3 ms·
It's because most mechanical engineering curricula, and as a consequence most mechanical engineers, overwhelmingly treat empirical distributions as point estima
by peepeepoopoo112 2y ago
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.
- foxglacier 2y agoDo 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?
- peepeepoopoo112 2y agoBecause 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.
- foxglacier 2y agoThat statistics would still point to some factor of safety. You're saying it would be lower than what's currently used but would it really? FoS allows for other things besides material behavior, such as uncertainty in loads and construction.