Originally Posted by Gravy
Jay is just another of those people who thinks engineering is easy stuff.
They don't realize that reality can be approximated very well linearly. You build a math model of your device, following all the known laws of physics. you see how it reacts to various inputs. If you have some uncertainty, you build a real, hardware model of part of the device. you subject your math model of that chunk of it to varying conditions, load, temperature, and boundary.
You then run those conditions on the real hardware. You adjust your math model to fit reality (not the other way around--we use NASTRAN, PATRAN,I-Deas, ALGOR, ANSYS and the like. I think its TWoo-ver5.1 that adjusts reality to the wishful thought process)
You do not have to test to failure; You do not have to test the entire structure--the world prior to failure is nice and linear. When you're trying to determine how you get to failure, you are in the linear relm.
Just as you don't have to plummet to the ground to verify that you don't keep accelerating when skydiving, you do not have to test to failure-or even to full load/temperature environments, to verify the validity of your model
Any engineer knows this. Wannabes will never understand it.