AI designed a wedge-air rocket engine from scratch in three weeks and it works

Artificial intelligence has reached new heights—in three weeks, from scratch, it designed a working wedge-air rocket engine, around which rocket scientists have been going in circles for more than 70 years. The model of the liquid rocket engine, then printed on a 3D printer, worked for 11 seconds in the fire test, developing a thrust of 5 kN.

Image source: LEAP 71

The engine design is carried out by LEAP 71, a company registered in Dubai (UAE). The German company AMCM is engaged in the production of its parts from copper alloy CuCrZr using additive printing and laser melting. Testing is carried out at the Airborne Engineering site in Wescott, UK. Earlier this year, LEAP 71 showed a prototype of a conventional liquid-propellant rocket engine, also designed by AI and manufactured on a 3D printer. The AI ​​had to work hard on the wedge-air rocket engine. If he designed a regular engine in two weeks, then the wedge-air engine project took three.

Wedge-air rocket engines (aerospike) were proposed in the 50s of the last century. They are interesting because of their partially open nozzle, which allows the flow of oncoming air flowing around the rocket to serve as a virtual second half of the nozzle. This means that the curvature of the nozzle will change as the rocket rises due to the gradual thinning of the air. It follows from this that a wedge-air rocket engine will be equally effective at all altitudes, while engines with a conventional nozzle are effective only in certain parts of the flight, which is why the rocket has several stages with different engines.

Interest in aerospike engines has returned amid the design of reusable rockets and spaceplanes. In a good way, the plane should not be multi-stage. Finally, wedge-air rocket engines should generally consume less fuel to deliver cargo into space. In light of the fight against warming and saving money in space, this is also important.

The LEAP 71 company has created the Noyron neural network, which is taught to design mechanisms and any design solutions without the use of CAD programs. The company has successfully shown AI working in rocket engine design, but also claims that Noyron is capable of designing not only rocket engines, but also toys and heavy equipment. The program is given input parameters, and the output is a finished device. It looks like another profession is facing transformation under the pressure of AI. This time it is the work of a design engineer, although people are still doing a good job even designing wedge-air engines, if necessary.