Just remember that they have teams for a reason
Shenyang next gen combat aircraft thread
- Thread starter Blitzo
- Start date
I just had a crazzyyyy thought (purely speculative, no basis on my part to believe this is actually the case).
What if they’re using neural nets to do FCS development? That might imply a very different approach to the kinds of flight testing you’d have to do to develop the FCS, since the data you’d be collecting for a NN based approach would probably require exploring the flight control envelop limits for training purposes, rather than validating and refining the pre-stimulated parameters you’d normally determine in pre-flight modeling.
Again purely *baseless* speculation on my part, but maybe a fun thought to contemplate.
Airbus plans to use neural network AI to help design flight control systems for the FCAS next gen project. (, )
Given how Chief Designer at Shenyang (Wang Yongqing) has disclosed that his team uses DeepSeek AI to help design next-gen stealth fighters like J-50, I wouldn't be surprised if neural networks is part of the development tools. () Wang said: “This technology is already demonstrating considerable operational value,” he said, “providing novel methodologies and pathways for future aerospace R&D.” Wang emphasized that artificial intelligence is not merely a peripheral tool but is being embedded into core workflows.
I'm not up to date on the state of the art fixed wing controls, I know there are a large body of research into NN based control and a lot of demos on small drones, but I would expect mission critical manned aircraft to still use deterministic control laws.
State space control isn't really that fancy, it just require a lot of high order modelling, can solve a complex control problem holistically and can achieve optimal control for a given set of observers and actuators. Something like J-50 with complex dynamics can be solved easier and better with this holestic approaches than linear methods, but at end of the day I'd expect the general KISS rule still hold.
You don’t necessarily need to run the whole control laws via a NN. What I had in mind was using a NN approach to map your state space. More of a development cadence hack was what I had in mind. Like I said I don’t have any basis for thinking the J-50 is going with this approach. Was just a fun thought.
Can we please post CGs and PS images even of real aircraft in the correct thread unless there is something specific to discuss?
Not sure if I should link these then, as they are CG images. But at least they're of the Shenyang's plane.
More CG images at the link below.
That Tomcat guy seems to have done a lot of 3D models of a lot of planes. But he seems to have zero presence on the non-Chinese internet.
More CG images at the link below.
That Tomcat guy seems to have done a lot of 3D models of a lot of planes. But he seems to have zero presence on the non-Chinese internet.
Last edited:
It is! He also did an F-47 render which I quite liked.