Enhancing Performance in Turbulent Pipe Flows with Shallow Dimples

ABSTRACT

Pipe flows underpin a wide range of engineering systems, the performance of which are primarily contingent on the heat transfer efficiency and pumping power required at a given flow rate. Dimples, in particular, have received considerable attention because they have been shown to be able to enhance heat transfer significantly whilst only incurring a relatively small drag increase penalty. In the present work, the effect of shallow dimples on thermo-aerodynamic performance in fully developed turbulent pipe flow is investigated using Large Eddy Simulation. The findings demonstrate that shallow dimples are able to reduce the pumping power required at a given flow rate whilst producing a modest improvement in heat transfer. Furthermore, the effect of particular dimple geometric parameters - namely, dimple diameter and shape - on flow topology and performance are explicated.

Learning-Based IMU Correction for UAV Localization in GPS-Denied Environments

ABSTRACT

Reliable localization is essential for stable UAV flight. However, systems that depend on GPS, cameras, or LiDAR can struggle in cluttered, low-light, or signal-poor environments. In this work, we present a learning-based approach using AirIMU, which relies only on IMU data. The model learns to correct sensor noise and estimate uncertainty through end-to-end training. We evaluate its performance on the EuRoC dataset and integrate it with the PX4 flight controller to enable real-time improvements in state estimation. This approach aims to improve UAV performance in environments where traditional sensing methods are unreliable.