The world's first liquid rocket engine designed entirely by artificial intelligence was recently tested in the UK.
The design took less than two weeks after approval specifications, and it took a few more days to 3D print. The engine started on the first try after assembly.
Finishing the parts and assembling the engine, carried out by staff at the University of Sheffield, took the most time. As a result, people were left with only physical work, while the creative task was performed by algorithms.
The project of developing complex engineering structures with the help of AI is implemented by the company LEAP 71 from Dubai. The company's specialists have developed a Noyron computing model with a compact geometric core PicoGK, capable of creating complex physical objects.
The neural network-engineered engine uses a kerosene/liquid oxygen fuel pair. During tests at the Airborne Engineering test site in Wescott, Great Britain, the 5 kN (500 kg) engine warmed up for 3.5 seconds, then reached full power and ran for 12 seconds, developing a thrust of 20,000 horsepower. This is sufficient for use in the upper stage of the rocket. The Noyron model can create a new engine modification in less than 15 minutes on a regular computer.
The engine components, manufactured in Germany by AMCM from CuCrZr copper alloy by additive printing on an EOS M290 printer, supplied innovative cooling system. The kerosene supply through built-in 0.8 mm channels prevented the copper from melting at combustion chamber temperatures of up to 3000 °C, so the engine body only heated up to 250 °C. A failure in the cooling system could have turned the engine into liquid copper, but the system worked without complaint. For fuel injection, a coaxial vortex nozzle was used, recognized as the most advanced solution today.