stress update algorithm in UMAT

[snowden]

recently, I am reading some books on the stress update algorithm. Because, it should be input into UMAT if you want to write a UMAT into ABAQUS. For a complicate constitutive model of material, possibly it is difficult.

what do you think of it? :lol:

[Jorgen]

Yes, writing a UMAT for ABAQUS or any other FE package is major undertaking requiring significant effort. What books did you read and what material model are you interested in?

Jorgen

[snowden]

Dear Dr.Jorgen:

I am reading a classical book: nonlinear finite element for continua and structures, written by Ted Belytschko. Chapt 5 concentrates the stress update algorithm of constitutive eqaution.

Did you read this book?

I want to discuss some problem with you. would you mind talk with me? :) :)

[Jorgen]

What are your problems? I saw from your other post that you are working with soft soil.

Jorgen

[snowden]

Dear Jorgen:

MY problem is how to update the stress for the following kind of flow rule?

viscoplastic strain rate=a scaling fuction *partial derivative of yield function with respect to stress component

if the scaling function (scalar quantity) is a function of stress component or strain conponent, how to update stress? That is different from commom plastic model in which the scaling function is not a function but a scalar value.

what do you think of it?

thank you very much

[Jorgen]

Hmm, maybe I am not quite following your arguments, but the flow rule that you specified basically says that the viscoplastic strain rate is a function of the current stress and strain rate. If this is true, why can you not simply us a traditional forward projection to get the viscoplastic strain (at t+dt) and then get the stress (at t+dt) at that state?

Jorgen

[snowden]

HI, Jorgen

thank you for your comments.

I think you are right. Actually, the implicit integration is needed to get the viscoplastic stain. I have derived and obtained it using backward Euler method. Now, I am confused about how to chose state variable (residual viscoplstic strain or residualstress )and apply Newton iteration?

many thanks

snowden

[Jorgen]

I would pick the plastic strain, if possible. It is often possible to use different statevariables. The actual choice should not matter as long as you can integrate forward in time.

- Jorgen

[snowden]

Hi Jorgen:

thank you.

In some cases, if the forward integration is used, disconvergence may occur. So I use implicit integration method.

I agree with you that the choice of the state variable does not matter for Newton iteration. But, for specified problem, appropriate state variables should be used avoiding some difficulty in differentiating some variables.

I have another problem. I read some papers and I found that some people used elastic predictor as the first step in stress update algorithm for inelastic constitutive model, but some people did not used it. In my opinion, we should use elastic predictor first.

what do you think of this?

thank you for your advice.

[Jorgen]

As you know, there are many different approaches to write a stress update algorithm. My experience is that dependent on the material model, there are different approaches that can be used successfully. Which algorithm to use will be dependent on the desired accuracy and effciency.

[snowden]

Hi, Jorgen:

Thank you

Yes, Ithink so.

For elastroplastic constitutive model, in theory, I think we should check all three residuals: plastic strain, yield surface and stress. Of course, this not the actual method to find an appropriate algorithm.

Can you give me a deep explaination for your comments? (Which algorithm to use will be dependent on the desired accuracy and effciency.)

thank you