Increment Size effect

[ZohrehA]

Dear ABAQUS users,

I have a question regarding increment size.
I ran a model with 2 different increment size: 0.05 & then 0.01 (initial & max increment size), the minimum increment size are the same for both case and it is 5E-11.
But I got two different results (thermal history).
The model is a simple heat transfer (transient) model, with two steps heating and cooling. The elements type are standard and linear.

Could anyone help me that why the results are changing by increment size? And how I can find the best increment size to have reliable results?

Regards,
Zohreh

 

[FEA way]

Too high increment size may cause Abaqus to skip some effects in transient analyses. It's better to check if you're not missing something by rerunning with a lower increment size. Sometimes you will know the minimum necessary increment size because e.g. your amplitude's resolution will determine it (Abaqus may skip changes in amplitudes if you don't set a low enough increment size).

However, in heat transfer analyses, there are some special considerations. First of all, there's the DELTMX parameter which sets the maximum temperature change allowed in an increment. You can use low values to increase the accuracy. On the other hand, there's the concept of minimum usable time increment. Basically, too low (with respect to the mesh size and material properties) increment sizes in transient heat transfer analyses may result in spurious oscillations with second-order elements. To avoid that, you can use the following formula to estimate the initial increment size in such analyses:

Δt >= ((ρ*c)/(6*k))*(Δl^2)

where: Δt - time increment, ρ - density, c - specific heat, k - conductivity, Δl - distance between nodes of elements near the surface with the highest temperature gradient.

 

[ZohrehA]

Dear FEA way,

Thanks for your answer.

I have some more questions:

1- According to my simulation the highest temperature with increment size of 0.01 is about 1100°c. Is it ok to set DELTMX 2000? Or I must choose even lower?

2- About increment size: you say better to check lower inc. size, you mean lower than 0.01?

3- I didn't get this point: "Basically, too low increment sizes in transient heat transfer analyses may result in spurious oscillations with second-order elements."

4- According to your formula the time increment must be 0.0015 s. But it takes too much time! 0.0015 s is not too low increment size regarding spurious oscillations?!

 

[FEA way]

DELTMX sets the maximum allowed temperature change at any node in a single increment. So if it's 2000 then Abaqus will allow jumps in temperature of almost 2000 degrees which may not be good for accuracy. On the other hand, too low DELTMX may cause convergence issues.

The choice of the increment size depends on the problem but if you can rerun with a lower increment size and care about accuracy then it might be good to try. It's like with mesh refinement.

The problem of those oscillations is explained in detail in the documentation chapter Analysis --> Analysis Procedures --> Heat Transfer and Thermal-Stress Analysis --> Uncoupled Heat Transfer Analysis (paragraph "Spurious Oscillations due to Small Time Increments").

The formula is used to calculate the minimum usable increment size so you should use a higher one. This is just a lower limit.

 

[ZohrehA]

Thanks for your answer.

Is there a way that I can find maximum stable increment size that give me accurate results?

I should explain more what I am doing:
I have an experimental thermal history from a thermal spray coating process. The aim of simulation is finding the amount of unknown heat input during this process.
For that I have to change the amount of heat input in Load module by try and error and then compare the simulation thermal history with the experiment one, to catch the best fit. Then I can say this heat input is the real heat input of the process.
The increment size affect the simulation thermal history too much. This means I cannot find which heat input is the real one. Because by decreasing increment size with the same heat input, thermal history graph is higher. So I don't know that I have to change the heat input or the increment size is not suitable.
So I have to know the maximum increment size that give me the accurate and real results of temperature to find the real heat input in reasonable time (avoiding too much computational time).
I hope I was able to explain my problem.

And one more point: the elements in my model are first-order elements. The formula that you had written is also valid for these type of elements also works?