Properly increment size

I want to ask, how to properly choose increment size in nonlinear, implicit study?
During analyze I get an error and I have no idea how to fix this problem.

You can try adjusting incrementation settings (reduce the initial and minimum increment, increase the maximum number of increments) but it’s likely that it won’t help with convergence problems. To handle them you should take a closer look at your model. If there are no errors such as wrong values or inconsistent units, consider changing the modeling approach. Especially if contact is involved in this analysis, you may have to use some tricks to aid convergence (displacement control instead of force control, stabilization with springs and so on). If you want us to help you with that, please provide the details regarding this simulation and possibly share the files.

The simulation is crankshaft rotation about 6.28rad with gravity affected. This is my input deck:

it’s nothing special but my best results are rotate maybe 3/4 of 6.28rad. I have no idea what is a reason of decrease increment time.
At the picture below we can see when increment decrese:

and finally increment achieve value 100 (max increments) and my study is failed at this moment. Increase max increments value is not a good idea in this case imo.

If it’s not a problem, please share the whole input file so that we can take a look at the complete setup of this analysis and run some tests. If you can’t do it, maybe screenshots showing the model with mesh and load/BC symbols will suffice.

I would add plasticity to this model if there is a chance of yielding. And when it comes to incrementation, we use these settings as standard for the first trials with nonlinear models:

0.1, 1, 1e-7, 1

But, as I’ve mentioned above, adjustment of the incrementation settings rarely helps obtain a converged solution, unless the problems start right at the beginning of the analysis.

You could also try with another solver, different than PaStiX.

I have experience with no convergence during not correctly material model but in my simulation the reason is different.

This is my model, how I mentioned, it’s cranshaft rotation simulation. I don’t care about stress, and strain, I want to know only displacements and in next simulation forces (probably I will be write to get some advise too :face_with_hand_over_mouth:)
After changing increment size, my best result is:

simulation should be end in inc. 10. so it’s so close (if I increase max inc. analyze probably will be finished but I am curious why inc. decreasing).
But I don’t understand what is a reason of changing inc time :man_shrugging:

I would try with a simplified version of this model first - an equivalent regular shaft (basically a cylinder) with the same settings to see how it behaves. Apart from that, check the results up to the last frame, you may notice what causes non-convergence. Check not only the deformed shape and field outputs (stresses, strains) but also selected history outputs (e.g. torque). You could investigate the residuals as well but it’s not so easy to interpret this data.

1 Like

let’s start the investigation:
my increment decrease between 4 and 5 step:

This is this line in convergence sheet:

so the residual force is not big.
At the picture below:

we can see last iteration in 4 step and small value of residual force but a little big value of inc. of displ.
And here, solver is decreasing inc. size.

What do you think about add controls card to the input deck?

Could you advise me, which control can be needed?

Mati, I have made several cranckshaft simulations (all no lineal static by the moment) even including the rods and pistons with a complete (360°) rotation. As other says, start with a simple model/mesh to get first the right bc, then improve the mesh. Feel free to send me the model and I will try to help you.

The sole reduction of the time increment by the solver is not so bad, the problems arise when the increment drops by several orders of magnitude and there is an obvious difficulty in convergence. It is often helpful to make a graph of these residuals (you can copy them into a spreadsheet).

I would not change the solver settings for now, this is the last resort and in practice rarely helps to achieve convergence.

I assume that you don’t want to share the model here, but if you can send it to me I’ll try to do some tests on it, check different options and maybe convergence can be achieved. Especially since you are not far from a complete solution.

Guess who made such nice simulation :slight_smile:


we all together :wink: i guess

Could you share BC of this simulation? Or send to me input deck?

I just looked and only can find on my email as an attachment, but is a very old version of PrePoMax file, if I’m ok is version 0.47


Thank you. This gifs are from abaqus?

Nop, they are from CalculiX results postproceed in Paraview. All those analysis were no lineal static, I could impose the force made by the gas being compressed at diferent angles of rotation of the cranckshaft (as a variable pressure at the head of pistons), but what I was not able to made was compute the model in dynamic state (steady at least), so we could see the difference in dynamic inertial loads at different velocities of the compressor.

1 Like

i found some files for these example,

1 Like