It's been surprising to me how poorly understood this field is. Even something as popular a the von Mises criterion seems to have got its popularity because Tresca was a more "pure" result of the supposed failure mechanisms but it needed rounded corners to fit dirty real world materials and von Mises has rounded corners, so it'll be good enough.
I suppose it doesn't matter to most engineers because nobody's going to design with a safety factor of 1.1 or even know the loads anywhere close to that accuracy. Also nobody needs to exploit the greater compressive strength vs tensile of most materials so no need for those asymmetric models.
It's amazing to me how many people working in artificially simplified fields look at real world mechanics and think it must be simple. Only the models are simple, and they are still impossible to solve analytically, with tensor projection mechanics as freshman methods. FEM is used on anything beyond a toy model and everyone building things knows that the ability to obtain and assemble materials is a far bigger constraint than measuring any ideal material property. Carbon fiber composites and single crystal nickel rotor blades are actually convenient simplifications. Properly designed they allow for more ideal structures and lower the dimensionality.
To give an example. TI built a multi $B business on DMD micromirror based projection systems. The millions of mirror hinges which oscillate at several kHz in each projector are built from thin film Aluminum under several % strain. They were most of a decade in and most of $B of investment before it was shown that the hinges happened to be thin enough to be single crystal in the high strain region allowing them to operate. Aluminum fatigue models weren't something the EEs were worried about, and they happened to be right!
Complexity is built on abstraction. Some things are VERY difficult to abstract. Solving coupled electro-thermo-mechanical equations for heterogeneous 3D objects is one of them.
The mechanics of failure are both very well understood in some domains and applications whilst other fields languish.
Various aviation materials have 10 of thousands of papers studying fatigue failure. The area of my PhD research program was in biomechanics working on knee meniscus as a fibrous composite material.
In the entire field of biomechanics and soft tissue material science there were 2 papers that did fatigue experiments in tendons that I ever found.
I figured out how to improve specimen preparation techniques to be able to perform experiments for a third paper (at the time) on soft tissue fatigue and the first ever on fibrocartilage. It was fun.
Yet somehow medical devices are made to fix spinal discs and torn meniscus were surgically removed for decades, all without very rudimentary engineering data on the materials. Then again, visiting a spinal clinic with my father you can see the results of that with how little we can really treat.
my wife had experimental spinal intervertebral disc replacement surgery 20y ago. Nothing was "fixed" they just replaced the soft tissue with something that resembles a polyurethane o-ring. Turns out the USPHS didn't believe in this procedure, so they just stopped doing it.
I had 3 courses of material science at university (30y ago) and back then the theoretical models where nowhere: being within 2 orders of magnitude of reality was already a success. Engineers would just build a big formula with plenty of constants that allowed the formula to fit reality.
Last year, my son had a course of material science at university and beforehand he did not believe me about how clueless we humans used to be back then. Turns out nothing has changed.
> It's been surprising to me how poorly understood this field is. Even something as popular a the von Mises criterion seems to have got its popularity because Tresca was a more "pure" result of the supposed failure mechanisms but it needed rounded corners to fit dirty real world materials and von Mises has rounded corners, so it'll be good enough.
I don't think this is an educated take. Tresca and von Mises are yield criteria that match experimental observations on controlled environments. Fatigue is not the same thing as yield failure. Fatigue can culminate in failure, but it is due to propagation of damage throughout time caused by repeated loads. Damage can assume multiple forms. This means that the structural system actually changes throughout time, and stresses propagate differently and concentrate on both places where damage forms and propagates and places where load is redistributed.
Fatigue is highly dependent on load history and manufacturing techniques/quality control. There are a few classes of fatigue such as low-cycle fatigue and high-cycle fatigue. You cannot model this with an equation. The best you can do is spread sensors around to monitor load history and make educated guesses on when specific parts of a structural system is reaching the end of its service life.
To add to the complexity, fatigue is also dependent on the microstructure of a structural material. The concept of very high cycle fatigue is relatively new. We had decades of engineering products being developed targeting high-cycle fatigue limits hoping it to be very robust or even eternal. Except we started to see structural systems in the real world, such as airframes, failing due to fatigue when they had a service life of very low loads. Some airframes designed to be practically eternal, such as the C-130, started to show weird fatigue patterns when subjected to neglectible load cycles. It turned out that the metal crystals were causing stress concentrations that caused microcracks even at very low loads. This was not known to even be a possibility, until things started to crack with no apparent reason, and engineers started troubleshooting.
But engineers still manage to do wizardry with the limits of what we know.
If you want to look at a domain where close to nothing is known, do not look at fatigue or yield criteria. Look at damping. The whole concept of damping is created by engineers hand-waving over the fact that energy leaves the system. The techniques in place have more to do with the convenience of mathematical models than with the actual phenomena. The main class of theological models is basically a linear model that's fitted to observations. There are still PhD thesis being done on how to monitor damping in specific problem settings.
Do Tresca or von Mises really fit anything well? I've only ever seen data that looks like this [1] which is all over the place. Can we even control shear stress accurately in a lab? Tensile and compression tests are easy but shear stress tends to be non-uniform through a sample. Then there's the mystery (afaik) world of uniform tri-axial tension which seems to be impossible to measure.
I wasn't talking about fatigue, but damping too for sure. Analysts often use Rayleigh damping even though it makes no sense physically, uses parameters that aren't material properties, and just happens to be a convenient formula that works with some solution techniques.
It's because most mechanical engineering curricula, and as a consequence most mechanical engineers, overwhelmingly treat empirical distributions as point estimates. It's a bad habit that we inherited from classical physics, where it makes more sense to treat physical constants as exact point estimates, versus the very empirical and probabilistic world of engineering.
Do you mean a quantity like tensile yield strength should be thought of as a distribution, not a value with an uncertainty? It's hard for me to really see the difference. A real part will fail somewhere which will have some error from prediction and what difference does it make why there's error as long as you can quantify it?
Because the proper way to quantify it is to represent it as a distribution. Right now a bunch of highly approximated heuristics, like factor of safety, are used in place of proper statistical practice. We would have much safer and higher performing engineering designs if we were to properly model the material parameters in our calculations using probabilistic methods.
I read this as “Failure of Theory for Material Science and Engineering” and had to laugh.
So much of this field involves experimentation and heuristics. As real life is really complicated to model.
It's been surprising to me how poorly understood this field is. Even something as popular a the von Mises criterion seems to have got its popularity because Tresca was a more "pure" result of the supposed failure mechanisms but it needed rounded corners to fit dirty real world materials and von Mises has rounded corners, so it'll be good enough.
I suppose it doesn't matter to most engineers because nobody's going to design with a safety factor of 1.1 or even know the loads anywhere close to that accuracy. Also nobody needs to exploit the greater compressive strength vs tensile of most materials so no need for those asymmetric models.
It's amazing to me how many people working in artificially simplified fields look at real world mechanics and think it must be simple. Only the models are simple, and they are still impossible to solve analytically, with tensor projection mechanics as freshman methods. FEM is used on anything beyond a toy model and everyone building things knows that the ability to obtain and assemble materials is a far bigger constraint than measuring any ideal material property. Carbon fiber composites and single crystal nickel rotor blades are actually convenient simplifications. Properly designed they allow for more ideal structures and lower the dimensionality.
To give an example. TI built a multi $B business on DMD micromirror based projection systems. The millions of mirror hinges which oscillate at several kHz in each projector are built from thin film Aluminum under several % strain. They were most of a decade in and most of $B of investment before it was shown that the hinges happened to be thin enough to be single crystal in the high strain region allowing them to operate. Aluminum fatigue models weren't something the EEs were worried about, and they happened to be right!
Complexity is built on abstraction. Some things are VERY difficult to abstract. Solving coupled electro-thermo-mechanical equations for heterogeneous 3D objects is one of them.
The mechanics of failure are both very well understood in some domains and applications whilst other fields languish.
Various aviation materials have 10 of thousands of papers studying fatigue failure. The area of my PhD research program was in biomechanics working on knee meniscus as a fibrous composite material.
In the entire field of biomechanics and soft tissue material science there were 2 papers that did fatigue experiments in tendons that I ever found.
I figured out how to improve specimen preparation techniques to be able to perform experiments for a third paper (at the time) on soft tissue fatigue and the first ever on fibrocartilage. It was fun.
Yet somehow medical devices are made to fix spinal discs and torn meniscus were surgically removed for decades, all without very rudimentary engineering data on the materials. Then again, visiting a spinal clinic with my father you can see the results of that with how little we can really treat.
my wife had experimental spinal intervertebral disc replacement surgery 20y ago. Nothing was "fixed" they just replaced the soft tissue with something that resembles a polyurethane o-ring. Turns out the USPHS didn't believe in this procedure, so they just stopped doing it.
I had 3 courses of material science at university (30y ago) and back then the theoretical models where nowhere: being within 2 orders of magnitude of reality was already a success. Engineers would just build a big formula with plenty of constants that allowed the formula to fit reality.
Last year, my son had a course of material science at university and beforehand he did not believe me about how clueless we humans used to be back then. Turns out nothing has changed.
> It's been surprising to me how poorly understood this field is. Even something as popular a the von Mises criterion seems to have got its popularity because Tresca was a more "pure" result of the supposed failure mechanisms but it needed rounded corners to fit dirty real world materials and von Mises has rounded corners, so it'll be good enough.
I don't think this is an educated take. Tresca and von Mises are yield criteria that match experimental observations on controlled environments. Fatigue is not the same thing as yield failure. Fatigue can culminate in failure, but it is due to propagation of damage throughout time caused by repeated loads. Damage can assume multiple forms. This means that the structural system actually changes throughout time, and stresses propagate differently and concentrate on both places where damage forms and propagates and places where load is redistributed.
Fatigue is highly dependent on load history and manufacturing techniques/quality control. There are a few classes of fatigue such as low-cycle fatigue and high-cycle fatigue. You cannot model this with an equation. The best you can do is spread sensors around to monitor load history and make educated guesses on when specific parts of a structural system is reaching the end of its service life.
To add to the complexity, fatigue is also dependent on the microstructure of a structural material. The concept of very high cycle fatigue is relatively new. We had decades of engineering products being developed targeting high-cycle fatigue limits hoping it to be very robust or even eternal. Except we started to see structural systems in the real world, such as airframes, failing due to fatigue when they had a service life of very low loads. Some airframes designed to be practically eternal, such as the C-130, started to show weird fatigue patterns when subjected to neglectible load cycles. It turned out that the metal crystals were causing stress concentrations that caused microcracks even at very low loads. This was not known to even be a possibility, until things started to crack with no apparent reason, and engineers started troubleshooting.
But engineers still manage to do wizardry with the limits of what we know.
If you want to look at a domain where close to nothing is known, do not look at fatigue or yield criteria. Look at damping. The whole concept of damping is created by engineers hand-waving over the fact that energy leaves the system. The techniques in place have more to do with the convenience of mathematical models than with the actual phenomena. The main class of theological models is basically a linear model that's fitted to observations. There are still PhD thesis being done on how to monitor damping in specific problem settings.
Do Tresca or von Mises really fit anything well? I've only ever seen data that looks like this [1] which is all over the place. Can we even control shear stress accurately in a lab? Tensile and compression tests are easy but shear stress tends to be non-uniform through a sample. Then there's the mystery (afaik) world of uniform tri-axial tension which seems to be impossible to measure.
I wasn't talking about fatigue, but damping too for sure. Analysts often use Rayleigh damping even though it makes no sense physically, uses parameters that aren't material properties, and just happens to be a convenient formula that works with some solution techniques.
[1] http://www.w.continuummechanics.org/images/vonmises/Experime...
It's because most mechanical engineering curricula, and as a consequence most mechanical engineers, overwhelmingly treat empirical distributions as point estimates. It's a bad habit that we inherited from classical physics, where it makes more sense to treat physical constants as exact point estimates, versus the very empirical and probabilistic world of engineering.
Do you mean a quantity like tensile yield strength should be thought of as a distribution, not a value with an uncertainty? It's hard for me to really see the difference. A real part will fail somewhere which will have some error from prediction and what difference does it make why there's error as long as you can quantify it?
Because the proper way to quantify it is to represent it as a distribution. Right now a bunch of highly approximated heuristics, like factor of safety, are used in place of proper statistical practice. We would have much safer and higher performing engineering designs if we were to properly model the material parameters in our calculations using probabilistic methods.