Entries for April 1, 2018
-
Variational Formulation of Elasticity
$ \newcommand{\argmin}{\mathop{\rm argmin}\nolimits} \newcommand{\cof}{\mathop{\rm cof}\nolimits} \newcommand{\sym}{\mathop{\rm sym}\nolimits} \newcommand{\invtra}{^{-T}} \newcommand{\eps}{\epsilon} \newcommand{\var}{\Delta} \newcommand{\Vvphi}{\Delta\Bvarphi} \newcommand{\vvphi}{\delta\Bvarphi} \newcommand{\BFC}{\boldsymbol{\mathsf{C}}} \newcommand{\BFc}{\boldsymbol{\mathsf{c}}} \newcommand{\push}{\Bvarphi_\ast} \newcommand{\pull}{\Bvarphi^\ast} $There are many books that give an outline of hyperelasticity, but there are few that try to help the reader implement solutions, and even fewer that manage to do it in a concise manner. Peter Wriggers’ Nonlinear Finite Element Methods is a great reference for those who like to roll up their sleeves and get lost in theory. It helped me understand a lot about how solutions to hyperelastic and inelastic problems are implemented.
One thing did not quite fit my taste though—it was very formal in the way that it didn’t give out indicial expressions. And if it wasn’t clear up until this point, I love indicial expressions, because they pack enough information to implement a solution in a single line. Almost all books skip these because they seem cluttered and the professors who wrote them think they’re trivial to derive. In fact, they are not. So below, I’ll try to derive indicial expressions for the update equations of hyperelasticity.
In the case of a hyperelastic material, there exists a strain energy function
Ψ:F↦Ψ(F)(1)which describes the elastic energy stored in the solid, i.e. energy density per unit mass of the reference configuration. The total energy stored in B is described by the the stored energy functional
E(φ):=∫BΨ(F)dm=∫Bρ0Ψ(F)dV(2)The loads acting on the body also form a potential:
L(φ):=∫Bρ0Γˉ⋅φdV+∫∂BtTˉ⋅φdA(3)where Γˉ and Tˉ are the prescribed body forces per unit mass and surface tractions respectively, where T=PN with Cauchy’s stress theorem.
The potential energy of B for deformation φ is defined as
Π(φ):=E(φ)−L(φ)(4)Thus the variational formulation reads
Find φ∈V such that the functional
Π(φ)=∫Bρ0Ψ(F)dV−∫Bρ0Γˉ⋅φdV−∫∂BtTˉ⋅φdA(5)is minimized for φ=φˉ on ∂Bu.
The solution is one that minimizes the potential energy:
φ∗=argminφ∈VΠ(φ)(6)A stationary point for Π means that its first variation vanishes: ΔΠ=0.
ΔΠ=DφΠ⋅δφ=:G(φ,δφ)=∫Bρ0∂F∂Ψ:∇(δφ)dV−∫Bρ0Γˉ⋅δφdV−∫∂BTˉ⋅δφdA(7)Using P=FS and P=ρ0∂Ψ/∂F,
ρ0∂F∂Ψ:∇(δφ)=FS:∇(δφ)=S:FT∇(δφ)(8)The symmetric part of the term on the right hand side of the contraction is equal to the variation of the Green-Lagrange strain tensor:
ΔE=DφE⋅δφ=dϵd21[∇(φ+ϵδφ)T∇(φ+ϵδφ)−I]ϵ=0=21[∇(δφ)TF+FT∇(δφ)](9)Substituting, we obtain the semilinear form G in terms of the second Piola-Kirchhoff stress tensor:
G(φ,δφ)=∫BS:ΔEdV−∫Bρ0Γˉ⋅δφdV−∫∂BTˉ⋅δφdA=0(10)We can write a Eulerian version of this form by pushing-forward the stresses and strains. The Almansi strain e is the pull-back of the Green-Lagrange strain E and vice versa:
e=φ∗(E)=F−TEF−1andE=φ∗(e)=FTEF(11)Commutative diagram for the pull-back and push-forward relationships of the Green-Lagrange and Almansi strain tensors.
Thus we can deduce the variation of the Almansi strain
Δe=F−TΔEF−1=21[∇(δφ)F−1+F−T∇(δφ)T]=21[∇x(δφ)+∇x(δφ)T](12)where we have used the identity
∇X(⋅)F−1=∇x(⋅).(13)The second Piola-Kirchhoff stress is the pull-back of the Kirchhoff stress τ:
S=φ∗(τ)=F−1τF−T(14)Then it is evident that
S:ΔE=(F−1τF−T):(FTΔeF)=τ:Δe(15)We can thus write the Eulerian version of (10):
G(φ,δφ)=∫Bτ:ΔedV−∫Bρ0Γˉ⋅δφdV−∫∂BTˉ⋅δφdA=0(16)Introducing the Cauchy stresses σ=τ/J, we can also transport the integrals to the current configuration
G(φ,δφ)=∫Sσ:Δedv−∫Sργˉ⋅δφdv−∫∂Sttˉ⋅δφda=0(17)Here, we substituted the following differential identities:
ρ0ΓdV=ργdv(18)for the body forces, and
TdA=PNdA=σJF−TNdA=σnda=tda(19)for the surface tractions, where we used the Piola identity.
Linearization of the Variational Formulation
We linearize G:
LinGφˉ=G(φˉ,δφ)+ΔGφˉ=0(20)Then we have the variational setting
a(Δφ,δφ)=b(δφ)(21)where
a(Δφ,δφ)=ΔGφˉandb(δφ)=−G(φˉ,δφ)(22)Commutative diagram of the linearized solution procedure. Each iteration brings the current iterate φˉ closer to the optimum value φ∗.
Mappings between line elements belonging to the tangent spaces of the linearization.
The variation ΔG is calculated as
ΔG=DφG⋅Δφ=∫B[ΔS:ΔE+S:Δ(ΔE)]dV(23)Consecutive variations of the Green-Lagrange strain tensor is calculated as
Δ(ΔE)=DφΔE⋅Δφ=21[∇(δφ)T∇(Δφ)+∇(Δφ)T∇(δφ)](24)The term on the left is calculated as
ΔS=DφS⋅Δφ=∂C∂S:ΔC=2∂C∂S:ΔE(25)where we substitute the Lagrangian elasticity tensor
C:=2∂C∂S=4ρ0∂C∂C∂2Ψ(26)and ΔE is calculated in the same manner as ΔE:
ΔE=21[∇(Δφ)TF+FT∇(Δφ)](27)Then the variational forms of the linearized setting are
a(Δφ,δφ)b(δφ)=∫BΔEˉ:Cˉ:ΔEˉ+Sˉ:[∇(δφ)T∇(Δφ)]dV=−∫BSˉ:ΔEˉdV+∫Bρ0Γˉ⋅δφdV+∫∂BTˉ⋅δφdA(28)where the bars denote evaluation φ=φˉ of dependent variables.
Eulerian Version of the Linearization
We also have the following relationship between the Lagrangian and Eulerian elasticity tensors
cabcd=FaAFbBFcCFdDCABCD(29)Substituting Eulerian expansions, we obtain the following identity:
ΔE:C:ΔE=(FTΔeF):C:(FTΔeF)=FaAΔeabFbBCABCDFcCΔecdFdD=ΔeabcabcdΔecd=Δe:c:Δe(30)Thus we have
S:[∇(δφ)T∇(Δφ)]=[F−1τF−T]:[∇(δφ)T∇(Δφ)]=τ:[(∇(δφ)F−1)T∇(Δφ)F−1]=τ:[∇x(δφ)T∇x(Δφ)](31)With these results at hand, we can write the Eulerian version of our variational formulation:
a(Δφ,δφ)b(δφ)=∫BΔeˉ:cˉ:Δeˉ+τˉ:[∇xˉ(δφ)T∇xˉ(Δφ)]dV=−∫Bτˉ:ΔeˉdV+∫Bρ0Γˉ⋅δφdV+∫∂BTˉ⋅δφdA(32)If we introduce the Cauchy stress tensor σ and corresponding elasticity tensor cσ=c/J, our variational formulation can be expressed completely in terms of Eulerian quantities:
a(Δφ,δφ)b(δφ)=∫SˉΔeˉ:cˉσ:Δeˉ+σˉ:[∇xˉ(δφ)T∇xˉ(Δφ)]dvˉ=−∫Sˉσˉ:Δeˉdvˉ+∫Sˉργˉ⋅δφdvˉ+∫∂Sˉttˉ⋅δφdaˉ(33)We have the following relationships of the differential forms:
dvˉ=Jˉdvandnˉdaˉ=cofFˉNdA(34)where Fˉ=∇Xφˉ and Jˉ=detFˉ.
Discretization of the Lagrangian Form
We use the following FE discretization:
φh=γ=1∑nnφγNγ=γ=1∑nna=1∑ndφaγeaNγ(35)where nn is the number of element nodes and nd is the number of spatial dimensions.
We use the same discretization for δφ and Δφ. Then the linear system at hand becomes
δ=1∑nnb=1∑ndAabγδΔφbδ=baγ(36)for a=1,…,nd and γ=1,…,nn where the A and b are calculated from the variational forms as
Aabγδbaγ=a(ebNδ,eaNγ)=b(eaNγ)(37)For detailed derivation, see the post Vectorial Finite Elements.
For discretized gradients, we have the following relationship
∇X(eaNγ)=(ea⊗Bγ)(38)where Bγ:=∇XNγ. For the first term in a, we can get rid of the symmetries:
sym(FˉT∇(eaNγ)):Cˉ:sym(FˉT∇(ebNδ))=(FˉT(ea⊗Bγ)):Cˉ:(FˉT(eb⊗Bδ))=FˉaABBγCˉABCDFˉbCBDδ(39)and for the second term, we have
Sˉ:[∇(eaNγ)T∇(ebNδ)]=Sˉ:[(ea⊗Bγ)T(eb⊗Bδ)]=Sˉ:[Bγ⊗Bδ]gab=BAγSˉABBBδgab(40)where gab are the components of the Eulerian metric tensor.
For the first term in b, we have
Sˉ:sym(FˉT∇(eaNγ))=Sˉ:(FˉT(ea⊗Bγ))=SˉABFˉaABBγ(41)Remaining terms can be calculated in a straightforward manner. We then have for A and b:
Aabγδbaγ=∫BFˉaABBγCˉABCDFˉbCBDδ+BAγSˉABBBδgabdV=−∫BSˉABFˉaABBγdV+∫Bρ0ΓˉaNγdV+∫∂BtTˉaNγdA(42)The lowercase indices in γˉ and Tˉ might be confusing, but in fact
Γa(X,t)Ta(X,t)=γa(x,t)∘φ(X,t)=ta(x,t)∘φ(X,t)(43)The system is solved for Δφ at each Newton iteration with the following update equation:
φ←φˉ+Δφ(44)Discretization of the Eulerian Form
Discretization of the Eulerian formulation parallels that of Lagrangian.
Aabγδbaγ=∫BBˉcγcˉacbdBˉdδ+BˉeγτˉefBˉfδgabdV=−∫BτˉabBˉbγdV+∫Bρ0ΓˉaNγdV+∫∂BtTˉaNγdAorAabγδbaγ=∫SˉBˉcγcˉacbdσBˉdδ+BˉeγσˉefBˉfδgabdvˉ=−∫SˉσˉabBˉbγdvˉ+∫SˉργˉaNγdvˉ+∫∂SˉttˉaNγdaˉ(45)(46)Here, Bˉγ=∇xˉNγ denote the spatial gradients of the shape functions. One way of calculating is Bˉγ=Fˉ−TBγ, similar to (13).
The update equation (44) holds for the Eulerian version.
Conclusion
The equations above in boxes contain all the information needed to implement the nonlinear solution scheme of hyperelasticity.