Frame durability - testing with science

Discussion in 'Cycling Equipment' started by Andrew Price, Apr 13, 2003.

Thread Status:
Not open for further replies.
  1. Andrew Price

    Andrew Price Guest

    Would be interested what the ng thinks of the frame durability tests recorded at -

    http://www.efbe.de/etour109.htm

    Surprising, to a non engineer that the lighter frames did so well. Would be interested to know if
    those with the technical knowledge consider the tests valid - one point strikes me that may have
    been insufficient examples tested.

    Look forward to hearing , best Andrew
     
    Tags:


  2. Peter Cole

    Peter Cole Guest

    "Andrew Price" <[email protected]> wrote in message
    news:[email protected]...
    > Would be interested what the ng thinks of the frame durability tests recorded at -
    >
    > http://www.efbe.de/etour109.htm
    >
    >
    > Surprising, to a non engineer that the lighter frames did so well. Would be interested to know if
    > those with the technical knowledge consider the tests valid - one point strikes me that may have
    > been insufficient examples tested.

    The article has been discussed a few times here. One criticism is that the test was unrealistic in
    that it used a relatively few cycles of a too high load, which might generate misleading results.
    That point taken, it still resulted with frames failing in classic manners. Both of the lugged steel
    frames failed with cracks around the lower head tube lug, exactly the way my steel frame failed in
    real use. While the test may not have been perfect, I think it did generate meaningful results, I
    don't find the outcomes too surprising.
     
  3. Jon Isaacs

    Jon Isaacs Guest

    >Would be interested what the ng thinks of the frame durability tests recorded at -
    >
    >http://www.efbe.de/etour109.htm
    >
    >
    >Surprising, to a non engineer that the lighter frames did so well. Would be interested to know if
    >those with the technical knowledge consider the tests valid - one point strikes me that may have
    >been insufficient examples tested.
    >
    >Look forward to hearing , best Andrew

    THese tests have been discussed here several times. The biggest problem I see with these tests are
    the large loads and few number of loading cycles.

    The load starts at 270 lbs for 100,000 cycles and then is increased to about 290 lbs for the next
    100,000 cycles.

    Rarely would a frame seem 270lbs in a cyclic manner, even with a large and relatively strong rider.
    Then consider that the 200,000 cycles amounts to about 500-1000 miles of riding.

    These are supposed to be fatigue tests and it would seem that by using a large load one can perform
    the test more quickly.

    However this is not true with fatigue tests. Fatigue is something that happens over millions and
    millions of cycles and there is a definite relationship between the fatigue properties of a material
    and the stress or load. Some materials have a endurance limit which means if the stress/load is kept
    below there endurance limit, the fatigue crack does not grow. Steel and Titanium are such materials.

    This means that if you increase the load to a load that is unusual, you may move past the endurance
    limit and thus the fatigue crack will propagate.

    Other materials like Aluminum have no fatigue limit so they require different design techniques to
    deal with fatigue. In general they must be over built from a strengh standpoint to deal with the
    fatigue issues.

    When doing a fatigue test or other engineering test, it is important to understand the failure
    criteria and the failure mechanisms and make sure that the test conditions are a good approximation
    of the real world loading conditions so that the results are indeed valid for the real world.

    Most any bike will last 1000 miles regardless of how it is ridden. THe real question is whether it
    will last 5000 miles or 50,000 miles and that cannot be determined by tests such as this one.

    On the other hand, if you are interested in knowing if Darren Hill or Marty Nostein could ride these
    bikes in match sprint competition without failure, these tests might be appropriate.

    The other issue is that in most cases, frame failure is due to manufacturing problems rather than
    design problems, though in the case of some "stupid light" frames, they are designed to have a
    limited life span.

    jon isaacs
     
  4. Jon Isaacs

    Jon Isaacs Guest

    >The article has been discussed a few times here. One criticism is that the test was unrealistic in
    >that it used a relatively few cycles of a too high load, which might generate misleading results.
    >That point taken, it still resulted with frames failing in classic manners. Both of the lugged
    >steel frames failed with cracks around the lower head tube lug, exactly the way my steel frame
    >failed in real use. While the test may not have been perfect, I think it did generate meaningful
    >results, I don't find the outcomes too surprising.

    The fact that a frame suffers low cycle fatigue failure at a weak point is not surprising. However
    whether a frame would sufferer similar failure given use within normal bounds is not within the
    realm of this test to determine.

    This is more of a strength test that a fatigue test and it is fatigue rather than strength that
    causes frames to fail.

    jon isaacs
     
  5. Jon makes some good points regarding the test and part design. Like just about everything, though,
    compromises for whatever reason are sometimes made.

    It would be nice to have the luxury of time in order to complete millions of cycles during a test
    like this on a large number of samples- speaking from personal experience, though, this simply is
    not practical.

    In my development of a stem durabiltiy evaluation protocol, I used a progressive load schedule.
    The amount of cumulative damage (if _any_) that occured at the small loads just doesn't matter
    near the end of the load schedule. In the future, the starting load I will use will be increased
    in order to shorten the amount of time on the fixture (current protocol takes 10-14 days with it
    operating 24/7).

    The EFBE is a comparative test protocol (which is more than likely developed according to DIN
    standards) that sheds some light on the frame durability subject. I would rather have this
    information than no information at all.

    --
    =======================
    Kraig Willett RBT Promotional rates: www.biketechreview.com/rbt.htm
    =======================
     
  6. Jim Beam

    Jim Beam Guest

    > However this is not true with fatigue tests. Fatigue is something that happens over millions and
    > millions of cycles and there is a definite relationship between the fatigue properties of a
    > material and the stress or load. Some materials have a endurance limit which means if the
    > stress/load is kept below there endurance limit, the fatigue crack does not grow. Steel and
    > Titanium are such materials.

    Kinda sorta. Fatigue, by definition, is cyclic load. Whether that's 10 cycles or 10^10 cycles. True,
    very low cycle fatigue has a different failure regime than high cycle fatigue, but once you get over
    10^3, there's no dispute which regime dominates, and from then on, you're in well known territory.
    On that basis, these tests are relevant.

    Regarding endurance limits, a lot of carbon steels and Ti alloys exhibit endurance limits, but this
    rarely applies to the high-strength alloys because endurance, iirc, is due to strain aging. High
    alloy content tends to remove the strain aging mechanism. As modern frames are usually made of the
    high-strength alloys, it's too simplistic to state that steel & Ti won't fatigue with lower loads.

    Jim
     
  7. Peter

    Peter Guest

    Jon Isaacs wrote:

    > The load starts at 270 lbs for 100,000 cycles and then is increased to about 290 lbs for the next
    > 100,000 cycles.
    >
    > Rarely would a frame seem 270lbs in a cyclic manner, even with a large and relatively strong
    > rider. Then consider that the 200,000 cycles amounts to about 500-1000 miles of riding.

    Only if the test is supposed to model a bike frame eventually failing purely as a result of the
    forces imposed by steady pedaling. Frankly I'd expect any reasonable road frame to last almost
    forever if only ridden in such a manner (say for hour rides on a smooth track), even if used by a
    strong rider. And they should certainly last a lifetime under the lower pedalling forces that I can
    exert - nevertheless, I have had two frames fail in normal use.

    OTOH, a road frame in normal use will occasionally see forces in the large range used in the test
    when hitting bumps or when the rider momentarily exerts himself very strongly. These events may only
    happen once or twice a mile, so the 200,000 cycles of the test would correspond to 100,000 or more
    miles of riding.

    > These are supposed to be fatigue tests and it would seem that by using a large load one can
    > perform the test more quickly.
    >
    > However this is not true with fatigue tests. Fatigue is something that happens over millions and
    > millions of cycles and there is a definite relationship between the fatigue properties of a
    > material and the stress or load.

    Fatigue shows a strong relationship between the size of the stress and the number of cycles that
    can be endured before failure. http://www.asme.org/igti/resources/articles/fatigue2.html shows a
    sample fatigue curve - on this curve, an increase of only about 20% in the applied stress results
    in the number of cycles to failure decreasing by more than a factor of 10, and a doubling of the
    stress drops the number of cycles by over 1000. The result is that even a relatively rare stress
    will dominate in the creation of fatigue failure if it is significantly larger than the more normal
    stresses encountered. The rider who puts a normal pedaling load on the frame 250 times per mile
    would see fatigue failure just as soon from the bumps that he hits only once every four miles, if
    those bumps result in twice the stress load of a normal pedal stroke. I'd also note that fatigue
    tests are frequently carried out in the range of 10s of thousands of cycles up to millions of
    cycles as shown in the figure cited above. The frame test using 200,000 cycles is not atypical of
    industry testing.

    (BTW, I am currently doing real-world fatigue testing on a 60 cm Cannondale frame that is 14 years
    old. At about 6000 miles per year it's doing fine so far. As long as I don't hit more than one
    significant bump per mile it should last for many more years.)
     
  8. Danziggidy

    Danziggidy Guest

    Just to add my two cents,

    It may just be the anal retentive engineer in me but testing one frame only (which as far as I can
    tell is what this company did) tells us absolutely nothing. I could build a frame that would outlast
    any of those in the tests, but come production you'll have variance in the quality. That would be
    why those of us involved with quality assurance sample MANY products on a line before we draw
    conlusions. IMHO the method of loading is the least of the problems with this test.

    Jon

    "Andrew Price" <[email protected]> wrote in message
    news:[email protected]...
    > Would be interested what the ng thinks of the frame durability tests recorded at -
    >
    > http://www.efbe.de/etour109.htm
    >
    >
    > Surprising, to a non engineer that the lighter frames did so well. Would
    be
    > interested to know if those with the technical knowledge consider the
    tests
    > valid - one point strikes me that may have been insufficient examples tested.
    >
    > Look forward to hearing , best Andrew
     
  9. Peter

    Peter Guest

    Jon Isaacs wrote:
    >>The EFBE is a comparative test protocol (which is more than likely developed according to DIN
    >>standards) that sheds some light on the frame durability subject. I would rather have this
    >>information than no information at all.
    >
    >
    > One just has to be careful interpreting this data, IMHO.

    True.

    > It is often used to say that Aluminum frames are more fatigue resistant than steel frames when
    > what it really says is that aluminum frames must be stronger to achieve the same real world
    > fatigue life of a steel frame.

    Given that the test was of a limited number of frames of different construction methods and with
    only one sample of each frame, I don't see how any general conclusions can be drawn about properties
    of any frame material.

    I certainly don't see anything in the results that would make me draw the conclusion you do above,
    there are far too many variables and no controls.
     
  10. Jim Beam

    Jim Beam Guest

    > This page states the endurance limit of 4340 is about 450 MPa.

    This reinforces the point about low-alloy steel being unrepresentative does it not?

    > 3. Tests which use unrealistically large loads move to different parts of S-N curves which makes
    > comparison tricky and prediction difficult unless the actual fatigue life is "backed out" of
    > the tests using some modeling assumptions.

    The whole point of mapping an S-N surve is that it /does/ allow you to extrapolate. Reliably. Unless
    the material exhibits an endurance limit, most metals give a good approximation of a straight line
    when plotting S vs. log N.

    > One obviously cannot draw any real conclusions from these tests because the actual geometries and
    > therefore stresses are unknown.

    If the intent is to trash this test, the only basis on which to do it, in my opinion, is the small
    sample size. The testing methods however are sound.

    Jim.
     
  11. > I personally think that most frame failures are the result of
    manufacturing
    > defects or later damage.

    While it's probably true that "most" frame failures are the result of manufacturing defects or
    later damage, there's a trend lately towards building at the edge of what's possible with a
    given material, with the result being frames that fail simply because there's no margin of error
    in the design.

    Our headlong pursuit of ultimate lightness, ignoring that different materials have differing
    strengths and weaknesses, is idiotic. Just because you can safely build a carbon frame at under 2.5
    lbs doesn't mean that a benchmark has been set that requires all materials be used to create frames
    of similar weight, and yet that's the way things have been heading. It's little surprise that we see
    such frames fail.

    --Mike-- Chain Reaction Bicycles http://www.ChainReactionBicycles.com

    "Jon Isaacs" <[email protected]> wrote in message
    news:[email protected]...
    > >Only if the test is supposed to model a bike frame eventually failing purely as a result of the
    > >forces imposed by steady pedaling. Frankly I'd expect any reasonable road frame to last almost
    > >forever if only ridden in such a manner (say for hour rides on a smooth track), even if
    used
    > by a strong rider. And they should certainly last a lifetime under
    > >the lower pedalling forces that I can exert - nevertheless, I have had two frames fail in
    > >normal use.
    >
    > As far as I can see and remember, this test simulates fatigue failure from pedalling loads. So
    > that is really the scope of this test, how long will
    the
    > fame handle large pedalling loads.
    >
    > >OTOH, a road frame in normal use will occasionally see forces in the large range used in the test
    > >when hitting bumps or when the rider momentarily exerts himself very strongly. These events may
    > >only happen once or twice a mile, so the 200,000 cycles of the test would correspond to 100,000
    > >or more miles of riding.
    > >
    >
    > This test did not simulate bump loadings, so I suggest that without
    further
    > analysis, it is not possible to draw any conclusions about the fatigue
    life
    > from bumps.
    >
    > >The rider who puts a normal pedaling load on the frame 250 times per mile would see fatigue
    > >failure just as soon from the bumps that he hits only once every four miles, if those bumps
    > >result in twice the stress load of
    a
    > normal pedal stroke.
    >
    > I agree but IMMSMW, these tests did not address the loads imposed by
    anything
    > other than cyclic pedalling.
    >
    > I personally think that most frame failures are the result of
    manufacturing
    > defects or later damage.
    >
    > Jon Isaacs
     
  12. Peter

    Peter Guest

    Jon Isaacs wrote:
    >>Only if the test is supposed to model a bike frame eventually failing purely as a result of the
    >>forces imposed by steady pedaling. Frankly I'd expect any reasonable road frame to last almost
    >>forever if only ridden in such a manner (say for hour rides on a smooth track), even if used
    >
    > by a strong rider. And they should certainly last a lifetime under
    >
    >>the lower pedalling forces that I can exert - nevertheless, I have had two frames fail in
    >>normal use.
    >
    >
    > As far as I can see and remember, this test simulates fatigue failure from pedalling loads.

    The test setup imposed a set of 100,000 forces of ~270 lbs and 100,000 forces of ~290 lbs on the end
    of the crank. Whether this is from a normal pedal stroke, an unusually strong pedal stroke say for
    starting a sprint, or from a rider having the force of his pedal stroke amplified by hitting a bump
    doesn't matter as long as the result is this number of cycles with the tested forces. Given how
    quickly the number of 'cycles to failure' decreases with increasing force, I'd guess the dominant
    contributor to fatigue failure in real-world cycling would be the relatively small number of pedal
    strokes that have a much larger than average amount of force rather than the majority of pedal
    strokes with a normal amount of force. The test setup was designed to closely simulate the force at
    the start of a sprint with the rider standing, leaning the bike appreciably, and pushing very hard
    on the pedals for a few strokes. So it does simulate the heaviest loads that will be encountered on
    only a small fraction of the total number of pedal strokes. As such, both the forces and the number
    of cycles seem within the range that a frame may see during real riding use over a life of around
    100000 miles although it will obviously vary greatly depending on the rider and type of riding.

    > So that is really the scope of this test, how long will the fame handle large pedalling loads.
    >
    >
    >>OTOH, a road frame in normal use will occasionally see forces in the large range used in the test
    >>when hitting bumps or when the rider momentarily exerts himself very strongly. These events may
    >>only happen once or twice a mile, so the 200,000 cycles of the test would correspond to 100,000 or
    >>more miles of riding.
    >>
    >
    >
    > This test did not simulate bump loadings, so I suggest that without further analysis, it is not
    > possible to draw any conclusions about the fatigue life from bumps.
    >
    >
    >>The rider who puts a normal pedaling load on the frame 250 times per mile would see fatigue
    >>failure just as soon from the bumps that he hits only once every four miles, if those bumps result
    >>in twice the stress load of a
    >
    > normal pedal stroke.
    >
    > I agree but IMMSMW, these tests did not address the loads imposed by anything other than cyclic
    > pedalling.
    >
    > I personally think that most frame failures are the result of manufacturing defects or
    > later damage.
    >
    > Jon Isaacs
     
Loading...
Thread Status:
Not open for further replies.
Loading...