Arthur T Knackerbracket has processed the following story:
The Southampton team developed a "nervous system" for drones using optical fibers to continuously monitor strains and stresses, similar to how nerves relay information in the human body. Unlike traditional electrical monitoring, this system uses light signals, which helps avoid the interference issues common in electronic setups.
The monitoring system works through a technique called "optical speckle," where specific images are projected based on detected strains and stresses. AI algorithms then analyze these patterns to assess potential damage, alerting crews when issues arise.
Initial test flights integrating the nervous system on an undergraduate drone project were promising. According to an aerospace student leading the project, the live data showed that fiber optic technology could significantly extend flight duration by reducing the need for manual inspections.
"The drone was first developed to deliver life-saving equipment like defibrillators in emergencies, but it's served as an excellent test platform for the optical fiber nervous system. What really excited us was seeing the live data from the fiber system. It showed that the technology could keep drones operational longer without extensive ground crews," said aerospace engineering graduate Toby King-Cline, who led the student team.
The researchers believe their self-monitoring system holds immense potential across industries. They aim to commercialize the technology as early as 2025.
Drones capable of continuously assessing their own structural integrity could prove invaluable for applications such as cargo transport, emergency response deliveries, and other sectors that would benefit from extended flight times without frequent landing requirements.
Although initial testing was conducted on a small student drone, these smaller drones don't typically need full inspections between flights. The technology would likely be most beneficial for larger cargo drones that endure greater operational stress.
(Score: 3, Interesting) by PiMuNu on Friday November 08, @09:23AM (3 children)
> AI algorithms then analyze these patterns
This looks like a terrible application of neural networks. One would hope that stresses and strains make a linear or quadratic distortion of any image; which should be easily analysed using conventional algorithms. If the image distortion is highly nonlinear so that a neural network approach is required then I don't really believe the results.
(Score: 3, Interesting) by Username on Friday November 08, @10:37AM (2 children)
I believe AI is being used as a buzzword replacement for computer, and nervous for optical. I have limited experience with optical fiber, but I do remember using a light and special scope to see if I cracked the glass correctly for a splice. If you bend the fiber you can get the light to shine more brightly on one side or the other. I assume it's as linear as that. Probably reads the oscillation created by the fans or jets or whatever like a waveform and does x if amplitude exceeds a thresholds or z for frequency. Though I'm not sure if this would be better than a normal rpm sensor and vibration sensor.
(Score: 2) by PiMuNu on Friday November 08, @12:01PM
It just makes me a little sad.
I pointed out to a grad student who wanted to write a Machine Learning paper (probably about 8 years ago) her definition of Machine Learning just meant "any program that has state" which, in turn, means just about any program at all.
(Score: 2) by VLM on Friday November 08, @08:41PM
Yes in the 90s we got bulk fiber and terminated our own ends on ST/SC connectors and that was the process. Special disposal container for the snapped off fiber ends. I got into that game around ST connector era.
I believe now a days the "cool kids" like buying premade SC/APC or duplex LC or SFP+ cables off amazon.
People who don't do fiber don't understand there's like 50 connectors and they're all expensive and they're all incompatible, but some are wildly more popular than others.
RF microwave electronics are the same. There must be 100 variations on little bitty RF connectors, all terrifying expensive.
(Score: 2) by VLM on Friday November 08, @08:35PM
An interesting display of probably the same physics is blasting a fiber with a VERY expensive OTDR you can see crazy stuff as you move/deform the fiber, even single mode fiber will look "messed with" if it is, indeed, messed with.
For some reason they can't just call it "bent fiber" they have to call it microbends and macrobends. As I probably incorrectly understand it from a former employer, macrobend effect is light leaking out the sides (even very little can be easily measured with modern equipment) so its strongly wavelength dependent vs microbends are more like a microwave interdigital filter or a diffraction grating where multiple little bends, added together, equal something like a cavity, and it acts like an interdigital filter and its pretty broadband attenuation it'll F up all the wavelengths in your passband about equally. So ironically you can get macrobend losses from a single bend, whereas microbend losses take periodic disturbances added together over a length of some fiber. Macrobend is like the single mode fiber is sort of like a lens that focuses light down the bore, but if you deform a lens the loss will increase, so bending a fiber always likewise causes (generally measurable) losses.
Most OTDRs are cheaper and lower performance that what I'm talking about because you only need to "see" precisely enough to find the excavator or bulldozer that dug up and snipped your fiber. The attentuation is about zero until it bounces off some "Caterpillar Mfg Co Yellow Paint" at some unfortunate jobsite. The really expensive ones are used in labs to develop and QC connectors and splices and stuff like that. The kind of thing you used to buy from HP and if you had to ask how much it costs, you can't afford it and it comes with your own personal field engineer to make sure it works.
Anyway, yeah I bet you could cheapie it up and if you don't care about location/distance you just make a cool interference pattern and try to analyze the resulting pattern or more likely change in pattern (more change is almost certainly more bad)