Wind-Tolerant, Autonomous Drones Could Cut Bridge Inspection Costs by 30%

dvtt@lemmings.world to Technology@lemmy.world – 174 points –
researchinenglish.com
19

Either bridge inspectors get paid a hell of a lot less per hour than I think they do, or the company doing the drone inspections is way overcharging. One relatively inexperienced pilot could fly the full structure of the bridge throw it into point cloud software and drop off a highly detailed 3d model of every inch of the bridge to be inspected from the comfort of someone's desk.

No OSHA, no driving a crane truck, no having to retract and re-extend the bucket around every structural element, no netting, no walkways or temporary bridges needing to be made.

I'm no expert on inspecting bridges, but I'd think that you still would need a professional inspector to do the inspecting, only that they would save the time of actually travelling out to the bridge themselves and instead could do it in their office, no?

And then there are probably things which still need to be done on site, such as non-visual inspections (ultrasound, X-Ray, Vibration testing, Tourque measuring on bolts, paint thickness,...? IDK)

Not just travel time, but also climbing around on the bridge itself.

at the very least they would identify all the parts that need to be looked at in person.

The amount of time it takes a body to hand inspect every inch should be the lions share of the time. if it's not, you need a new bridge :)

Visual inspection is an important starting point to determine if you need more extensive testing. You get a sense of the area, bridge type, and age of the bridge. This would be great for younger bridges that are low risk but should have a visual once over every couple years or so.

30% cost reduction vs current drone systems which can't tolerate wind. I'm guessing more air+recharge time, damaged drones, skilled pilots, etc. all add up

When you do bridge inspection you have to be real close to the structure. And basically stationary as well. You are looking for minuscule gaps and other damages.

You can easily do a 3D map from afar but that won’t have anywhere near the resolution required.

close for a camera is just resolution. drone is there to provide an angle. This is a technical problem , not a time one.

It’s not just resolution

Its also stability/focus and speed

Get the camera with highest zoom you can and plop it in a drone and you will get shitty photos. Drones have a ton of vibrations.

"Its also stability/focus and speed" "Drones have a ton of vibrations."

I think you're conflating 'drone' with a $1000 consumer grade DJI. You can put whatever light, sensor or camera you want any one of a dozen stabilized platforms that are good up to 50 lbs. They're still not any harder to fly, it's still a one person job, you could make it two people and have the second person aiming and shooting you can do the work twice as fast.

"highest zoom you can" I didn't say that, because it would be an intensely stupid statement to make.

If you want sub millimeter resolution of the entire structure, that's doable. You want infrared? FLIR? all doable.

If you used a stabilized 50MP mirrorless with a fixed lens to take pictures from 10ft away, you could get 2.62 pixels per mm. If you reeled that in to 5ft at a time, you'd get 5.25 pixels per mm.

On a gimbal, with a stabilized lens, vibration would be a non issue. But if you want to beat that horse, you have 6 blades at 10krpm. The maximum frequency of the vibrations would be around 1000hz. (6*10000rpm/60) which means if you're shooting faster than 1/1000 there's no time for vibration even if it was completely un-stabilized. That's easily doable with a light source. That said shooting at 800 would be more than adequate.

And there are cameras with higher resolution and faster sensors available. I'm just taking an upper end off the shelf Nikon at the moment to make the point.

5.25 pixels per millimeter does not sound like very good resolution, especially for close inspection... Did you leave out a modifier or am I misreading something here?

Nah I'm kind of tired of chasing these goal posts here. Go find somebody else to argue with.

I didn't move any goal posts, nor did I argue with you, I was really just asking a question, but you seem to have a chip on your shoulder about something so have a good one

First of all the other guy that replied to you wasn’t me. With all that rant you didn’t even notice that.

Second of all I actually work as en engineer in the Drone industry. I’ve written flight control software so I know a thing or two.

Third I actually know people working with drones in that specific industry in two different companies actually. In both cases these drones aim to get as close as possible to their target in order to do the inspection because that’s required. If a bridge has a crack that hasn’t been spotted and is already visible from a far, that bridge is already in a bad state. You are looking at basically micro gaps in your preventative maintenance. And no these aren’t 1k consumer grade djis (though DJI has fucking amazing hardware). I’m talking 5 to 6 figures and sometimes that just a lease.

Lastly that’s all very fancy math for vibrations. You should go work in state estimation if you can cancel it that easily. In the real world you have vibrations across the entire spectrum. Yes there are certain peaks and one of them is related to the motor rpm but it’s an extremely complex problem. Gimbals aim to stabilize the feed from a attitude pov with some damping but they aren’t foolproof.

Oh, no Jimmy, I'm tired of your goalposts moving as well. I'm just more or less over the whole argument honestly where everyone wants to cherry pick every part of the discussion they can without actually arguing my actual original point they attacked.

So, you're a professional 'engineer' in 'the drone industry' whose written 'flight control software'. What are we talking about here? Are you programming flight controllers with your own home written code to handle the PID? Are you writing PC software to interface and locate the drones? Or are you programming dronelink or similar to gather ariel footage for customers? I've been at it since we were putting our own code on stm32's and putting crap together out of spare parts. I've programmed my fair share of control systems and balanced countless PID loops.

So anyway, you're a professional engineer in the drone industry whose written flight control software. Let's assume you actually understand vibration and have some form of civil or mechanical engineering degree.

Who has tight connections with two companies of engineers in drone based bridge inspection,

who's arguing with me that drone footage isn't clear enough due to vibration to the point can't be used in bridge inspection to reduce costs

Then you admit that drones are already being used in bridge inspection, but only at very close distances.

So, what, you're arguing that better equipment can't be do it at a distance of 10 ft? Your argument keeps changing and your credentials seem to be far overreaching for a proper appeal to authority on machine vision.

What is your actual argument with me? My original statement that you contested is that bridge inspections by drone should reduce the cost by a hell of a lot more than 30%. Is that patently incorrect? because there appears to be an actual study out there not done a company trying to sell their services showing a reduction of 70%

Just don't respond. I'm over it. Block me, I'll block you, we can move on with our lives.

You are way too defensive and triggered for the “authority” you supposedly have. And I do believe you. Also you gotta learn to chill and read. I’m not moving goal posts here, you just like to rant. Also big lol at your first paragraph in this comment. Hilarious irony there. Loved it.

Answering your comment and trying to wrap the discussion (I’m not gonna block you lol, what are you even talking about about). What I disagreed with your original argument is that doing a bridge inspector isn’t just getting some semi-pro pilot with an expensive but generic drone with a fancy camera. Snap some photos from a few meters distance and bam, job is done and you inspect one side of a small bridge in 1 flight. That’s it. It was pretty clearly laid out in my comments. To get any meaningful data you basically want/need to get as close as possible. And this usually means actually touching the object you are inspecting. You allegedly have experience in this field so I don’t need to tell you that “regular” drones (and by regular I mean not collision tolerant) don’t like to fly very close to large metallic structures. I’m sure you can guess why.

Not that it matters but yes, I have more than 5 years of experience as a Robotics Engineer working with drones. For my masters thesis I designed, simulated and coded my own control and estimation loops for a drone with tilting motors. You don’t even have to be from a Stem background to tune PIDs. They are stupidly simple but for most cases they are more than good enough. And yes I have a mechanical engineering degree (MSc). Anything that I said is wrong?

I didn’t disagree about the costs reduction. First these drones are still somewhat new in what I assume is an old school field. Also expensive af. Second usually they are usually performed by service providers. If a bridge inspections costs X and the provider can do it with a drone for like 1/4, then he isn’t going to charge that. He’s going to charge 3/4 because it’s a win/win situation. And because of the first point maybe you don’t have that many providers. Also because the drones need to be specialized, and not just any drone for aerial photography or mapping. Lastly regulation. That is always a PITA and more so with drones.

Edit: I actually went and read the article and the 30% are not at all related to bridge inspection cost. It’s basically a 30% reduction in some trajectory planning cost function because it takes into account the wind conditions. So yea kind of a useless discussion

See your way would create jobs and big business want less employees. So expensive autonomous drones it is !

I'm not sure costs are the issue so much as a staggering amount of work for a frighteningly small number of trained workers