On process-step parallel computability and linear superposition of mechanical responses in additive manufacturing process simulation

Abstract

We study linearity assumptions in the transient macroscale mechanical aspect of additive manufacturing (AM) process simulation. Linearity assumptions are often resorted to in combination with calibrated inelastic deformation components to arrive at computationally tractable yet reasonably accurate AM process models. We point out that linearity assumptions permit the independent computation of the response increment in each step of the AM process, and the total mechanical response is the superposition of all the process-step increments. In effect, process-step increments are computed with respect to the stress-free reference configuration in each step. The implication is that the mechanical response increment in each linearised AM process step may be computed in parallel. Trivial process-step-wise parallel computability breaks down, however, if nonlinearity (i.e. geometric or material) is modelled. In our investigation the influence of geometric nonlinearity on part distortion is small (but this is of course part-geometry specific), and more realistic stresses are obtained by imposing a nonlinear elastoplastic material law after the parallel computation and superposition of the linear AM response increments. It is demonstrated that simulation wall-clock time is reduced by exploiting process-step parallel computability in the linear regime. Moreover, numerical experiments suggest that process-step parallelization scales better (in wall-clock time) than conventional parallelization in the sequential computation of each response increment.