Tangent operators and design sensitivity formulations for transient non‐linear coupled problems with applications to elastoplasticity

Abstract

SUMMARYTangent operators and design sensitivities are derived for transient non-linear coupled problems. The solution process and the formation of tangent operators are presented in a systematic manner and sensitivities for a generalized response functional are formulated via both the direct differentiation and adjoint methods. The derived formulations are suitable for finite element implementations. Analyses of systems, with materials that exhibit history dependent response, may be obtained directly by applying the analyses of transient %on-linear coupled systems. Rate-independent elastoplasticity is investigated as a case study and a problem with an analytical solution is analysed for demonstration purposes.

Adaptive FEM with goal‐oriented error estimation and an approximation of the dual problem for inelastic problems

“…(5) there is a linear equations system with the transposed tangential stiffness matrix from the primal problem (1st equation) and some kind of "evolution" for the dual internal variables (2nd equation). A similar result for sensitivity analysis was derived in [9]. Application of eq.…”

On gradient‐based optimization strategies for inverse problems in metal forming

“…Two popular alternatives to the Direct Difference Method (FDM), which exactly compute the sensitivities in the discretized setting, are the Direct Differentiation Method (DDM) and the Adjoint Differentiation Method (ADM), see [37] for an application to elasto-plasticity. Both are based on a transformation of the primal problem, either into corresponding pseudo problems with pseudo load vectors to compute the sensitivities with respect to each component of the design parameter vector (DDM) or into corresponding adjoint problems with Lagrangian multipliers for each component of the response function components (ADM).…”

“…An analytically calculation of the Jacobian J (µ) = ∂ϕ(µ) ∂µ is very cumbersome, especially in nonlinear elasto-plasticity [37]. As a remedy, the Jacobian is approximated by the forward finite difference method, see Eq.…”

Topology optimization of inertia driven dosing units

“…The nested formulation is usually applied in topology optimization as it ensures that the physics in every design iteration is modeled correctly. Furthermore, it yields the possibility to abort the optimization procedure at any intermediate iteration and, assuming that the design is feasible, achieve an optimized design The gradient information (sensitivities) that are needed to drive the optimization process is obtained using the adjoint method (Michaleris et al, 1994). It involves a single linear system solve for each functional i.e.…”