Machine learning optimized fabrication of bi-modal capillary wick structures using high voltage press sinter method (PSM) for efficient and compact EVTOL thermal management

The global need to reduce CO2 emissions and the gradual shift towards a more sustainable means of transportation has resulted in the realization of a new mode of transit- Urban Air Mobility (UAM) as a means of efficiently providing safe, accessible and on-demand mode of transportation. The increase in the road-enabled mode of transportation has created an increase in road congestion, leading to a generation of interest from aerospace companies to capture this emerging market. To realize this need for providing on-demand air service, short take-off and landing vehicles (sTOL) are being explored via conventional take-off and landing vehicles (CTOL) or vertical take-off and landing vehicles (VTOL). With the recent development of distributed electric propulsion (DEP) in EVs, attempts are being made to implement electrified propulsion systems in VTOLs. However, the efficiency of these aircraft depends on the efficient thermal management of the propulsion system. Extensive research and testing in the field of thermal management has led to two-phase, liquid-vapor-based heat pipes consisting of porous wick structures with controlled porosity and permeability are being used for components operating at megawatt power levels. This research deals with the fabrication of bi-modal capillary wick structures through a novel, innovative, in-house developed High Voltage Press Sinter Method (PSM). Wicks manufactured using this technique were examined for their hydrodynamic functionality based on the sintering parameters optimized using a statistical design of experiments (DOE) based study, which was validated and verified using a fully discretized three-dimensional numerical model. The wicks were also characterized by their structural stability and strength using an in-house characterization technique. To check for the predictability and repeatability of the manufacturing process, a machine learning algorithm using Random Forest Regressor has been implemented to predict the functionality of the wick in terms of porosity, heat flux, and weight retention from the input parameters.

Author

• Khan, Talha

Identifier

hdl:10057/30353

OAI identifier oai:identifier

oai:soar.wichita.edu:10057/30353