I was trying to use the alternative definition for the neo-hookean model (the one commented out in the code, $W=C{1}\left(I{1}-3-2 \ln J\right)+D{1}(J-1)^{2} ; J=\operatorname{det}(\boldsymbol{F})=\lambda{1} \lambda{2} \lambda{3}$). But it seems to be more unstable than the uncommented definition in the code ($W=C{1}\left(J^{-2 / 3} I{1}-3\right)+\left(\frac{C{1}}{6}+\frac{D{1}}{4}\right)\left(J^{2}+\frac{1}{J^{2}}-2\right)$), either using fmincon or Newton's method.
But the first definition seems more natural, with $\ln J$ ensuring it's inversion-free and $(J-1)^2$ ensuring it's volume-preserving. Is there any way to make Bartels support the first definition better (maybe without the $\ln J$ since that may cause trouble)? Thank you! :)
I was trying to use the alternative definition for the neo-hookean model (the one commented out in the code, $W=C{1}\left(I{1}-3-2 \ln J\right)+D{1}(J-1)^{2} ; J=\operatorname{det}(\boldsymbol{F})=\lambda{1} \lambda{2} \lambda{3}$). But it seems to be more unstable than the uncommented definition in the code ($W=C{1}\left(J^{-2 / 3} I{1}-3\right)+\left(\frac{C{1}}{6}+\frac{D{1}}{4}\right)\left(J^{2}+\frac{1}{J^{2}}-2\right)$), either using fmincon or Newton's method.
But the first definition seems more natural, with $\ln J$ ensuring it's inversion-free and $(J-1)^2$ ensuring it's volume-preserving. Is there any way to make Bartels support the first definition better (maybe without the $\ln J$ since that may cause trouble)? Thank you! :)