Abstract
Availability of potential drone-delivery services for medical operations are expected to be limited by extreme weather conditions. Hence reliable wind predictions systems need to be developed for accessing wind and turbulence conditions along the drone flight paths that can account for the impact of different scales (from buildings to terrains) . The aim is to provide accurate and fast wind flow conditions in drone flight path in vicinity of urban hospitals. In this direction, two modelling frameworks have been developed: (A) A micro-scale CFD model involving Large Eddy simulation and RANS turbulence models that are used in a nested multi-scale framework with two other established weather models (HARMONIE and SIMRA model), and (B) a non-intrusive reduced order model for urban wind flows involving machine learning. The deep learning based non-intrusive models is a first-step towards faster models for digital-twin platform for real-time turbulence alert systems. This report presents the details of both the models and methodology along with the validation with experimental MAST measurements for a realistic case study (Oslo hospital). A brief recommendation on potential drone safety operations based on turbulence maps is given along with scope of future work. For further details, the readers can read the 3 articles \cite{mandar2020,2T,3T} that are published as a part of scientific dissemination from this work. This work was conducted based on Research council of Norway funded "Aerial transport of biological materials" project (project number 282207).