A2.4 Modeling Study of Hydrogen and/or Ammonia Combustion Safety

Cosmonautics does not stand still, designers set themselves new, increasingly ambitious tasks – and governments allocate fabulous money for their implementation. However, all the work and money invested, and most importantly, the lives of astronauts can be at great risk if they do not thoroughly work out all aspects of flight safety: protection from space debris, fire hazards in habitable platforms, reliability of power plants, emergency situations during takeoff and landing, reliable relay systems, computer software, serviceability of communications and much more. Currently, there is an active search for new types of fuels characterized by a low carbon content in the emitted combustion products. Hydrogen is the most environmentally friendly fuel, which is why hydrogen energy has become one of the promising areas for the development of the planet's energy complex. However, hydrogen mixed with air produces explosive gas, which is extremely explosive. Therefore, the efforts of scientists in many countries are aimed at creating combustible mixtures based on hydrogen, which would not be as explosive as mixtures of hydrogen with air. The design of new-fuel engines requires knowledge of chemical kinetics to understand ignition, flame propagation, and pollutant release. The behavior of combustion and emissions in gas turbines is a manifestation of the nonlinear interaction between several physical and chemical processes that are controlled by the configuration and operation of the engine. Therefore, it is very important to have reliable experiments, high accuracy and computational efficiency of a platform for mathematical modeling of hydrodynamic processes in chemically reacting gas streams. Modeling gas-dynamic processes in turbulent flows, taking into account the multi–stage chemical kinetics of chain reactions, is a rather resource-intensive task even for modern supercomputers. At the same time, as shown by previous studies, modeling detailed kinetics can take up to 80% of computing resources in the process of integrating a rigid system of differential equations. Therefore, in order to reduce the need for resources and, consequently, the computing time, it is planned to develop an architecture and train a neural network for predictive modeling of the kinetic chain reaction mechanism, followed by embedding this block into the gas dynamic modeling system. The time spent on creating and training a neural network will more than pay off in the future as a result of reducing the time of computational predictive modeling.