Stress Analysis of Bicycle Frames
Mike Wahl
the sole subject of finding stresses in bike frames. I know that simulating
the forces on a bike can be very complex, but I am an ME with nothing better
to do. Besides I am curious in finding the Factor of Safety that is used on
different bikes. I plan on doing the stress analysis on a computer, but I
need to have a basic idea of what to look for, so any help would be
appreciated.
Mike Wahl
mw...@bobcat.ent.ohiou.edu
Rob Horvatich,EVEO,,
Hey Mike
I am familiar with the process of finding stresses in car structures, not all that different but much more complicated than bicycle frames.
I would have to say that this may not be as bad as it seems to you.
I actually did this in an FEA class as an undergrad.
What difficulty level is up to you on how much detail you want to include.
The analysis I would imagine to be a simple static analysis.
Many codes are available at the PC level that will handle such tasks of such size.
The tough part would be to accurately model the frame you have in mind.
Do you want beam elements with section properties or shell elements to model the geometry.
You will need information regarding wall thicknesses and butted sections to accurately represent the frame.
Either buy a frame and disect it or contact the manufacturer which I don't feel will be overly willing to release information like that.
As for simulating forces on a bike frame, this is up to you.
You can only approximate frame loads frome riding and jumping.
Unlike some industrial equipment where the loads are known and predictable, a bicycle frame can be subjected to an unpredictable magnatude of force.
ie. a very heavy guy taking big air.
This makes it next to impossible to calculate an overall safety factor.
I am not aware of any standard tests that are conducted to simulate real world durability for bicycles.
I can only imagine you can place a heavy rider (say 250lb) and subject him to say a 5g load and calculate the reaction the frame has to take.
Again, these are approximations that you will need to determine the validity of.
In CAE analysis you can not get away from approximations.
Hope this helps to point you in the right direction in detail needed to perform such an analysis.
---
ROBERT HORVATICH
|0/\ FORD MOTOR COMPANY
0/ \ /\
/ / \/\
"You can't take life too seriously, you don't get out alive", Bugs Bunny
All thought described by me belong and represent no other
Matt Bushore
spots on the frame, some acceleromters, a mini CCD camera, and a laptop
in a fanny pack, then go beat the hell out of the bike and see what's
really happening to it?
Would a telemtry system automaticaly make you a technoweenie?
Rob Horvatich,EVEO,,
Don't forget the strain gauges.
Can we use your laptop?
But hey, isn't that what the FEA simulation is supposed to predict without all that cool techno instrumentation.
You are correct in that it would be the most accurate way to know what is really happening.
Great way to correlate FEA data.
Jim Papadopoulos
>But hey, isn't that what the FEA simulation is supposed to predict without all that cool techno instrumentation.
>You are correct in that it would be the most accurate way to know what is really happening.
act on a frame, in what combinations. Thus instrumented riding
is _essential_ to determine the loads (and build a model for predicting
them on different circumstances).
Once the loads are known, o fcourse FEA should do the job ....
but better measure to make sure. If there are residual stresses
for example, or drawn tubes are anisotropic because of texture,
the numerical simulation won't predict the right stresses (or strains).
Jim
Rob Horvatich,EVEO,,
Hey Jim
You are correct in posting that the loads need to be extracted out of the frame.
This is usually a more advanced step in a design analysis process where a prototype can actually be built.
But since you already have a frame in mind, why not.
You are incorrect in saying a numerical simulation will not predict anisotropic material behavior.
FEA codes exist that are fully capable of processing such data, they just need to be defined.
john.d.unruh
>In article E...@oucsace.cs.ohiou.edu, mw...@bobcat.ent.ohiou.edu (Mike Wahl ) writes:
>> I was wondering if anyone knows of any articles, papers, books written on
>>the sole subject of finding stresses in bike frames.
>
>
>I would have to say that this may not be as bad as it seems to you.
>I actually did this in an FEA class as an undergrad.
I remember an issue of Bicycling mag, probably 5 or 6 years ago, had
an article with a high level overview of FEA of a bike frame. There was
a 3-D color plot showing the forces on the frame during pedaling. They
had a university do the analysis. I think Lehigh, which is close to
the Bicycling mag HQ. The result was different than the gospel as
preached by bike builders.
John Unruh
Antony Pringle
But surely there must be some data out there which deals with
the forces imparted to a frame subject to various inputs...
Does anyone know the month/year of that Bicycling Mag. back-issue
which had the FEA of a bike frame? I remember reading it
at the time - it would be neat to find it and actually understand
it... and it might even give me some ideas for the postgrad
work.
I noticed that the ad which the FEA package 'COSMOS' uses (its
in the ASME publication: Mechanical Engineering) shows an FE plot
of a bike near the BB area. Wonder what loads they used???
--
ap...@freenet.carleton.ca | Tony Pringle
| Carleton University School of
| Mechanical and Aerospace
| Engineering.
Matt Bushore
|>
|>
|> But surely there must be some data out there which deals with
|> the forces imparted to a frame subject to various inputs...
Maybe they have simulated simple pedalling motion or even
a sprint, but
I think its nearly impossible to even fathom what is happening
to a bike during jumping and crossing up or hopping on the front wheel.
Well, maybe not impossible, but there are a wonderous quantity of
permutations to bicycle movement and loadings, with more thought of every
single day.
Rob Horvatich,EVEO,,
That is why in engineering we make so many assumptions.
The real world is too damn complicated to figure out exact solutions for what we look for.
Some of the forces you may have in mind may not be considered significant for certain type of analysis and thus not considered.
They may exist, just are not considere significant.
Of course these assumptions need to be varified, otherwise your analysis is invalid.
Happy trails.........Rob
Jim Papadopoulos
> How difficult would it be to throw some piezo transducers on a few
>spots on the frame, some acceleromters, a mini CCD camera, and a laptop
>in a fanny pack, then go beat the hell out of the bike and see what's
>really happening to it?
job of this yet.
If the interest is in designing entirely new frames,
then what one needs are not frame stresses per se, but
LOADS -- in many cases the loads should not be affected
by bicycle structural design, so they can be used as FEA
boundary conditions.
The difficulty is: a bicycle is loaded at the saddle, steerer,
bottom bracket, and two axles (or two ground contacts).
To determine those loads while riding outdoors
requires some fancy work placing strain gages, or rdesigning
some parts, so that the force and moment applied to a given
part of the frame is measured accurately without any crosstalk.
This is a great thesis topic for some lucky person!
JP
Jim Papadopoulos
>I remember an issue of Bicycling mag, probably 5 or 6 years ago, had
>an article with a high level overview of FEA of a bike frame. There was
>a 3-D color plot showing the forces on the frame during pedaling. They
>had a university do the analysis. I think Lehigh, which is close to
>the Bicycling mag HQ. The result was different than the gospel as
>preached by bike builders.
(The main thrust of the work appears to have been to
generate that picture.)
The boundary conditions, as near as I could tell, were
wrong (based on Tarantula loading device), and in any event just a
single special case.
The students used a shell model to conclude something about titanium
which could have ben learned from considering (not even analysing)
a beam model.
(..if my memory serves me, which is not always the case.)
Bottom line, I think the merit of that study was less than zero...
asking bad questions of a pointless model in a special case
(and possibly wrong) loading situation.
But, take a look and decide for yourself!!!
I know of a couple of studies which were somewhat better
(including one by my student Sarah Trilling, at Cornell)
.. no pretty pictures though!!!
Even so, without a way of thinking about the many possible
loads (singly or together) it's hard to say much about
more efficient designs.
JP
Jim Papadopoulos
>You are correct in posting that the loads need to be extracted out of the frame.
>This is usually a more advanced step in a design analysis process where a prototype can actually be built.
>But since you already have a frame in mind, why not.
>You are incorrect in saying a numerical simulation will not predict anisotropic material behavior.
>FEA codes exist that are fully capable of processing such data, they just need to be defined.
>
If you don't know what loads to apply, what god does it do you
to have a frame in mind?? For example does a hard-landing
bike cause a 200lb, 400lb, 600lb or other force on the rear
contact?? Does it depend on the rider weight, rider skill, and
height of drop? What about the side forces, simultaneously
or sequentially -- 50, 100, 150 ---?
You can't logically take the strength from a current frame, unless
you forswear any changes -- if you take all possible load
cases that a current frame can just barely endure (a 'surface'
in a 'loading space' of about 3+3+3+6+6 dimensions), and demand
that the new frame also endure them, then you have something equally
strong but probably inefficient. If you take just a few points from
the load 'surface', then you risk having the new design
too weak for some untested loadings.
Bikes are a big messcompared tosomething like a car (or even simpler,
a car suspension link).
As for the anisotropy, of course you're right. What I meant was that
without measurements one is tempted to assume isotropy in the model,
and it might not be borne out in practice.
As I see it, the big problem in bicycle design is trying to
cut weight way down, which means reducing factors of ignorance,
yet we are quite ignorant of loads and of the detailed
behaviour of highly stressed materials (in notches, yielding
somewhat, with reversed loading.)
----"A Mad Bicycle Scientist Flame"
Rob Horvatich,EVEO,,
>
>If the interest is in designing entirely new frames,
>then what one needs are not frame stresses per se, but
>LOADS
You are very correct here.
Problem is when starting out with a blank piece of paper, those loads are not known.
You would have to take a "similar" bicycle and gather that data and use that as your assumption for design until a prototype is built.
-- in many cases the loads should not be affected
>by bicycle structural design, so they can be used as FEA
>boundary conditions.
This is not 100% accurate.
Different geometries and designs will yield different road loads.
To what magnatude is questionable (I wouldn't think all that much).
Most bicycles don't have a large deviation from the normal dimensions (relative position of BB, axles, handlebars, seat).
This maintains a consistent rider position so I question how much these seemingly small differences will effect the overall equation.
I am not talking here about recumbents and tandems.
These dramatic changes in geometry will definitely yield a dramatically different set of loads.
The addition of suspension will significantly change road loads.
This will reduce some loads to the frame, and rider.
>This is a great thesis topic for some lucky person!
You bet.
Have a good day..........Rob
Rob Horvatich,EVEO,,
---In article bicyc...@delphi.com, Jim Papadopoulos <bicyc...@delphi.com> () writes:
>Rob Horvatich,EVEO,, <rho...@eve054.ford.com> writes:
>
>>You are correct in posting that the loads need to be extracted out of the frame.
>>This is usually a more advanced step in a design analysis process where a prototype can actually be built.
>>But since you already have a frame in mind, why not
>
>
>
He can input loads into his FEA model taken off of the same frame ridden in some sort of condition, (say off a big jump for maximum load).
>If you don't know what loads to apply, what god does it do you
>to have a frame in mind?? For example does a hard-landing
>bike cause a 200lb, 400lb, 600lb or other force on the rear
>contact?? Does it depend on the rider weight, rider skill, and
>height of drop? What about the side forces, simultaneously
>or sequentially -- 50, 100, 150 ---?
This is the information we need to extract from the existing bicycle he has in mind to analyze.
The sensitiveity of those variable can also be evaluated while collecting data.
I would imagine rider weight and terrain ridden would be two major variable.
>
>You can't logically take the strength from a current frame, unless
>you forswear any changes.
Here again, I was talking about evaluating an existing design, not changing one.
You are accurate with this statement, I think we were talking about two different things.
In reality this is how the initial road loads for a vehicle are prepared.
By taking a similar vehicle with similar projected weight and dimensions, this initial input is used to drive the design, then later verified.
This is assuming that the final design will be similar.
This is much better than pulling numbers out of the air to drive your design.
Bicycle hardpoints don't vary as much as cars so a common set of loads may possibly be standardized.
This would definitely need to be evaluated and I am not speaking from experience on this part.
>
>Bikes are a big mess compared to something like a car (or even simpler,
>a car suspension link).
>
What I had in mind whas a sheet metal structure consisting of hundreds of pieces that have thousands of spot weldeds as well as bonded together
Not real simple geometry or joining techniques.
I know, I have modeled them to a point where all you see is a big mess of lines on the screen (talk about eye strain).
Also the analysis that is done is much more complex.
I was thinking in terms of large deformation dynamic analysis for crash.
To predict accurately how the front end of a vehicle will crush is truely complex.
Parts going far beyond yielding to fold up like an acordian.
It still amazes me how close these analysis predict.
I don't believe bicycles are designed with crush zones or with a consistent failure safe mode.
I may be wrong though.
>As I see it, the big problem in bicycle design is trying to
>cut weight way down, which means reducing factors of ignorance,
>
You are very very correct here.
The more material you shave off, the more you better know about how that material is being used.
This mean using less assumptions and more testing of the product.
Not having the customers be your test facility.
>----"A Mad Bicycle Scientist Flame"
right on.........Rob
Erik Asphaug x2773
complex creatures. Finite-element codes can be quite useful in
predicting the response of _subsets_ of the bike frame to a variety
of stress loads (gradual load, sharp load, torsion, shear). But
to model the entire bike is a pipe dream...
For one thing, close to 10,000,000 nodes (in 3d) would be required
at minimum to resolve the spokes, the hollow frame, the various
joints... For another thing, the answer would be dependent upon the
mechanical properties you assume for your metals. This sounds easy
but it's not always, especially when you're dealing with several
different kinds of metal which can behave differently depending on the
loading rate.
So... you could spend a decade getting "the answer" by modeling the beast
on a computer in a brightly lit cubicle at a national lab, or you could
get "the answer" by test-riding a bunch of new ideas down sunny
back roads. Computers at best can test and facilitate creative ideas,
but the real analysis will always have to take place in the field.
-Erik Asphaug
Rob Horvatich,EVEO,,
>For one thing, close to 10,000,000 nodes (in 3d) would be required
>at minimum to resolve the spokes, the hollow frame, the various
>joints... For another thing, the answer would be dependent upon the
>mechanical properties you assume for your metals. This sounds easy
>but it's not always, especially when you're dealing with several
>different kinds of metal which can behave differently depending on the
>loading rate.
What type of analysis would require that many nodes and elements?
What are you trying to predict.
Why do you think it takes so many elements to define the geometry and capture the results accurately.
I have never seen a model of that size before and would not want to pay for the processing time.
I am not saying they don't exist, but for something on this level and size, I don't see why I would need so many elements.
>
>So... you could spend a decade getting "the answer" by modeling the beast
>on a computer in a brightly lit cubicle at a national lab, or you could
>get "the answer" by test-riding a bunch of new ideas down sunny
>back roads. Computers at best can test and facilitate creative ideas,
>but the real analysis will always have to take place in the field.
What "answer" are you looking for?
It usually isn't sunny back roads that kill bicycle frames.
If it is, I don't want to be one of those testing guys to pick a seat stay from my behind after big air or a pothole.
Remember, we were talking about predicting the stresses in a bicycle frame.
It is not complicated to physically model in FEA.
It is complicated to predict what loading that will be applied.
Road and MTB frames will not experience the same loads obviously.
We do not care about all stresses induced by all loads on the frame.
We do care about the loads that may potentially break it.
You are correct in that the analysis has to be validated with field testing.
To rely soley on what the computer spits out is foolish.
I am not disagreeing with the you on the idea that a bicycle should not be soley designed on a computer.
I don't think they are.
A computer can be a fabulous tool to tell you what ideas to dump before you spend any time and money building and testing.
I am curious how you arrived at some of your figures.
Have a good day......Rob
---
Erik Asphaug x2773
>In article l...@organpipe.uug.arizona.edu, asp...@curly.lpl.arizona.edu (Erik Asphaug x2773) writes:
>>For one thing, close to 10,000,000 nodes (in 3d) would be required
>>at minimum to resolve the spokes, the hollow frame, the various
>>joints... For another thing, the answer would be dependent upon the
>>mechanical properties you assume for your metals. This sounds easy
>>but it's not always, especially when you're dealing with several
>>different kinds of metal which can behave differently depending on the
>>loading rate.
>
>What type of analysis would require that many nodes and elements?
>What are you trying to predict.
>Why do you think it takes so many elements to define the geometry and capture the results accurately.
>I have never seen a model of that size before and would not want to pay for the processing time.
>I am not saying they don't exist, but for something on this level and size, I don't see why I would need so many elements.
>>
Well, we currently use a 150,000 particle SPH code as the resolution
marginally sufficient to show the explicit fracture of a 6 cm diameter
sphere of basalt. This is pretty much the state of the art for
fracture, although I don't quite know where finite element (equilibrated
stress) codes stand. Needless to say, a sphere of uniform rock is a
far simpler system than a bicycle!
Engineers at Sandia use about 30,000 nodes to model (with fairly low
accuracy) the bending of an aluminum tube.
Engineers at the Phillips lab (also in Albuquerque) use over 1,000,000
nodes in an SPH code to simulate (again, with low accuracy) the impact
of a spherical meteoroid into a communications satellite. I know this
is inaccurate because you can observe unphysical instabilities in the
simulation video.
And paradoxically, the lower the strain rate, the longer the computation
takes. Shocks (hitting your Cannondale with a bazooka blast) is a far
simpler calculation than gradual stress (riding over a bump, or simply
applying a downstroke to the pedal).
Fast codes for engineering use linear elements for segments of the
frame, but if you want to model the bending of the frame, and the
interplay between frame joints, you must resolve it at fairly high
resolution. Remember that in 3D, 10,000,000 nodes is only 300 nodes
along each linear dimension of the bike.
- Erik Asphaug
Jim Papadopoulos
>You can't logically take the strength from a current frame, unless
>you forswear any changes -- if you take all possible load
>cases that a current frame can just barely endure (a 'surface'
>in a 'loading space' of about 3+3+3+6+6 dimensions), and demand
>that the new frame also endure them, then you have something equally
>strong but probably inefficient. If you take just a few points from
>the load 'surface', then you risk having the new design
>too weak for some untested loadings.
>
>Bikes are a big messcompared tosomething like a car (or even simpler,
and level-headed, but now you're prescribing years of supercomputer
time to design a frame --- what gives? Did you have a rough day???
The Critic
Jim Papadopoulos
>In article <CpCx2...@freenet.carleton.ca>, ap...@FreeNet.Carleton.CA (Antony Pringle) writes:
>|>
>|>
>|> But surely there must be some data out there which deals with
>|> the forces imparted to a frame subject to various inputs...
>
> Maybe they have simulated simple pedalling motion or even
>a sprint, but
>
> I think its nearly impossible to even fathom what is happening
and also talked with many in the industry, and very little
turned up.
Furthermore, what I have found is almost useless -- stresses not loads,
or loads without reaction forces, certain very special cases
(single rider, single maneuver, single 'style'). Perhaps someone
like HajoZ knows of better work done abroad .....
Now with simpler portable dat acq. systems, there is beginning
to be data on (say) handlebars, but it would be a whole lot more
useful if it was broken down into different components
(e.g. bump, sprint etc.) so we could project how they'd
differ for a lighter, stronger person (say).
I totally agree with Matt ("it's nearly impossible...") --
in fact I think it's worse than he imagines.
Have a nice day, everybody.
Jim
Jim Papadopoulos
>Well, we currently use a 150,000 particle SPH code as the resolution
>marginally sufficient to show the explicit fracture of a 6 cm diameter
>sphere of basalt. This is pretty much the state of the art for
>fracture, although I don't quite know where finite element (equilibrated
>stress) codes stand. Needless to say, a sphere of uniform rock is a
>far simpler system than a bicycle!
On the one hand, to be acurate about the tip of a crack
which is within a notch that is yielding plastically
in reversed structural loading requires great sophistication
and computational complexity. Anyone who thinks
otherwise is delusional.
On the other hand, many questions are NOT about such hard
problems, but rather simple things such as overall structural
elasticity or member moments for simple frame structures --
so for some purposes it's perfectly fine (indeed preferable)
to model a frame with 30-60 nodes (depending on butting, tapers,
etc.)
Once a frame model is performed, and the tubes are designed
(this is iterative becasue of statical indeterminacy),
then it makes sense to look closely at each joint in turn,
when it transmits maximal moments, and use several hundred nodes
(or even several thousand) for that joint alone.
The greatest difficulty I think comes when the structure
is loaded near failure and all kinds of wrinkling or crack
growth or reversed plastic flow occur, somtimes some specialised
software may approximate what happens, otherwise
the fea code will be hard-pressed to be accurate.
Just my opinion .....
Jim
Jim Papadopoulos
>Steady on, Jim old ch!!! You're normally so perspicacious
>and level-headed, but now you're prescribing years of supercomputer
>time to design a frame --- what gives? Did you have a rough day???
Perhaps I was a little hasty in my posting. First of all,
we can reduce the load-space dimension by 6, assuming
the bike frame itself is approximately in equilibrium.
Secondly, I must concede that if you build a 'standard' bicycle
to carry each of a few major loads safely, it will probably be
strong enough to survive other 'untested' loads, and maybe even
typical load _combinations_.
(But then, what do you have -- maybe just a completely normal
standard bike?!)
The fact still remains, that you need at least 5-10 load cases to make a
good stab at a design.
(Magnitudes, orientations, locations)
But whoever doesn't know these loads shouldn't feel too
bad, few if any manufacturers do either.
Thankfully, bikes are not designed purely by computer!
Jim P.
Jim Papadopoulos
>---In article bicyc...@delphi.com, Jim Papadopoulos <bicyc...@delphi.com> () writes:
>>Rob Horvatich,EVEO,, <rho...@eve054.ford.com> writes:
>>
>>>You are correct in posting that the loads need to be extracted out of the frame.
>>>This is usually a more advanced step in a design analysis process where a prototype can actually be built.
>>>But since you already have a frame in mind, why not
>>
>>
>>I'm afraid I don't really get your point about "why not".
>The original poster wanted to do FEA work on existing frames to determine information such as SF.
>He can input loads into his FEA model taken off of the same frame ridden in some sort of condition, (say off a big jump for maximum load).
>
initial posting somewhat and had wrongly assumed that it
was about bicycle design.
I wanted to discuss another (more minor) point by email,
but none of my email has ever gotten through to you. (Returned
from Ford for wrong host name or account name).
If you are willing to receive email please post or email a working
address to me.
Thanks, Jim
Rob Horvatich,EVEO,,
>In article <2qgljq$6...@eve120.cpd.ford.com> rho...@eve054.ford.com writes:
>>In article l...@organpipe.uug.arizona.edu, asp...@curly.lpl.arizona.edu (Erik Asphaug
>accuracy) the bending of an aluminum tube.
>
>Engineers at the Phillips lab (also in Albuquerque) use over 1,000,000
>nodes in an SPH code to simulate (again, with low accuracy) the impact
>of a spherical meteoroid into a communications satellite. I know this
>is inaccurate because you can observe unphysical instabilities in the
>simulation video.
>
>And paradoxically, the lower the strain rate, the longer the computation
>takes. Shocks (hitting your Cannondale with a bazooka blast) is a far
>simpler calculation than gradual stress (riding over a bump, or simply
>applying a downstroke to the pedal).
>
>Fast codes for engineering use linear elements for segments of the
>frame, but if you want to model the bending of the frame, and the
>interplay between frame joints, you must resolve it at fairly high
>resolution. Remember that in 3D, 10,000,000 nodes is only 300 nodes
>along each linear dimension of the bike.
Hey Erik
Sounds like you are talking about different types of analysis than I was.
I was describing just a simple static analysis with maximum loads to measure distributed stress.
How we come up with an acceptable maximum loads is a different story.
You are talking about simulations of nonlinear dynamic events (impact loads) and fracture mechanics.
I don't see the usefullness of either in the prediction of stress distributions in a frame with the loading events it may be subjected to.
Maybe you can help me out here?
I have seen aluminum column crush simulations with as little as 5,000 elements with very good correlation to actual tests.
Sounds like the code/model you are talking about is not fully representative of the structures behavior.
Were those poor results made more accurate by simply increasing the number of elements?
I find this interesting because aluminum behavior is far from mysterious and I have seen it accurately modeled before (linear, nonlinear analysis).
I admit I am not familiar with SPH and your use/definition of nodes.
I am talking about first modeling the frame crudely by beam elements.
More detailed analysis would include linear quad shell elements to define the tube structure.
I would estimate a detailed analysis with an element size of approx 5-10mm (small) in high stress areas would give acceptable results.
The model size would be approximately 15,000-20,000 elements.
This would mean the number of nodes is approximately the same as elements since there are not a lot of free edges.
Maybe you can see why I questioned the significant differences in figures.
Any comments?
Have a good day........Rob
Rob Horvatich,EVEO,,
>Secondly, I must concede that if you build a 'standard' bicycle
>to carry each of a few major loads safely, it will probably be
>strong enough to survive other 'untested' loads, and maybe even
>typical load _combinations_.
>
Direct hit! Hole in one!
I couldn't (didn't) have said it better myself.
What you described is exactly what most FEA includes, or should I say excludes.
You can not practically understand everything that is going on in a structure.
Even a simple beam bending problem has its share of complicated details if you want to go that far with it.
For example, working on a micro scale with putting into consideration heat generation from internal friction.
These effect are not considerations for most engineers.
For all practical purposes much of those details aren't needed for most considerations.
That is why us engineers throw in that safety factor, or is it fudge factor, to take care of the rest of the unknown that might do undesireable things to our designs.
With a lightweight bicycle frame that has minimized material usage, there is much less room for error than in other areas less weight consious.
>(But then, what do you have -- maybe just a completely normal
>standard bike?!)
An ideal design for a bicycle frame unfortunately may not lend itself to be manufacture at a cost effective level.
Other factors involved, like manufacturing, that effect the final design and may have little to do with the initial design criteria.
>But whoever doesn't know these loads shouldn't feel too
>bad, few if any manufacturers do either.
Sometimes we know them, sometimes we don't.
Sometimes we just we don't care to work with them because the final effects are minimal and not worth the extra expense to include them.
This is only possible if you can afford not to include them.
Secondly you still have to validate what the computer tells you.
>Thankfully, bikes are not designed purely by computer!
I don't know of anything that is.
Thak goodness
Have a good day all............Rob
Erik Asphaug x2773
>>Engineers at Sandia use about 30,000 nodes to model (with fairly low
>>accuracy) the bending of an aluminum tube.
Maybe you can see why I questioned the significant differences in figures.
Any comments?
Have a good day........Rob
Hi Rob,
I think you're right that in the case of nondynamic stress loads there
are much more efficient and accurate codes out there, so the 10^7 node
estimate is probably overly pessimistic... unless you want to look at
the overall onset of material failure in any specific way.
The "can crushing" code uses mesh elements confined to a planar lattice,
and thus fails at the onset of crumpling. The "crumpling" (i.e. failure)
that occurs is an artifact of code instability, not physics. I believe
you can model the cans pretty accurately using a few thousand nodes
if you're only looking at gradual bending motion within the elastic
regime.
In general though, I would use detailed analyses to gain a general
understanding of the behavior of critical bike elements under a wide
variety of stress/strain, and then perhaps model the relatively simple
regions (tube seqments) in a more efficient manner. But by my experience,
such "composite" numerical models can be a nightmare, with difficulties
similar to the butting of carbon fiber to aluminum!
Cheers,
Erik
Rob Horvatich,EVEO,,
Jim,
I have difficulty with this mail thing too.
I tried emailing you but don't know if it was delivered.
Sometimes it returns to sender even if it get through.
Maybe I forgot to lick the stamp?
Antony Pringle
Why all the talk about failure (crushing, crumpling, buckling
etc.)???? Surely, a 'bicycle frame designer' (if there really
is such a beast) would not care about modelling the failure
of the design.... we are talking about a bicycle here,
not a car or a school bus!
I would assume that the only deformation which would have to
be considered would be ELLASTIC since a bicycle should be
designed such that no plastic deformation occured in "normal"
use.
It would not take 10^7 nodes to model a frame! Remember that
a simple two-noded FE beam models an isotropic beam
EXACTLY. Assuming isotropy can be assumed (and I think that
it would be a resonable assumption for a metallic frame), a
frame could likely be modelled using beam elements. Beam
FE results (displacements) could then be used as input to
more accurate 3D solid models (20-noded isoparametric
quad elements) of the lugs/welds/joints.
Of course, that brings us to an old problem, what loads
would you use in the boundary conditions......?
And what loading modes should be considered (ie. sprint
'mode', climbing in the saddle 'mode', etc) to ensure
that all load extremes are considered??
Raymond Willemann
This discussion seems to have revolved around whether or not a complete
analysis of failure is possible. Has any of the analysis that's been
done in the past produced suggestions for how to select a frame? For
example, should a tall rider really look for a taller frame? Or does
a smaller frame actually sufficiently stronger to justify simply using
a taller seat post and stem. Or, what do the stress analyses say about
the importance of using the shortest possible rear stays in a racing
frame. Has anything useful come from the analyses?
Ray
Jim Papadopoulos
>For all practical purposes much of those details aren't needed for most considerations.
>That is why us engineers throw in that safety factor, or is it fudge factor, to take care of the rest of the unknown that might do undesireable things to our designs.
>
>With a lightweight bicycle frame that has minimized material usage, there is much less room for error than in other areas less weight consious.
>
>
innovation in bike frames, as opposed to duplication, requires
more knowledge than the industry currently possesses (*as far as I can
tell*). (That is, if you want to make a good stab with a computer-
based design.)
What I mean is, it you look at sprinting loads, you will
design a bike that is all down tube. If you look at seated
grind-it-out pedalling, a nice cantilevered seat tube will do
the tick. To deal with substantial front braking, you may
increase bending strength of the forks and head area.
But now you have designed for these threee major loadings, will
the frame survive all others?? (Rear wheel impact, standing
and leaning the frame, swerving rapidly etc.???) My belief is, that
if one had done a great job with the first three load cases, making
a super light frame which could safely handle them all, then
it might well be absurdly weak when loaded differently (like by a rock
that bounces up and dents a tube).
Thus, anyone who is tempted to strive for real structural efficiency,
had better have a lot more knowledge of loading!!
Maybe that's what you were saying but I didn't grasp it???
Apologies if so.
Jim
Jim Papadopoulos
>etc.)???? Surely, a 'bicycle frame designer' (if there really
>is such a beast) would not care about modelling the failure
>of the design.... we are talking about a bicycle here,
>not a car or a school bus!
>
>I would assume that the only deformation which would have to
>be considered would be ELLASTIC since a bicycle should be
>designed such that no plastic deformation occured in "normal"
>use.
Light frames do sometimes fatigue. If residual stress and notches
are not the problem, I think we have to conclude that
low-cycle fatigue is the culprit ( because, for every thoussand or
ten thousand modest loads, there is one twice or thrice the size --
loads like that have a disproportionate effect on fatigue.)
Well, low cycle fatigue is pretty close to plastic flow, or
reversed plastic flow etc.
That is to say, most of the time bikes should be perfectly
elastic (away from sharp notches which we hope are absent).
But their ultimate survival appears to me
to relate to how they handle those occasional much larger loads
(bumps, hard startups, panic braking, swerves, even crashes).
We know that abusive riders break many more frames. I maintain that
this is telling us we are on the edge in terms of elastic behaviour,
and improving endurance while reducing weight will not involve
us in the purely elastic again.
(Or, we can lower stresses by going to beer-can tubes, and then
deal with all kinds of buckling concerns in those high-load
situations.)
Does this address your concerns, I'd be interested in your response.
Jim
bui tho xuan
>...
>innovation in bike frames, as opposed to duplication, requires
>more knowledge than the industry currently possesses (*as far as I can
>tell*). (That is, if you want to make a good stab with a computer-
>based design.)
>Thus, anyone who is tempted to strive for real structural efficiency,
>had better have a lot more knowledge of loading!!
And don't forget to test until failure after we're all done
with this brand new, shiny design. A great deal of material
data out there, specifically, fatigue data, are measured for
comparison purpose. By this I mean metallurgist would measure
fatigue life in identical repeating conditions for different
materials and then compare them.
Since the loading history of a bike is anything but constant,
it's unlikely that you can obtain fatigue data that will exactly
match it. Result? You still have to test to failure. The
aerospace industry probably has the best design capability out
there, but guess what they do with critical components? They test
'em 'til they break.
tho
Antony Pringle
We seem to be discussing a topic which, IMHO, could be divided
into to very distinct subjects, ie: STRESS analysis and
FAILURE analysis.
Stress analysis is where FEA shines. It can be used to
accurately predict the distribution of stresses in a
frame, assuming that the input loads are known (which we
seem to have decided is something quite unknown). It
is primarily used to model an elastic response. Although
many FEA codes can accurately predict inelastic/buckling/
large deformation behaviour, it would not be particularly
useful, since the purpose of stress analysis would likely
be to give a designer 'the big picture' and assist him/her
in building a stiffer/lighter bike.
Failure analysis (for a bicycle) must consider both fast
failures and fatigue. Fast failures (buckling, shearing,
etc.) could be accurately modelled using various
theoretical or numerical (FEA) techniques, but wouldn't
have to be too seriously considered by a designer
since it would be impossible to consider all possible
modes of failure (and besides, a bike should be designed
to provided a certain strength/stiffness/'feel' to weight
ratio, not to survive a collision with an 18 wheeler, say).
A more interesting aspect of failure analysis is crack
growth analysis. There exist, in aerospace, many very
applicable models of crack propagation and crack
failure prediction.
My recipe for a comprehensive stress and failure analysis
of a bicycle frame (ie. DREAM THESIS) would include:
- instrumenting a frame with strain gauges
- let someone ride it for a few months, over various terrain
and obtain a complete load distribution and time spectrum
- use the load data obtained to perform in-depth stress
analysis of the frame, using FEA, for those load profiles
which are considered critical; construct a test jig to
apply critical loads to bike and check FEA results
with strain gauge and photoelastic results
- re-design the frame using the stress data obtained
- using the time load spectrum, perform lab testing of
the frame, and consult existing failure data to
predict the useful life of the frame
- ride the frame 'til it breaks!!
Jim Papadopoulos
>We seem to be discussing a topic which, IMHO, could be divided
>into to very distinct subjects, ie: STRESS analysis and
>FAILURE analysis.
>
>Stress analysis is where FEA shines. It can be used to
>accurately predict the distribution of stresses in a
>frame, assuming that the input loads are known (which we
>seem to have decided is something quite unknown). It
>is primarily used to model an elastic response. Although
>many FEA codes can accurately predict inelastic/buckling/
>large deformation behaviour, it would not be particularly
>useful, since the purpose of stress analysis would likely
>be to give a designer 'the big picture' and assist him/her
>in building a stiffer/lighter bike.
>
>Failure analysis (for a bicycle) must consider both fast
>failures and fatigue. Fast failures (buckling, shearing,
>etc.) could be accurately modelled using various
>theoretical or numerical (FEA) techniques, but wouldn't
>have to be too seriously considered by a designer
and the tools available to address them. But still I feel that
we are disagreeing on something substantial, whether due to
my error or yours.
How to phrase this ......
If you look at a fairly simple part like a handlebar or stem, I
think if you or I 'design' one to compete in today's marketplace,
we will find that some of the loads it sees cause a small amount
of yielding. If we used FEA to redesign to remain elastic
under known loads, the stem will weigh some or a lot more than
the competition!!! So realistic design to remain elastic
to me does not look like an option. [I am not going into
the value of lower weight; let's just say the market demands it.]
But you may say, design it to remain elastic at 80% of the peak load
measured on the trail!
But here's the thing, I believe that depending on details of the
nonlinear material behaviour, and aspects of how this yielding is confined,
and how much springback occurs, etc. the lifetime for this kind of loading
could be dramatically affected. With care and expertise, I imagine
FEA can help in this situation --- but it won't be pretty.
Bottom line, I absolutely agree with your suggestion to
do some linear-elastic beam modelling to get started, and if that's
all I can do it's still not to be sneezed at. But for lighter-weight
frames, for investigating various strategies for structural efficiency,
all kinds of nasty phenomena will come into play. I intentionally bought
a nonlinear code (Ansys) just so I could explore those kinds of
problems at my desk, so I wouldn't have to go through the hassle
of making a weird part only to find that it buckles and
loses strength unacceptably when meeting that rare overload.
Do we have an argument? Or merely an issue of semantics?
I do welcome further input, because struggling with exactly these
issues is now the purpose of my business (BICYCLE R&D).
Jim
Hans-Joachim Zierke
bicyc...@delphi.com (Jim Papadopoulos) wrote:
> I guess I didn't quote enough, but never mind. My main point is that
> innovation in bike frames, as opposed to duplication, requires
> more knowledge than the industry currently possesses (*as far as I can
> tell*). (That is, if you want to make a good stab with a computer-
> based design.)
All this is currently tried. While the current "carbon" frames are horribly
expensive to manufacture, it is tried to use Thermoplast (don't know
wether the word is the same in English) in pre-fabricated sheets for the
manufacturing process. This is no longer garage work, the main player in
the game is DuPont.
Some people in the automotive industry seem to like using bicycles as a
playing field for this manufacturing technology. And I'm quite sure that
the FEA capacity of BMW is a little better than the typical equipment of
the bicycle industry.
All this still does not mean success, though. I think we'll get knowledge
in 1997 or so.
hajo
Hans-Joachim Zierke
ap...@FreeNet.Carleton.CA (Antony Pringle) wrote:
> My recipe for a comprehensive stress and failure analysis
> of a bicycle frame (ie. DREAM THESIS) would include:
> - instrumenting a frame with strain gauges
> - let someone ride it for a few months, over various terrain
> and obtain a complete load distribution and time spectrum
> - use the load data obtained to perform in-depth stress
> analysis of the frame, using FEA, for those load profiles
> which are considered critical; construct a test jig to
> apply critical loads to bike and check FEA results
> with strain gauge and photoelastic results
All this was done by the RWTH Aachen around 1985-1988. They used the data for
construction of test equipment they sold to the industry.
The approach is the following: Construct a test device out of this data
that apllies a little more dynamic load than found in real life and be
safe.
Most bicycle frames will brake on their test devices.
An ITM titanium stem will brake on their test device for handlebars. A
standard Cinelli stem won't.
hajo
Jim Papadopoulos
>But surely there must be some data out there which deals with
>the forces imparted to a frame subject to various inputs...
>
>which had the FEA of a bike frame? I remember reading it
Over the years there have been quite a few attempts to measure
SOME loads or SOME stresses.
100 years ago at least one investigator made some 'force-measuring
pedals', which recorded pedal force on a strip of graph paper.
Moulton measured crank torque, with a signal line from a recorder
while the cyclist rode in circles around it.
I have seen pictures (in Columbus ads??) of some italian
researchers, a thick bundle of cables from the ridden bike
to a chase car.
I think the people who developed the Teledyne Titan in the
early seventies did some strain gage work.
At MIT a student plastered several strain gages on a frame
and converted the outputs to audio frequencies so several
could be recorded on a cassete tape -- then sprinted
down the hall
There is an academic paper by Soden and A... , wherein a rider
pedals up a steep ramp
Maury Hull and students (at UC Davis) have published many
papers, involving high-quality data from instrumented
pedals, seat posts, etc. However the riding was (always?)
done indoors on a stationary or treadmill-mounted bike, and there
were never enough measurements to determine how the
bike was loaded (i.e. where the reactions were).
(By the way, don't trust the INTERPRETATIONS or projections
the authors make from their data.)
Bicycling magazine got portable data acquisition for measuring
handlebar loads on the trail, and published one article about
them.
Now they have sponsored some research (At Virginia Polytechnic???)
in which handlebar loads are again measured.
The folks at OFFROAD/Proflex measured the forces
when a bicycle is riddent over an instrumented bump.
[These items, all off the top of my head, are about half the cases
I've heard about, which in turn are probably about 1/3 the
cases that have ever come to pass, I bet.
As I search my memory, I'm not aware of anyone having
studied the full set of loadings, i.e. all forces on the bike,
in a number of major loading dituations.
Probably I would recommend talking to Hull if you want to
know about top-quality load instrumentation and partial
data -- (916)752-6220 -- feel free to tell him I sent you.]
According to my recollection, the Bicycling cover was between
1985 and 1988. If I get teh exact date I'll post it.
Good luck with your search -- if you learn anything I'd be interested in
hearing it, and if you want ot work to dig up some of
the studies I've mentioned I should be able to give you
more specific information.
Jim
Jim Papadopoulos
>expensive to manufacture, it is tried to use Thermoplast (don't know
>wether the word is the same in English) in pre-fabricated sheets for the
>manufacturing process. This is no longer garage work, the main player in
>the game is DuPont.
>
>Some people in the automotive industry seem to like using bicycles as a
>playing field for this manufacturing technology. And I'm quite sure that
>the FEA capacity of BMW is a little better than the typical equipment of
>the bicycle industry.
(SMC), a sheet of random fibers and uncured matrix which can be deformed
in a press like sheet metal, then cured for strength??
FEA CAPACITY: At least in the U.S. and British cycle industry, my
opinion is that CAPACITY was never the issue -- rather there has
been little or no effort to be rational about loads and
stresses -- to perfrom engineering, if you will. Conversely,
even if a company gets a supercomputer and a full-blown
nonlinear fea code, it guaranteees nothing about what they'll
learn. When I spoke at length to 'engineers' at many
bicycle companies a few years ago, the most important use they
could think of for fea was generating impressive high-tech
pictures. (This was not necessarily their preference, I should
add.)
If you (or anyone else) knows of any sensible work delineating
loads and performing reasonable structural analyses, I'd
be very interested to hear about it.
The best effort I can recall (and I'm not saying it was all
that thorough) was outlined in a Bike Tech article --
structural analysis by Pegasus Research for (I think) Trek.
Your projection of reasonable structural engineering in 1997
makes me happy - I hope it comes to pass!
Hans-Joachim Zierke
bicyc...@delphi.com (Jim Papadopoulos) wrote:
> THERMOPLAST: I wonder if that's the same as Sheet Molding Compound
> (SMC), a sheet of random fibers and uncured matrix which can be deformed
> in a press like sheet metal, then cured for strength??
I finally found my DuPont stuff. (Tidyness is for whimps, a genius is in
control of the chaos.)
DuPont TEPEX (Thermoplastic Engineered Preforms)
DuPont MCS (Moldable Composite Sheets)
The DuPont USA people in Newark might be happy to help you.
Dr. Sturgeon (302) - 733 - 8878
> The best effort I can recall (and I'm not saying it was all
> that thorough) was outlined in a Bike Tech article --
> structural analysis by Pegasus Research for (I think) Trek.
The best effort I can recall was financed by Mannesmann Roehrenwerke and
carried out by the RWTH Aachen.
There was an old, qualified senior engineer, Mr. Suermann, at Mannesmann.
He is interested in bikes and used this data to sell Mannesmann tubes to
the industry.
For Mannesmann, the bike industry is peanuts. After the retirement of
Mr. Suermann, I did not hear a lot. They still sell tubes via Oria in
Italy.
> Your projection of reasonable structural engineering in 1997
> makes me happy - I hope it comes to pass!
My projection is that we will see some output in production runs. I won't
judge about the engineering before I can ride that output.
You will also misunderstanding me to be happy about it. _If_ this
technology works, this will be the end of the small makers, and there will
be a few automated production plants in the world to make that frames.
Exactly this explains the nervousness in the industry.
After ten or twenty years of this being hip, people will buy steel frames
from framebuilders as a luxury object like mechanic watches from Switzerland
today.
hajo