IoT Sensors Monitor a 1,400-Kilometer Canal in China | Hacker News Reader

IoT Sensors Monitor a 1,400-Kilometer Canal in China

spectrum.ieee.org

74 pointssohkamyung33 comments

> Although some sections of the canal had access to a fiber optic internet connection, other areas had no access and passed through remote areas with spotty or absolutely no cellular network service. As a solution, Yang and his team developed the so-called Smart Gateway to receive data continuously from local sensors and then transmit it to a cloud server using whatever signal was available at the moment. That could include fiber, Ethernet, 2G, 3G, 4G, Wi-Fi, or Zigbee.

If you're building such a massive infrastructure project, why not just run fiber along its length?

Or design devices with P2P in mind. The devices with a network connection forward on data received from peers.

I always wondered about that myself. I've seen people discussing what the best technology is for monitoring remote oil pipelines, should they use cell spectrum and ensure every part of the pipe is covered, cell and ensure each transmitter can connect m2m to relay messages to where coverage is, or some other tech...and no one suggested "Hey...why don't we just run fiber along the pipe?" There may have been a good reason that wasn't obvious to me...or it may have been people legitimately didn't think of it, too enamored by the idea of connectivity in a "remote" location.

I could think of a few potential reasons:

- Maintenance: It could be very difficult to find and fix a fiber break. AFAIK, there's no easy way to determine where a break in the fiber is, you generally end up replacing the entire broken segment, which in this case could be miles.

- Reliability: If the fiber breaks, all the upstream sensors go down. Especially since a fiber break could also mean a pipeline break, you just handicapped yourself at the most critical time.

- Cost: Hundreds of miles of fiber, properly shielded from the elements, must cost a lot. (Of course, getting wireless signal to thousands of sensor stations would also be pricey, and I can't say how they'd compare.)

If you're interested, optical time-domain reflectometers are used to characterize fiber cables, including breaks/bends/faults/etc. Pretty standard and relatively cheap test equipment.

I was curious how accurate those are, found a datasheet for the PRO-OBL-201A ($1600, or effectively nothing compared to an oil pipeline). It's a portable "optical break detector", full blown reflectometers might have more features.

They list the maximum range as 100 Km, with accuracy of +/- (0.8 m + 0.001% x Distance). So measuring to a break at maximum range, you'll be within 1.8 meters.

iirc, tunnels use this to know where exactly fires occur.

None of those are a problem if you're talking about building a canal. From my previous experience, 80% to 90% of the cost of building a fiber network is construction, especially digging ditches, hanging poles, etc. Materials account for the remaining 10%-20%. RoW costs are not included.

IIRC, the materials cost for building long haul fiber is a couple dollar per meter or less, especially in China.

If you're already building a massive structure like a canal, embedding conduits into the infrastructure is a tiny marginal cost. And Fiber maintenance is very, very low in protected environments (such as the electric-fence protected canal).

Not only that, you could easily build two conduits, one on each side of the canal, creating route diversity.

So for laying hundreds of miles (let's say 1000km) of fiber where construction is already happening, the total cost would be a few million USD.

You mention maintenance of the fibre, but I think the bigger reason would be maintenance of the pipeline. Pipelines require frequent maintenance and inspection, and expecting the workers doing that maintenance to avoid slicing the fibre is probably optimistic.

When the cable breaks you then loose all upstream sensors. Pipes sometimes explode, waterways burst. So it isnt ever just one cable. You need many levels of backup and as a general rule want the data moving on a network not folowing the pipe.

You could use a bidirectional ring network and have ptp wifi (canals tend to be mostly straight) as a back up and use GSM or sat phone as a tertiary backup.

So as kind of a broad reply to a lot of the "well, the cable could break" - yes. Yes it could.

I'm not saying don't have some other way of transmitting data...just rather than worrying about "how do we relay connectivity between senors using a wireless protocol", and/or "How do we connect sensors that aren't in cell coverage areas to the broader internet", fiber can be used to connect sensors in relays. Put a cellular chip every 10 miles, say, and then connect those, with sensors, with fiber. So basically cell chip -> sensor -> sensor -> ... -> cell chip -> repeat. Any single split of one of those 10 mile stretches, and you have lost nothing (as communication can still go the other way, that is, the sensor north of the split can still send data north, the sensor south of the split can still send data south, say). And you know where the split happened.

"Workers will just cut the fiber" - err, really? I assume installing oil pipelines, given the mechanical complexity of it, involves more than just sticking it together and welding it shut (and removing it involves more than just slicing it open and dropping a new one in its place). An extra step or two to ensure the fiber line is properly hooked up hardly seems any more complex than what they're already doing.

I wonder if they couldn't find a LoRaWAN[1] option that would work for their hardware? It seems better-suited than most of the alternatives.

1: https://en.wikipedia.org/wiki/LPWAN#LoRaWAN

the only problem with LoRa is that it is proprietary... but I just did a quick search and there appear to be open source alternatives now, like this: https://www.cablelabs.com/cablelabs-announces-open-source-lo...

I think that's only some of the software stack, LoRa transceiver chips themselves AFAIK still only are made by a few companies, on license from Semtech? I think even with that, LoRaWAN is one of the more open and easy to get parts for LPWAN technologies though.

Maybe something like a LoRa network might wind up being more durable and flexible over the long term?

If you have a tiny low-energy sensor that only wakes up for a few milliseconds out of every hour and can be powered by something like a little peltier element or turbine or pile-driven geothermal shaft, what's likely to go wrong besides someone stealing it? And with the design worked out, how much would a unit cost, $100 even with a rugged enclosure and special pricing for government jobs?

You'd only have to maintain a handful of base stations as opposed to umpteen dozen fiber terminations, and you could freely move sensor nodes or install new ones depending on what you learn or changes to the program.

But that said, it sounds like these sensors are A) along an existing infrastructure project like you said and I misunderstood, and B) also need to monitor for intruders which presumably also involves general surveillance. So...yeah, what gives?

Well, whatever, it looks cool.

Isn't that the fallacy that car salesmen use to sell overpriced extras. "If you're already spending $20,000 on a car, what difference does a $500 stereo make?" The complicated mixture of wireless networks might be cheaper than fiber. Looking at it another way, why would anyone use wireless for a fixed installation when they could "just" lay wires around their house or up their street. They're already built an expensive house or street.

I am pretty sure they did, and most of the article is bullshit.

Probably because they'd want to minimize correlation between the thing being monitored breaking and the thing monitoring it breaking.

1. The IoT network is built after the canal. 2. The South-to-North Water Project is not a single entirely new canal. It consists of many small canals and dams that connect existing canals and nature rivers.

It sure sounds like the definition of what is IoT has changed since the last time I checked. I always thought IoT was about attaching sensors to objects in daily life, such as in clothing, appliances, food packages, doors, etc. Building sensors along a canal to me sounds like a regular surveillance network.

The project is still innovative though.

> It sure sounds like the definition of what is IoT has changed since the last time I checked...

It’s almost like some kind of system for supervisory control and data acquisition...

IoT terminology gets used for a lot of different things, including fairly traditional SCADA. It does tend to be different in degree than historically--more sensors, more analysis, more real-time control, but it's certainly different from wearables, SmartHome, or even the broader set of sensors that go by things like smart cities.

I'm going to start calling my PLCs IoT devices from here on out.

That's a mouthful. There should be an acronym...

They solved the problem of connectivity but there are also many other problems many related to the data they receive from the sensors. It is probably not enough to simply store or monitor this data. Analyzing such amounts of data could be the next challenge.

its not exactly high data rates we are talking about is it

You've got that backwards. Things (i.e. things other than general purpose computers) were connected to the internet in industrial settings long before they were in homes.

There's an entire section of IIoT, the Industrial Internet of Things that most people don't know or think.

I'm guessing they called it IoT so more people would click on the title.

Hmm, dubious claim here with the "Smart Gateway"

>> “The Smart Gateway can learn the availability of the connection to the cloud. After a successful transmission, it will follow that network next time. Otherwise, it will try another one,” says Zhang.

Considering,

>> Video cameras were spaced every 500 meters along the entire structure.

And,

>> Yang and his team developed the so-called Smart Gateway to receive data continuously from local sensors and then transmit it to a cloud server using whatever signal was available at the moment. That could include fiber, Ethernet, 2G, 3G, 4G, Wi-Fi, or Zigbee.

I'd be interested to see how their network holds up when streaming video to a gateway via Zigbee or video to the cloud over 2G (or even 3G). Considering the system is supposed to go into "continuous stream" mode during a natural disaster or other emergency, I'd guess this is also when the network would collapse. Oof.

Adding IOT is just shameless pandering for clicks why not add bitcoin to it :-)

I helped build similar devices (custom S100 based systems) 30 years ago at BHR group for remotely monitoring experiments at remote sites and using sensors to monitor pipelines was standard back then.

Wish I had suggested we build our own modems out of BYTE for £30 instead of paying £300 for answer and £600 for answer originate.

It's an interesting project, I really wish that they would have talked more about how the system has been implemented. As in what system are they using and how close to "real time" are they getting the data? Those are a HUGE number of sensors and without even considering how they are getting the information back to any gateways, it would be a gargantuan task to build anything out that would be easily digestible my a person in a Control Room.

I would have thought that with all the effort to build that canal that they would have wired it all. it would have been a great opportunity to have done so considering all the effort expended to create the canals and dams.

is there some reason this was not undertaken at the time of construction?

Reply on news.ycombinator.com