Maximum deflection changes by the number of convergence

[dido32511]

Good day, 
I am trying to calculate the maximum deflection on steel beam and i am using Non-Linear analysis 

I used the attached settings and I noticed that every time I increase the number of convergence I get different result sometimes bigger  , I wonder what is the best number to use to calculate maximum deflection.

12 convergenceis the previous result of the research I am comparing with.

15 convergence10 convergence

[John_Holtz]

Hi @dido32511 

The answer is you should try 20 increments, 40 increments, 60 increments, and so on. Once you reach the required number of increments to get an accurate answer, you will see the results will be the same.

 Another way to ask your question is why do the results change so much with the number of increments. It may be possible to determine that from viewing all the results you have. For example,

1. Is the reaction force correct at the end of the analysis? (Right-click on the constraint in the model tree > SPC Summation) If not, is there some correlation between the reaction force and the number of steps? 
2. Does the material yield? If not, what result do you get from a linear analysis? (If the volume of yielding is small, what result do you get from a linear analysis?)
3. When does the material start to yield? Is it in between two increments? (Maybe the amount of displacement that occurs when it goes plastic is capture better with X steps instead of Y steps.)
4. How much of the cross-section yields? How many increments are calculated after the part yields?

John

[dido32511]

Thank you very much I will try and hopefully everything will be ok , I now have better understanding of the concept

[dido32511]

Hi Mr john @John_Holtz , I did some trial and error for another Beam which I have the experiment results for it in which the maximum deflection is 14.73 mm and the deflection I had from Nastran with 9 convergence is 14.45 m
The yielding stress for the material is 275 Mpa 
and the chart of material I used was elasto plastic , as you can see clearly the stress yields during the analysis 

My question is : Are the convergence similar to the loading criteria in real life ?
I mean when we watch any loading experience , the load keeps pressing and increasing on the beam so the load increases until it reaches the final capacity of the machine in which the material properties changes till it yields.

so is the convergence controls that point like is it related to the time of the loading ?
I did another shot by changing the convergence to 11 but I got the results of Inc 9 which should according to my understanding match the 9  convergence (isn't that correct ? ) but the results didn't match in any way neither in stress or deflection(29.12 mm ) you can refer to the attached stress and deflection in both conditions as 



I am so sorry for the confusion but what i am trying to understand is the effect of convergence on simulation of real life experience as the next stage will be simulating the  effect of using fiber on the beam to see how it will behave so knowing how to simulate the real experiments is essential 

I used the below settings

Convergence 9

[John_Holtz · Accepted solution]

Hi,

If you are changing the convergence criteria and error tolerances input to values of 9, 10, and so on, DON'T CHANGE THEM.

First, you should not be changing the default convergence criteria unless the analysis is unable to converge and/or you have a good reason to change them.

Second, you should understand what you are entering. Indicating the results are okay if they are within plus or minus 900% of the stress-strain curve (Displacement convergence = 9) is wrong. I can do a hand calculation and get a result more accurate than 900%. 🙂 If you want to change the error tolerance, please start with this article: Understanding convergence in a Nastran nonlinear analysis | Inventor Nastran | Autodesk Knowledge Ne...

Hopefully that will give you more accurate results regardless of the number of iterations used in the analysis.

John