I am interested in seeing if Cosmos Motion can be used to
analyze a rotational system such as an electric motor with a
harmonic forcing function. I am looking to work on a design of a
rotational damper to privide a smoother velocity profile. I will
need to model the forces/torques applied to the rotor shaft as well
as the rotational spring constant of the rotor and the possible
damping constants of the device I am working on in order to
evaluate. I would like to know if this is possible in Cosmos
Motion, has anyone used it for this, and how accurate can the
solutions be driven assuming correct assumptions for parameters
mentioned? Ideally I would like to be able to change the design and
re-run the analysis to obtain a velocity profile - without having
to verify with a physical emperiment for each interation I want to
run. I am hoping to arrive at a closer answer faster, which I know
is what FEA and simulation can do

Pete

Pete

Are you looking at studying/designing the actual motor as part of this or just assuming that it is provinding an enforced harmonic motion?

COSMOSMotion can handle both scenarios but it does determine the method of input and the need to account for rotational mass of the motor winding etc.

Out of the box, you can specify linear/nonlinear torsional springs/dampers. If you need a more complex force function, you can define them manually.

There is a lot of flexibility in terms of modeling the forces acting on the shaft. If you can model the rest of the mechanism that is applying the loads to the shaft, that is the best approach as it eliminates assumptions (unless you are referring to loads such as combustion forces, where as long as you have the force as a function of time or crank angle you can account for it).

From a results perspective, it is straightforward to chart velocity, force, acceleration etc as a function of time or any other result component (such as rotation angle).

COSMOSMotion does use a true physics engine, so it will give meaningful answers to what you would see in the real product.

Hope this helps.

Cheers,

Ian

Thanks for the reply. The entire motor is our design and we manufacture it for our products. Everything is modelled and available for analysis. I am desinging a damper for this particular motor to provide a more consistent velocity profile. This particular design is locked in and I now need to just focus on the damper. For other future designs, I may have more latitude with the overall design. My question was pertaining to more of a framework or approach to the design and whether Cosmos Motion could fit in for me. I suspected it could and I have some experience with it, not a lot yet. I thought I would ask to see if anyone else has used it for analysis such as this and if there were any potential pitfalls. The ability to change parameters in the model and re-run sounds much better to me than having to solve partial differential equations. I am hoping to acheive a good correlation between my virtual model and its velocity profile in Cosmos and data in the real world. I should be able to easily compare my Cosmos output of velocity (angular) verses time to the data from our velocity measuring device in the lab. This should help me acheive a solution for a critically damped mechanism in a shorter time period than multiple physical prototypes. I suspect the most dificult part for me will be to define a forcing function in Cosmos that accurately describes what happens in the real world. Do you have much experience with what I am about to embark on Ian. Any recommendations?

As an aside, if you're interested in understanding the torque being generated by the elctromagnets, you should take a look as COSMOSEMS (developed by electromagneticworks). This produce models the electromagnetic fields for actualy design of these elements to achieve a desired force. I don't know if what your doing falls into the standard motor geometry (where hand calcs can do the trick) or there is some unique geometry where actually modelling the fields generate by the geometry will provde a more accurate response.

I'm assuming that this is a servo motor type application (based on the harmonic input).

You can look at designing the damper in 2 methods.

1) Enforced harmonic motion

For this approach you use an enforced movement as a function of time for the motor (be applying a motion function to the shaft) and can tailor the damper to give a response as a function of the motor velocity position. This is good for when the prescribed motion is well defined. It also is a quick way of determining a damper response based on a movmement and specific time.

2) Applied torque force function

This is the most realistic way (and dynamic way) of studying the motor response. Since the motion is not enforced, it is just a matter of an applied force as a function time, the response will be based on the dynamics of the system (ie inertia of parts involved, reaction loads from whatever is connected to the motor and the torsional damper you are adding to the system.

By default COSMOSMotion does not account for motor controllers that use feedback to adjust input to achieve a desired movement. National Instruments actually have integration with COSMOSMotion (via the API) so that you can use Labview and your controller software to interact with the CAD geometry to actually prototype your system design.

You can emulate a very simple feedback controller within COSMOSMotion using some force functions that monitor specific results and adjust the force input based on some difference between a goal (for position, velocity, etc) and the current value. It is not as accurate as the Labview method, but can give you a good idea on the required force needed to make something move and be able to also limit the max force possible.

Cheers,

Ian

I looked into CosmosEMS and the price tag shocked too many folks here. As a method, I am interested in #2 that you wrote. As I mentioned, I think my biggest challenge may be to come up with a harmonic forcing function to describe what is really happening accurately enough that it represents real world data. I will attempt to work up to this by not considering the damper until I have a forcing function descibed accurately enough on the undamped model. This is for a control application, but motor control here is very easy - on or off. There is no PWM, inverter, etc... It is an AC permanent magnet hybrid motor design. The harmonic input comes from a rotating magnetic field created by 2 different phases of AC current in the coils on the stator. There are no windings on the rotor, only some permanent magnets and pole caps separated by an air gap. The pole cap teeth on the rotor are constantly chasing (attracted to stator teeth) the changing rotating magnetic field while current to the motor is on (think stepper motor). Another difficulty may be to accurately model the torsional stiffness of the system. I belive the torsional stiffness and the torque that generates rotational motion could possibly be two sides of the same coin. Is it possible to place a torsional spring in the same area that place the harmonic forcing function? I believe my equation of motion to be of the general form...

TorqueMotor * Sin XTheta = J * Theta(double dot) + Cv * Theta (dot) + Krot*Theta

J is the inertia that we have been mentioning and this is the easiest part as SolidWorks/Cosmos will caclulate that for me. Cv is the damping that I will be designing in my mechanism. Krot is the torsional spring that I need to model in Cosmos. This will be a value that I have rough equations for, but will need to find an accurate value for.

Pete

These the equations of motion are along those lines. COSMOSMotion does solve the PDE's of the system under the hood.

You can add a torsional spring to the system, but you may need to break some geometry to model it. For example, apply the torque to one segment of the shaft and then have this segement connect via a bushing (has both torsional stiffness and damping) to the second segment of the shaft. You can dial in the real stiffness of the shaft based on the geometry (and add some structural damping). I would only worry about doing a torsional spring to represent the shaft stiffness if you think the twist in the shaft is significant enough to impact the function.

So it the harmonic function a desired movement or are you more thinking about the attraction between the stator and rotor being oscillitory in nature. You can apply a torque function (or force at the rotor magnet location) which would vary as a function of position (ie vary based a harmonic function)

Normally the supplier of the bearings should have some inertia terms in their data sheets. It's been a long time since 've looked for that data. I would think the inertia of the bearing should be pretty small with respect to the rotor.

Cheers,

Ian

Also - I think what I need is to ignore the stator (and all non moving parts), apply a force from the stator to the rotor assembly, and provide a bushing between the stator and rotor assembly to model the stiffness that results from the interaction. I do not think I will need a torsional spring - I was mixing up my terminology. My rotational system stiffness that I come up with I will model as the torsional stiffness of the bushing - that is the Krot term that I listed in a previous post - and can be estimated from some rough hand calcs...

http://homepages.cae.wisc.edu/...ublications/index.htm

You can view pdfs of the theory behind it on this site, but really simply it is just conservation of energy by equating of forces (F - ma =0) and angular momentum.

The following gives a good overview across the board.

http://real.uwaterloo.ca/~mcphee/sd652/intro_sd652.ppt

Cheers,

Ian

I see MSC in the credits for Cosmos Motion. I assume the technology is mostly licensed from MSC. Interesting to know. Thanks for the help thus far. Interstingly... we have a new physical test bench almost complete that is using some Labview software and instrumentation that I am hoping to test these motors on. I do know that Labview and Cosmos can loosely work together. I would be very interested in looking at the physical results out of labview compared to the simulated results in Cosmos Motion. Right now I need to learn to crawl before I can walk and then run though... Back to setting up this simulation and understanding the nature of the problem.

Pete Yodis

Harold Beck and Sons

Pete

Sorry for the delay in responding. The best way to approach this is using phase shifts for separate harmonic functions (for the 3 phenomena you mention). You can do this in the function expression using the SHF function.

SHF (x, x0, a, w, phi, b)

x The independent variable in the function.

x0 The offset in the independent variable x.

a The amplitude of the harmonic function.

w The frequency of the harmonic function (in radians per unit of the independent variable, or degrees if you use a D suffix)

phi A phase shift in the harmonic function (in radians)

b The average value of displacement of the harmonic function.

To do this as a function of the rotation angle, you will need to use the joint markers for the shaft joint and use AZ(I,J) as the x-axis for the harmonic function.

The help (in 2007 and prior, covers how to obtain the marker id's).

So your total function expresison would be something like

SHF (AZ(34,33), 0, 12, 360D, 0, 0)+ SHF (AZ(34,33), 0, 3, 360D, 90D, 0)

+SHF (AZ(34,33), 0, 32, 360D, 270D, 0)

I hope this helps,

Cheers,

Ian

I'll give this a try. I'll admit I read the help files and it wasn't apparent that I could achieve this with joint markers. Thanks.

Pete

In your example you list the AZ(I,J) needed. What is the I value and the J value? I assumed these to be the markers - Is that correct? Do you need 2 markers - an I and a J, or would just one marker do... Also, would it make any difference if each component of torque was repesented as its own torque applied to the model so that you would have 3 torque values applied instead of 1 torque value with 3 components in the expression?

I & J are replaced with the marker id's. There is a button at the bottom of the function expression window that launches a dialog that browses the entities in the model to extract the marker ids.

It is important to use both I and J since the J marker gives you the reference frame with respect to which you are measuring. It can be on ground or another part, so hence the important.

Here is the definition out of the Solverfunctions.chm file (look in the installation directory for this):

The AZ function returns the rotational displacement of marker i about the z-axis of marker j, and accounts for angle wrapping.

Marker j defaults to the global coordinate system if it is not specified. This value is computed as follows: assume that rotations about the other two axes (x-, y-axes) of marker j are zero. Then AZ is the angle between the two x-axes (or the two y-axes). AZ is measured in a counter- clockwise sense from the x-axis of the J marker to the x-axis of the I marker.

As far as applying the torques separately, it makes no difference if that makes it easier for you to manage them.

Cheers,

Ian

Hi Ian - I stumbled on this topic and wondered how this analysis would be performed in SW 2010 Premium as it doesn't seem to be possible to access the markers?

Hi Keith,

Sorry for the delay in responding, have been off the forums for a little bit.

You can do all these things in 2010, but the marker approach has been superceded by just using plot results. If you want to use a result; rotation, displacement, velocity, acceleration, force etc, in an equation, then create a plot to measure what you are interested in and then you can select that plot result from in the function expression editor (any plot results are listed at the bottom of the property editor.

Cheers,

Ian