NonlinearJ2

Nonlinear General J2 Plasticity Model

Summary

This is an abstract material class thus cannot be used directly. This class defines a general plasticity model using J2 yielding criterion with associated flow rule and mixed hardening rule. The isotropic/kinematic hardening response can be customized.

To use this model, a derived class shall be defined first.

class YourJ2 final : public NonlinearJ2 {
// class definition
}

The derived class only needs to implement four pure virtual methods that define the isotropic and kinematic hardening rules.

virtual double compute_k(double) const = 0; // isotropic hardening
virtual double compute_dk(double) const = 0; // derivative isotropic
virtual double compute_h(double) const = 0; // kinematic hardening
virtual double compute_dh(double) const = 0; // derivative kinematic

All four methods take equivalent plastic strain as the input argument, on output, the corresponding quantities shall be provided.

The isotropic hardening function $K({\overline{\epsilon }}_p)$ defines the isotropic hardening rule, there are some requirements:

1. $K({\overline{\epsilon }}_p)$ should be non-negative,
2. $K({\overline{\epsilon }}_p=0)={\sigma }_y$ where ${\sigma }_y$ is the initial yielding stress.

There is no requirement for the kinematic hardening function $H({\overline{\epsilon }}_p)$. Both hardening rules can coexist. However, to successfully solve the trial status, there is an additional constraint that shall be applied on the model:

$$E+H^{\prime }({\overline{\epsilon }}_p)+K^{\prime }({\overline{\epsilon }}_p)⩾0\text{for all}{\overline{\epsilon }}_p$$

Otherwise, the local Newton iteration will fail.

Brief On Theory

The NonlinearJ2 abstract class defines an associative plasticity framework using the von Mises yield criterion, which is defined as follows.

$$F(\sigma ,{\overline{\epsilon }}_p)=\sqrt{{\displaystyle \frac{3}{2}}(s-\beta ({\overline{\epsilon }}_p)):(s-\beta ({\overline{\epsilon }}_p))}-{\sigma }_y({\overline{\epsilon }}_p)$$

where $\beta ({\overline{\epsilon }}_p)$ is the back stress depends on the equivalent plastic strain ${\overline{\epsilon }}_p$ and ${\sigma }_y({\overline{\epsilon }}_p)$ is the yield stress. Note

$$\sqrt{3J_2}=\sqrt{{\displaystyle \frac{3}{2}}(s-\beta ({\overline{\epsilon }}_p)):(s-\beta ({\overline{\epsilon }}_p))}$$

It is also called J2 plasticity model. A detailed discussion can be seen elsewhere. $\beta ({\overline{\epsilon }}_p)=H({\overline{\epsilon }}_p)$ and ${\sigma }_y({\overline{\epsilon }}_p)=K({\overline{\epsilon }}_p)$.

History Layout

location	paramater
initial_history(0)	accumulated plastic strain
initial_history(1-6)	back stress

Kinematic Hardening

The back stress $\beta ({\overline{\epsilon }}_p)$ defines a kinematic hardening response. For example a linear kinematic hardening could be defined as:

$$\beta ({\overline{\epsilon }}_p)=E_K{\overline{\epsilon }}_p$$

and the derivative

$${\displaystyle \frac{\mathrm{d}\beta ({\overline{\epsilon }}_p)}{\mathrm{d}{\overline{\epsilon }}_p}}=E_K$$

in which $E_K$ is the kinematic hardening stiffness.

In this case, user shall override the corresponding two methods with such an implmentation.

double SampleJ2::compute_h(const double p_strain) const { return e_kin * p_strain; }
double SampleJ2::compute_dh(const double) const { return e_kin; }

Of course, a nonlinear relationship could also be defined.

Isotropic Hardening

The isotropic hardening is defined by function ${\sigma }_y({\overline{\epsilon }}_p)$. The value ${\sigma }_y(0)$ should be the initial yield stress. Also, for a bilinear isotropic hardening response, user shall override the following two methods.

double SampleJ2::compute_k(const double p_strain) const { return e_iso * p_strain + yield_stress; }
double SampleJ2::compute_dk(const double) const { return e_iso; }

Here another polynomial based isotropic hardening function is shown as an additional example. The function is defined as

$${\sigma }_y={\sigma }_0(1+\sum _{i=1}^n{a}_i{\overline{\epsilon }}_p^i)$$

where $a_i$ are material constants. It is easy to see ${\sigma }_y{|}_{{\overline{\epsilon }}_p=0}={\sigma }_0$. The derivative is

$${\displaystyle \frac{\mathrm{d}{\sigma }_y}{\mathrm{d}{\overline{\epsilon }}_p}}={\sigma }_0\sum _{i=1}^{n}i{a}_i{\overline{\epsilon }}_p^{i-1}$$

To define such a hardening behavior, a vector shall be used to store all constants.

// PolyJ2.h
const vec poly_para; // poly_para is a vector stores a_i

The methods could be written as follows.

// PolyJ2.cpp
double PolyJ2::compute_k(const double p_strain) const {
 vec t_vec(poly_para.n_elem);

 t_vec(0) = 1.;
 for(uword I = 1; I < t_vec.n_elem; ++I) t_vec(I) = t_vec(I - 1) * p_strain;

 return yield_stress * dot(poly_para, t_vec);
}

double PolyJ2::compute_dk(const double p_strain) const {
 vec t_vec(poly_para.n_elem);

 t_vec(0) = 0.;
 t_vec(1) = 1.;
 for(uword I = 2; I < t_vec.n_elem; ++I) t_vec(I) = (double(I) + 1.) * pow(p_strain, double(I));

 return yield_stress * dot(poly_para, t_vec);
}

Example

A few different models are shown as examples. User can define arbitrary models.