Punit Pandey | Searcharxiv

arXiv2026

Development and validation of a sharp interface immersed boundary method for high-speed flows

This study presents an advanced sharp-interface immersed boundary method (IBM) integrated with the blastFOAM library on the OpenFOAM platform for high-speed compressible flow simulations. The developed solver extends the existing IBM techniques available in OpenFOAM to compressible regimes, tackling challenges such as shock waves, expansions, and dynamic geometries without needing body-fitted meshes. A novel contribution of this work is the implementation of a slip boundary condition for velocity at immersed surfaces, specifically designed to handle inviscid highspeed flows. The method also combines the second-order polynomial IBM reconstruction with multiple flux schemes such as Kurganov, Tadmor, HLL (Harten-Lax-van Leer), and AUSM+up (Advection Upstream Splitting Method Plus Upwind). The technique achieves significant accuracy across diverse high-speed flow conditions. Extensive validation is performed through supersonic flow cases over a wedge, a cylinder, an aerofoil, a sphere, and a moving piston. Results show excellent agreement with analytical and body-fitted solutions, with sharp resolution of shocks, minimal numerical oscillations, and shock reflections. A grid convergence study confirms the solver's reliability across varying mesh resolutions, while three-dimensional simulations highlight its capability for scaled-up applications. This solver provides a flexible, efficient, and accurate tool for capturing high-speed flow phenomena across various Mach numbers and geometries. It offers significant advantages in mesh handling, particularly for dynamic or intricate configurations, making it ideal for aerospace and engineering applications involving compressible flows.

physics.class-ph↗

arXiv2010

Reinforcement Learning by Comparing Immediate Reward

This paper introduces an approach to Reinforcement Learning Algorithm by comparing their immediate rewards using a variation of Q-Learning algorithm. Unlike the conventional Q-Learning, the proposed algorithm compares current reward with immediate reward of past move and work accordingly. Relative reward based Q-learning is an approach towards interactive learning. Q-Learning is a model free reinforcement learning method that used to learn the agents. It is observed that under normal circumstances algorithm take more episodes to reach optimal Q-value due to its normal reward or sometime negative reward. In this new form of algorithm agents select only those actions which have a higher immediate reward signal in comparison to previous one. The contribution of this article is the presentation of new Q-Learning Algorithm in order to maximize the performance of algorithm and reduce the number of episode required to reach optimal Q-value. Effectiveness of proposed algorithm is simulated in a 20 x20 Grid world deterministic environment and the result for the two forms of Q-Learning Algorithms is given.

cs.LG↗