Search

International audience; The Adaptive Particle Refinement (APR) method results in a coupling between two SPH domains composed by either larger particles (mothers) or smaller particles (daughters). APR is one of the most promising local refinement techniques for SPH. The 2D dam break and shock tube test cases are used for validation purpose, with satisfactory results in comparison with reference solutions. An oscillating 2D channel test case is also simulated. It is found that the APR method has a lack of robustness in Lagrangian simulations of flows going back and forth. The refinement process appears to be particularly sensitive to the pre-existing daughters within the refining mother neighbourhood. The present paper aims to improve the APR method robustness, performing the refinement of a mother regarding its neighbourhood. Therefore, an optimization problem is formulated to determine the accurate amount, positions and variables of the new daughters to create. For the oscillating channel test case, results are compared qualitatively to previous works. The proposed refinement improves the robustness.