We could say that it all started with the mythical film of the 1990s Red October huntingthe history of the most modern nuclear submarine in the Soviet Union, thanks to a revolutionary Silent Magnetohydrodynamic Propulsion System, which makes any sound undetectable. And now the Chinese update arrives: Leviathan operationthe story of An American nuclear submarine that uses high -tech acoustic holograms to deceive Chinese torpedoes and their human operators. The difference is that the Chinese version seems to be true.
A team of Chinese scientists, led by Wu Yajun, has revealed that they were working on An artificial intelligence system designed to overcome this type of underwater deception. In A study Posted in Command Control & SimulationThe authors claim that their system had unprecedented precision for torpedoes that moved at high speed.
Tested with data from high -speed torpedo scopes, Technology achieved an average success rate of 92.2 % by distinguishing real submarines of lures, even during tense exchangesaccording to the article.
This represents a great advance with respect to traditional systems that often do not reach the goal. For the Yajun team, the underwater war of the future depends on the deception of torpedoes through illusions. High -tech lures, such as those shown in Operation Leviathan, are used to replicate the acoustic signal of a ship, generate a wake of false bubbles to simulate an emergency turn or deploy in coordinated swarms to project ghost objectives on the sound screens.
These tactics are particularly effective against the so -called ultra -grape supercavitant torpedoes: Weapons that generate cavitation, or steam bubbles, around their helmets to reduce resistance. The resulting roar drowns the real echoes of the target and distort the acoustic footprints.
“The current objective recognition methods for China high -speed submarine vehicles are inadequate in saturated in advanced counter -demured environments, which requires the urgent development of new approaches to the extraction of characteristics and the identification of objectives -explains Yajun -. Only high -speed submarine systems, Equipped with long -range detection capabilities and high objective recognition ratesthey can offer enough operational efficacy. ”
The solution they proposed arose from an unconventional combination of physics and automatic learning. Given the scarcity of real combat data, the team began simulating lighter profiles through hydrodynamic models of bubble collapse patterns and turbulence. To do this, They used unprocessed data compiled from the high -speed torpedo test field of the EPL Navy.
These simulations were subsequently added to an “adversary generative network”: basically two AI models that antagonized each other. One of them, the generator, refined the signatures of the lures through the study of underwater physics and acoustic principles. His opponent, the discriminator, trained to detect failures in these falsifications by seven layers of analysis of sonic patterns.
After numerous training rounds, The system created a huge collection of artificial lighter profiless. According to the article, AI uses a specialized neuronal network architecture inspired by image recognition. The Sonar signals are subjected to an amplitude standardization process, filtered by correlation receptors to suppress noise and, finally, are represented as spectral miniatures by means of a mathematical tool known as Fourier transform.
These sonic snapshots pass through convolutionary layers in the neuronal network, optimized to detect anomalies in frequency modulation. Grouping operations average distortions such as bubble interference. The study states that, When facing the most sophisticated lures, detection rates went from 61.3 % to more than 80 %.
“With the continuous advances in modern underwater acoustics, Electronic technologies and artificial intelligence, the current underwater battlefield often contains multiple simultaneous threats Within a single operational area, including lures, electroacoustic countermeasures systems, electronic inhibitors and various weapons systems, ”concludes the study.