International Journal of Science, Engineering and TechnologyAuthors: Dr Thadakamalla Srinivasulu, Assistant Professor
Abstract: The numerical simulation of fluid flows continues to advance through the integration of machine learning, high‑order discretizations, mesh‑free frameworks, and emerging quantum‑inspired algorithms. This review synthesizes peer‑reviewed research published between 2025 and 2026, focusing on five transformative methodologies: physics‑informed neural networks augmented with lattice Boltzmann kinetics, hybrid high‑order formulations for turbulent flows, iterative high‑order smoothed particle hydrodynamics, p‑adaptive mesh‑free frameworks, and quantum‑assisted computational fluid dynamics (CFD). Key quantitative advances include two orders of magnitude improvement in smoothed particle hydrodynamics accuracy, mean absolute error reduction by a full order of magnitude in neural network solvers, computational cost savings up to 50% through p‑adaptivity, and machine learning acceleration of CFD simulations by up to 10,000 times. These developments collectively indicate a paradigm shift toward hybrid, adaptive, and cross‑paradigm numerical methods that address the longstanding trade‑offs between accuracy, stability, and computational cost in fluid dynamics research.
DOI: http://doi.org/10.5281/zenodo.20583590
