
Figure 21: With a stable arrangement, the CBUSH can transfer moment loads onto solid elements; however, the stiffness of the CBUSH
element is likewise transferred unlike the RBE3.
Page: 33 | 40
4.2.3
The last example shows how to use the CBUSH element as a torsion coupler between two hollow rods (see Figure 22). To join the two
cylinders that have a slight gap between them, a series of CBUSH elements were created between the two surfaces using Femaps Mesh /
Connect / Closest Link command. They are shown graphically as the dense mass of lines and numbers in the middle of two cylinders. As
discussed earlier, to allow the CBUSH elements to account for moments, one end of the CBUSH elements had SPCs applied in the 4, 5 and 6
DOFs. If one zooms in on the graphic, these numbers can almost be discerned.
For the torsion load application and to cleanly constrain the tube, RBE2s were used. The procedure was to create a user defined cylindrical
coordinate system centered down the axis of the cylinder. The nodes of both ends of the tube were then modified to have their nodal
output coordinate system in this cylindrical coordinate system (Femap command Modify / Update Other / Output CSys). In this manner, the
RBE2 connection dependent DOFs are in the coordinate system of R, theta and Z.
Given this starting point, the logical question is why the torsion load was applied using a RBE2 and not a RBE3 which has been advertised as
the perfect force smearing connection. This opens up a whole can of worms but maybe it is best that we peer into this can all together.
The difficulty or challenge with the RBE3 on this particular application is that the RBE3, with the default weighting factor of 1.0, distributes
the moment equally to all the attached nodes. For the torsion to be smoothly applied, the quantity of moment (resolved force per node)
needs to be scaled per the area fraction of the node. In other words, the shear (force/area) needs to be smooth across the end section. To
get this behavior with a RBE3 would require a bit of effort to calculate individual weighting factors for each ring of nodes. Whereas, if one
uses the RBE2, it just enforces a fixed displacement in the theta direction and we are good to go. The moral of this example is to not get
fixed in your ways but be cognizant of the objective youre trying to achieve.
At the other end, the end constraint using a RBE2 is easily explained since the RBE2 only has two active dependent DOFs in the theta and Z
direction. The independent node is SPCed in all six DOF and correctly constrains the structure.
Although this example appears a bit cumbersome, it covers a nice twist to the standard plug and chug examples where everything is
obvious and transparent.
Page: 34 | 40
Figure 22: The CBUSH elements with rotational constraints coupled with the RBE2 elements in a cylindrical coordinate system provides a
smooth torsion transition along the tube.
Page: 35 | 40
4.3
The CBUSH should be the preferred and automatic element choice when a spring-like element is needed.
Numerically, it adds stiffness terms directly to the stiffness matrix [K] and as such, user defined stiffness terms should be kept
reasonable, i.e., not set to 1.0e99.
Many advanced features can be enabled for it use with nonlinear and dynamic solutions.
It should be treated like a beam element with six DOFs. When attached to solid element that only have translational DOF, some
care needs to be taken to account for stability and load transfer.
It is a very useful element to make your models more accurate with less numerical expense.
5.
APPENDIX
5.1
Very few analysts avail themselves of this option. As discussed, only independent nodes of a MPC (RBE2 and RBE3) can have SPCs. Since
these elements werent quite complex enough, it was decided to allow the user to apply a constraint to the dependent DOF of the RBE3
element. This ability was added to allow the RBE3 to restrain the structure while simultaneously smearing loads or displacements. In
practice, it is most likely rarely used but it exists and therefore merits a bit of an explanation.
Figure 23 shows an example of a cylindrical tube with an applied internal pressure. The tube is simulated as freely expanding. To restrain
the free-body-motions of the structure (the six mechanisms of TX, TY, TZ, RX, RY and RZ) without restricting the uniform radial expansion
the following sequence of operations was performed:
1. Create a RBE3 with the dependent node at the center of the cylinder and the independent nodes around the circumference;
2. Apply a SPC with all six DOF fixed to the dependent node (should never work but here comes the trick);
3. Create a cylindrical coordinate system with its center at the dependent node of the RBE3 and Modify / Update Other / Output CSys
of the independent nodes (for convenience since you really only need to do the nodes within the Um set);
4. Open up the Um dialog box within the RBE3 and select the TY and TZ DOF and then three nodes within the independent set equally
spaced around the perimeter. This sequence is shown in Figure 23. Note: The reason to use a cylindrical coordinate system allows
All Rights Reserved 2013
Page: 36 | 40
the mapping of the TY and TZ onto the Theta and TZ directions of the nodes. When tied together in a set of six DOF, they restrain
the six free body mechanisms of the structure. Conceptually this is the hardest step of the process.
5. Analyze and interrogate.
Figure 23: An internal pressure is applied to the cylinder. The RBE3 provides constraint without rigidity. The dependent node is SPCed in
all six DOF at the center. This is impossible without transferring these SPCs to the Um DOF.
Page: 37 | 40
Figure 24: Setting up the RBE3 with the Um option to provide restraint to the internally pressurized cylinder.
5.1.1
We would be remiss not to show that this same problem can be solved using a RBE2 element. A cylindrical coordinate system is still
employed but the dependent DOFs that are engaged are only the TY ad TZ. Results from this model are shown in Figure 25. This is a
unique example and of course there are situations where the Um option would be useful but in this particular example there are options.
Page: 38 | 40
Figure 25: Another approach to solving the internal pressurized cylinder with an RBE2 connection.
5.1.2
Since we are beating this example to death, yes, one could get the same result using the CBUSH. The important trick is to set the CBUSH
Orientation CSys into the user defined cylindrical coordinate system defined for the RBE3 and RBE2 examples. Figure 26 shows the results
from this application.
Page: 39 | 40
Figure 26: Using the CBUSH Element to obtain the same result as shown for the RBE3 and RBE2 connections.
Page: 40 | 40