Efficient Hyper-Reduced Order Modelling of Mixed Contact Problems defined on Non-matching grids
Please login to view abstract download link
Contact mechanics problems are often time-consuming to solve due to the non-linearity and non-smoothness of the contact conditions. For contact problems treated with Lagrange multipliers (fulfilling Signorini contact conditions), the application of reduced-order modeling remains a challenge, in particular due to the non-negativity constraint on Lagrange multipliers. To overcome this issue, the Hybrid Hyper-Reduction (HHR) method has been recently proposed [1, 2]. The main point of this approach is to solve the reduced problem in a mesh sampling, called reduced integration domain (RID), reducing the number of dual unknowns without having to use a compressed dual basis. So far, the HHR method has only been applied to node-to-node contact pairing, which limits its application potential. The aim of this work is to present an extention of the HHR method to non-matching contact surfaces meshes allowing the treatment of more complex contact geometries. A node-to-surface pairing associated to a master-slave strategy is considered. This non-conforming mesh configuration leads to a generalized reduced saddle-point problem (non-symmetric extra diagonal blocks) to be solved. Moreover new constraints have to be taken into account in the RID construction in order the HHR problem to remain well-posed. The efficiency of the extended HHR method is demonstrated on industrial test cases derived from the nuclear fuel simulation field. The general framwork of nonlinear materials is easily handled thanks to the Hyper-Reduced approach, that considerably reduces the computational effort of constitutive relation integration. REFERENCES [1] J. Fauque, I. Ramière, and D. Ryckelynck. Hybrid hyper-reduced modeling for contact mechanics problems. IJNME, 115(1):117–139, 2018. [2] S. Le Berre, I. Ramière, J. Fauque, and D. Ryckelynck. Condition number and clustering-based efficiency improvement of reduced-order solvers for contact problems using lagrange multipliers. Mathematics, 10(9):1495, 2022.