Efficiency of PINN-LPCDD for long-time shock–bubble simulations

Investigate the computational efficiency of PINN-LPCDD relative to traditional high-precision numerical methods for long-time simulations of shock–bubble interaction problems.

Background

The paper compares the computational cost of PINN-LPCDD with traditional high-precision methods for a short-time two-dimensional shock–bubble interaction problem. Although the costs are comparable at the accuracy level used for the reference solution, the authors explicitly limit this comparison to short-time simulations and leave the efficiency of long-time simulations unresolved. Long-time performance is particularly relevant because the method’s accuracy may degrade through error accumulation and because the proposed time-staged training strategy has not yet been developed or evaluated.

References

Note that this comparison is limited to the short-time simulation of the shock--bubble interaction problem considered in this paper; the efficiency for long-time simulations remains to be further investigated in future work.

— PINN-LPCDD: Physics-Informed Neural Network with Learnable and Physically Consistent Domain Decomposition for Compressible Multi-Medium Flows  (2610.22903 - Zeng et al., 16 Oct 2026) in Appendix B, paragraph “2D cases”