AI-backed distributed acoustic sensing to monitor cities

AI-backed distributed acoustic sensing to monitor cities

Interviews |
By eeNews Europe

If big data is the new oil, as it has often been said, then Fotech offers the right tools to dig for more of it and feed the insatiable needs of tomorrow’s smart cities. eeNews Europe caught up with Fotech’s product line director Stuart Large about what the company considers the untapped potential of distributed acoustic sensing (DAS) for smart cities.

An advanced variant of telecom’s traditional Optical Time Domain Reflectometers (OTDR), distributed acoustic sensing relies on the analysis of the coherent Rayleigh backscatter noise signature in a fibre optic cable as pulsed light is sent into the fibre. DAS is optimised to measure small changes in the coherent Rayleigh noise structure that occurs from pulse to pulse. Since the coherent Rayleigh noise structure is generated interferometrically within the fibre by the relative locations and strengths of local scattering centres intrinsic to the structure of the glass, even very small physical (acoustical or vibrational) disturbances at a point in the fibre can make detectable changes in the interferometric signal.

Schematics of the LiveDETECT II dual channel technology,
delivering up to 100km of continuous monitoring from each
individual module.

Of course, special algorithms and processing nodes must be put in place to make sense of such faint changes, but the beauty of DAS is that it operates with standard telecommunications grade optical fibres, either tapping into existing telecoms infrastructures or into new purpose-laid monitoring fibres, effectively turning these fibres into a linear array of discrete vibration sensors.

DAS goes way beyond what traditional OTDRs can, stresses Large. “A telecom OTDR merely tells you what splices you have along the fibre, what connections you have and the quality of these splices, it doesn’t provide sensing. That same fibre can send vibrations from cars travelling along the road, from which can be derived the flow rate, the velocity, the number of cars and possible accidents. DAS is a far more sophisticated OTDR. There are about 20 different companies stating they have DAS, but only a small bunch of them has the software to actually extract meaningful information from the collected data”.

“DAS generates about one Terabyte of data per hour. You need an edge processor to go through that data and recognize the acoustical or vibrational signature of an event, such as a pipeline leaking, someone digging or a car passing by, and then report about it. AI requirements depend on the applications, whether it is to detect an anchor nearby a sea cable or someone jumping over a fence in a particular environment. In a duct or a buried pipe, the acoustic properties of the environment and the acoustic signatures are different. You need to train AI algorithms to better recognize what’s happening before setting up an alarm. The alarm can then be integrated with third party systems to show the event on a map, to pan nearby CCTV cameras to the location or even to send a UAV and get an early view of what is happening, shining light or issuing a warning” continued Large.

The Helios DAS interrogator carries the optics connections.

Talking about the edge processor, Large says it must cope with a huge amount of data, just as if it were processing data from thousands of microphones simultaneously. Algorithms, which Fotech refers to as Detection Streams, process the data in real time to recognise the signatures of the particular events they are interested in. There may be one stream detecting walking, another for digging etc… The Panoptes Alarm system then considers the events being reported by one or more DAS systems and applies logic to decide when an alarm needs to be raised.

But surely, tapping existing fibre networks for such valuable data may not remain free for much longer, especially when such networks belong to telecom operators.

“The economical business model is still being explored, and we are talking to telecom providers”, acknowledged Large. “When realizing the potential value of the data, some may want to rent out their fibre network to us on a monthly or an annual basis or provide data as a service to some of their customers”.

What Large envisages is that the insight provided by one DAS on one fibre could be sold to more than one customer, each consuming a different stream. That may be traffic flow management through speed control via smart traffic lights, or perimeter intrusion detection for specific zones, or even for the telecom companies themselves, monitoring their own networks. In that case, when excavation work would come too close to the buried fibre, the DAS would generate an alarm so someone could be sent out to stop the work and eventually reroute some of the traffic.

“We want to bring forward the know-how we have gathered from our established businesses in pipeline and cable monitoring to smart cities, where millions of kilometres of fibre are already laid in place”, explained Large.

The LiveDETECT system rack combines the Helios DAS interrogator
with UPS, KVM, Network Switches and at certain points, the
Panoptes Monitoring Module.
Logic is applied to decide when an alarm needs to be raised.
Alarms start with Green and advance in severity to Amber
and then Red.

“In the oil industry, once pipeline operators have realized the value of DAS for leak detection or intrusion events to within a few metres (such as digging nearby or hot-tapping) simply from using existing laid fibre networks, then they consider installing fibres with that specific purpose for their next projects”, Large noted, adding that a more careful fibre placement relative to the assets to be monitored yields better data.

But aren’t 5G and IoT sensor networks also promising to deliver such services to smart cities? eeNews Europe asked.

“We are still in an exploratory phase, we know that we can deliver certain capabilities which open new opportunities for tracking assets. In smart cities, DAS could discriminate between a train’s front and rear carriages and figure out where it is heading to, providing a stream of data to the control systems. DAS could also identify defects onboard trains, such as a flat wheel. For traffic monitoring, of course there are cameras and radars, V2X and in some cases inductive loops, but what we offer is a continuous sensor along a continuous journey, not just point sensors with blind spots in-between” argued Large.

Even with so-called floating point data provided by the geolocalisation of smartphones along transited roads, traffic accidents are more difficult to establish reliably on small countryside roads. Large also mentioned that such floating point data had been easily tricked in the past, as artist Simon Weckert demonstrated, creating virtual Google Map traffic jams by simply pulling a cart packed with borrowed smartphones using the app, along empty streets in Berlin.

Now being part of BP Launchpad, Large says Fotech no longer has to worry so much about cash, the company is hiring rapidly, it is encouraged to look at the bigger picture. BP Launchpad wants to grow it into a billion dollar business within the next five years.

Fotech –

Related articles:

Stanford researchers turn optical fibres telecom network into earthquake observatory

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles