Artificial intelligence-based model to predict the heat transfer coefficient in flow boiling of liquid hydrogen as fuel and cryogenic coolant in future hydrogen-powered cryo-electric aviation
Shahin Alipour Bonab, Mohammad Yazdani-Asrami
Abstract
• Liquid hydrogen seems to be the solution for decarbonization of aviation sector. • Safety of hydrogen aircraft relies on the storage tank and pipes reliability. • Artificial intelligence is a modern way to prediction heat transfer in LH. • Cascade Forward Neural Network predict the HTC of LH with 99.88% accuracy. • The results were extensively compared with existing Empirical correlation equations. Hydrogen is considered an environmentally friendly chemical energy carrier that can help the aviation industry to decarbonize and achieve Net-Zero without notable change in the configuration of the conventional propulsion engines of existing aircraft. In addition, cryo-electric aircraft which takes advantage of superconducting power systems seem to be an additional solution. Hydrogen can be used as a cryogenic coolant to cool down superconducting devices (including propulsion motor, cable, and fault current limiter) to a temperature of around 20 K. For both purposes, hydrogen must be stored in the liquid phase to make it dense and suitable for storage in the fuel tank of hydrogen-powered cryo-electric aircraft. As the unpredicted gasification of the hydrogen in the pipes is hazardous for the safety of the aircraft, it is of paramount importance to evaluate its heat transfer with the pipes to ensure that hydrogen temperature, pressure, and volume are under control during the flight. Advanced artificial intelligence (AI)-powered models can capture the complex patterns of the heat transfer coefficient (HTC) by considering numerous influential factors. These models can accurately estimate HTC within milliseconds, making them a safe and efficient solution for monitoring applications. Compared to less accurate empirical models or slower Computational Fluid Dynamics solutions, AI-powered models offer superior performance. For this purpose, a new AI model based on the Cascade forward neural network has been proposed to predict HTC with high accuracy, using the experimental data as the input of the model. The results demonstrated that the proposed model could predict HTC with 99.88% accuracy in terms of R-squared.