hello there
i'm writing some User Material (UMAT) subroutine for a multiplicative decomposition of the plastic and elastic deformation gradient (instead of the additive decomposition of the strain). I need this for a large non-linear analysis (abaqus standard/static).
In Abaqus standard manual one can read that when the NLGEOM switch is flagged on, the default strain measure is the logarithmic strain (LE), which is calculated from the left stretch tensor V = sqrt(FFt) ( F is the deformation gradient and Ft it's transpose)
In the UMAT subroutine, one gets on the one hand the current strain and the strain increment of the current timestep through the variables STRAN and DSTRAN, and also the current deformation gradient at the end of the step DFGRD1.
so, from my point of view, i should get the same strain by STRAN+DSTRAN like i should get by calculating it directly from DFGRD1. (-> FFt, taking eigenvalues (EW) and eigenvectors (EV), applying the logarithm to the eigenvalues und transforming it back by EV ln(EW) EV^(-1). all this is done through lapack and also tested with python scripts which use numarray library. so there are not many places where i could have made an error. and i really rechecked it carefully.)
but i don't get the same strains. there are errors about 1% in the direct components (which should be only numerical problems) and about 6% in the shear terms (i'm aware of the engineering shear strain measure). but the 6% seem to be out of the range of numerical inprecision.
has anybody any suggestion or seen anything similar? it would be very helpful!
many thanks!
ratko
i'm writing some User Material (UMAT) subroutine for a multiplicative decomposition of the plastic and elastic deformation gradient (instead of the additive decomposition of the strain). I need this for a large non-linear analysis (abaqus standard/static).
In Abaqus standard manual one can read that when the NLGEOM switch is flagged on, the default strain measure is the logarithmic strain (LE), which is calculated from the left stretch tensor V = sqrt(FFt) ( F is the deformation gradient and Ft it's transpose)
In the UMAT subroutine, one gets on the one hand the current strain and the strain increment of the current timestep through the variables STRAN and DSTRAN, and also the current deformation gradient at the end of the step DFGRD1.
so, from my point of view, i should get the same strain by STRAN+DSTRAN like i should get by calculating it directly from DFGRD1. (-> FFt, taking eigenvalues (EW) and eigenvectors (EV), applying the logarithm to the eigenvalues und transforming it back by EV ln(EW) EV^(-1). all this is done through lapack and also tested with python scripts which use numarray library. so there are not many places where i could have made an error. and i really rechecked it carefully.)
but i don't get the same strains. there are errors about 1% in the direct components (which should be only numerical problems) and about 6% in the shear terms (i'm aware of the engineering shear strain measure). but the 6% seem to be out of the range of numerical inprecision.
has anybody any suggestion or seen anything similar? it would be very helpful!
many thanks!
ratko
