Jike Han | Searcharxiv

arXiv2026

Elastoplastic inherent strain-based topology optimization for residual stress reduction in metal additive manufacturing

This paper proposes a topology optimization method for reducing the residual stress arising in the building process of metal additive manufacturing. First, a layer-by-layer process analysis model based on an elastoplastic inherent strain method is introduced. In this model, the incremental displacement is solved anew at each layer step, and the stress history is explicitly incorporated into the constitutive equation as the stress accumulated up to the previous step, which guarantees the stress continuity across layer interfaces without introducing activation strains. Next, the design sensitivity of this analysis model is derived based on the adjoint method. Taking the pair of the stress and the equivalent plastic strain as the state variables reduces the dependency between layer steps to a one-step recurrence, and the adjoint fields are constructed as a layer-by-layer reverse sweep that reuses the coefficient tensors obtained in the forward analysis. Consequently, the cost of the sensitivity analysis scales linearly with the number of layers and remains of the same order as that of the forward analysis. An optimization problem is then formulated based on the density method to minimize the P-norm of the residual stress at the completion of the building process under the volume and final-use compliance constraints, and the derived sensitivities are verified by comparison with central finite differences. Finally, the proposed method is demonstrated through two- and three-dimensional examples of residual stress minimization under a compliance constraint. The results clarify that, under the elastoplastic analysis, the maximum residual stress is bounded by the yield surface, and the optimization therefore reduces the extent of the yielded and plastic strain accumulating regions rather than the peak stress value.

cs.CE↗

arXiv2026

Topology optimization of conduction-radiation problems based on a ray-tracing approach

Thermal management is essential in space systems, where electronic devices must dissipate heat via radiative heat transfer. To achieve efficient designs of radiative cooling devices, structural optimization approaches such as topology optimization are required. While existing topology optimization methods have incorporated radiative heat transfer with certain simplifications, fully accounting for multidirectional mutual radiation remains challenging. To address this issue, this study proposes a density-based topology optimization method for conduction-radiation heat transfer problems that accounts for multidirectional mutual radiation. The proposed method integrates a zonal-method-based radiative heat transfer analysis incorporating a ray-tracing method into the finite element heat conduction analysis, capturing radiation effects during the optimization process. By treating the intermediate material densities that arise during the optimization as participating media, the proposed method enables a physically consistent evaluation of radiative heat transfer on implicitly represented structural boundaries. The analytical design sensitivities are derived using the adjoint method, and the accuracy is confirmed by the comparison with the numerical sensitivities obtained by the finite difference method. Numerical examples demonstrate the optimization of radiative heat sinks and radiation shields. The heat sink examples clarify how the balance between conduction and radiation governs the resulting designs, while the radiation shield examples produce multilayer insulation structures that are not obtained by conventional approaches.

cs.CE↗