Paul, All
This is the problem of clarity of objective. (the skills will follow as your questions here will be more surgical). 'Here's a conrod, what are the stresses?' is a poor question. It is just as bad as - 'here's an oil rig. Will it break?' or statements such as 'converge the conrod stresses' are not helpful. Where do we want converged stresses?
The single most important question to answer before you start is
What am I trying to find out?
Exercises such as this are difficult because the reader has no intuitive objective and the model has many potential directions. I will labour the point inexhaustively :
- Modal
- Do the linear assumptions for modal matter
- Fatigue
- Deflection envelope
- Load Time history for fatigue
- How big should the rounds be
- bending stresses in the main shaft
- Journal contact stresses
- Journal contact stresses at the other end
- Interaction with a crank
- Envelope of stresses at the round
- Envelope bending stresses
- Envelope shear stresses
- Stresses around lubrication slots/holes
- Effect of different constraint methods
- buckling (simple)
- buckling (real)
Each one of these questions requires the model be built in a different way.
No one model will answer all questions. Clarity of purpose drives the model design.
So back the system hard-wired measures. The global hard-wired measures are limited use and will lead you a merry dance. Use of the information they provide should be limited and then with care.
You have to be clear about what information you want from the model. This defines where you design the model to converge well. It will also define what and where you want to measure.
Without being too glib, everywhere else we only care about stiffness and ignore stresses making sure that the elements reaching sharp corners (for example) are far enough away (by mesh control) that they don't 'contaminate' the answer you want.
Look at the exercise again and decide what you want to know. Make a simple clear statement of what you want to know and have another go. I note that Steven's model still has sharp corners; what question was he trying to answer?
Example. if you want know what the stresses in the rounds at the big end are then you need rounds at the big end. You will also need a reasonable quality of mesh on/near the rounds (2-4 element edges for every 90deg of arc in the round cross section), you will also need user defined measures on the round. You could look at the convergence graph for the round's measures; they will be 'nice'. The rounds at the little end are irrelevant as is a fine mesh on the rest of the conrod. If you graph the hard-wired measure for stress it will still look like the graph in your first post (and will be reporting stresses at some sharp corner or other). The method of holding the model will be up for debate. You may need to try several to bound the answer. There is no right answer. There are bad answers.