Calibrate LS-DYNA *MAT_ADD_INELASTIC

Introduction

Recent versions of LS-DYNA support the ability to add inelasticity to almost any existing material model (like a hyperelastic model). This is quite a unique idea that does not, as far as I know, exist in any other FE software. Using this approach you can, for example, take any of the LS-DYNA hyperelastic models and then convert it to a multi-network model with flow elements. In some cases this may be useful for FE simulations of polymers. In this article I will show how you can explore and calibrate this type of model framework using MCalibration!

Figure: Rheological representaion of the *MAT_ADD_INELASTIC model from the LS-DYNA User’s Manual.

Example 1: Two Network Model

As a first example I will create a two-network model consisting of one network with a Neo-Hookean spring, and a second network with a Neo-Hookean spring in series with a Norton-Bailey creep flow component. This creep model has the following equation:

\(\dot{\varepsilon}_C = \displaystyle\left[ A \left( \frac{\sigma}{\sigma_0}\right)^n \left( \frac{T}{T_0} \right)^p \left( (1+m)(\varepsilon_0+\varepsilon_c) \right)^m \right]^{1/(1+m)}\)

This material model can be used in MCalibration using the following LS-DYNA template:

				
					$$ Units: [length]=mm, [force]=N, [time]=sec, [temperature]=K
$$ For LS-DYNA R13
*MAT_HYPERELASTIC_RUBBER
$$ Neo-Hookean
$$     mid,       ro,      pr
        1,    1.0e-9,  %PR=0.4995%
$$     c10
%C10=1.0%
*MAT_ADD_INELASTICITY
$$ Norton-Bailey Creep
$$     mid, nielinks
         1,         1
$$ nielaws,   weight
         1, %w=0.8%
$$     law,    model
         5,        3
$$       A,     sig0,       n,      T0,    p,    m, eps0
%A=1.0%, %sig0=3.0%, %n=4.0%, %T0=293.0%, %p=0.0%, %m=0.0%, %eps0=0.0%
				
			

The details of the *MAT_HYPERELASTIC_RUBBER and *MAT_ADD_INELASTICITY commands are provided in the LS-DYNA User’s Manual. Note 1: I converted some of the floating point values to MCalibration variables using the %variable=[value]% syntax. Note 2: LS-DYNA comments need to have two dollar signs in MCalibration. When using this material model in MCalibration it is necessary to specify that LS-DYNA is the solver that should be used, see the following Load Case Dialog image.

MCalibration can quickly perform stress-strain calculations once the material model and load case have been specified. Here’s an example in which I applied monotonic uniaxial tension with at constant strain rate of 0.01/s to an engineering strain of 0.5, and then unloaded back to zero strain using the same rate.

Example 2: Four Network Model

It is easy to work with and calibrate models with more parallel networks. The following example shows one case in which I defined a 4 network model consisting of Neo-Hookean springs and Norton-Bailey creep. 

				
					$$ Units: [length]=mm, [force]=N, [time]=sec, [temperature]=K
$$ For LS-DYNA R13
*MAT_HYPERELASTIC_RUBBER
$$ Neo-Hookean
$$     mid,       ro,      pr
        1,    1.0e-9,  %PR=0.4995%
$$     c10
%C10=1.0%
*MAT_ADD_INELASTICITY
$$ Norton-Bailey Creep
$$     mid, nielinks
         1,       3
$$ - Net 1 -
$$ nielaws,   weight
         1, %w1=0.3%
$$     law,    model
         5,        3
$$       A,     sig0,       n,      T0,    p,    m, eps0
%A=1.0%, %sig1=3.0%, %n=4.0%, %T0=293.0%, %p=0.0%, %m=0.0%, %eps0=0.0%
$$ - Net 2 -
$$ nielaws,   weight
         1, %w2=0.3%
$$     law,    model
         5,        3
$$       A,     sig0,       n,      T0,    p,    m, eps0
%A=1.0%, %sig2=3.0%, %n=4.0%, %T0=293.0%, %p=0.0%, %m=0.0%, %eps0=0.0%
$$ - Net 3 -
$$ nielaws,   weight
         1, %w3=0.3%
$$     law,    model
         5,        3
$$       A,     sig0,       n,      T0,    p,    m, eps0
%A=1.0%, %sig3=3.0%, %n=4.0%, %T0=293.0%, %p=0.0%, %m=0.0%, %eps0=0.0%
				
			

The following figure shows the load-unload predictions from this model.

Summary

  • MCalibration makes it is easy to both examine and calibrate the new LS-DYNA *MAT_ADD_INELASTICITY framework models.
  • My typically recommendation is to use 2 networks for rubber-like materials, and 3 networks for thermoplastics.
Facebook
Twitter
LinkedIn

More to explore

Residual Strain Prediction

Most tests are in displacement control, but residual strain is measured in load control. This article explains how to use this mixed data.

Leave a Comment