AI and ML give old fiber-optic cable a new reason for being

Much can be revealed by measuring and analyzing cable vibrations with AI/ML and distributed acoustic sensing.

In the news

There are about 2.5 billion miles of fiber-optic cable in the world. Most of it is buried, and in addition to transmitting kitten videos at high speeds, it’s got another trait: It vibrates at known rates when it’s disturbed, whether by an earthquake or the pounding feet of a high-school cross-country team. Even better, these cable vibrations can be recorded and measured through a technique known as distributed acoustic sensing, or DAS.

By deploying the proper equipment, scientists can analyze these cable vibrations to identify not only what caused a given disturbance but also when and where it occurred. The potential use cases are nearly endless, as fiber-optic cable is often found in places where other sensors are impractical: under the ocean floor, for instance, or along railroad tracks in remote areas.

Exploiting DAS begins with a device called an interrogator that shoots laser pulses down the cables and analyzes the bits of light that bounce back. Scientists can precisely measure the time it took for that signal to travel back to the interrogator, pinpointing the distance to within about 30 feet.

Aided by artificial intelligence capabilities, DAS is finding a raft of new applications—including monitoring cicadas and protecting elephants from trains.

The Cognizant take

In today’s data-saturated world, it’s perhaps no surprise that one of the challenges around DAS is an overabundance of information. Those cables never stop vibrating, after all, so machine learning (ML) is critical for gathering useful information for any given use case. While a single traditional sensor can gather information wherever it’s stationed, DAS grabs data along the entire length of a fiber-optic cable, and does so 24/7.

Robust ML is required to train the sensing computer on what the resting state of that length of cable is like; what it’s supposed to “listen” for; and what that phenomenon will “sound” like. All this is dependent, of course, on the use case.

Advances in AI/ML in recent years have improved signal processing, pattern recognition and anomaly detection, enabling faster and more accurate interpretation of sensory data. These advances have led to myriad use cases, from improving rail safety to tracking the movement of sea ice.

DAS is being used to monitor oil and gas pipelines. Indeed, it’s helping to enable the sustainable energy transition by detecting early signals of failure—and thus directing predictive maintenance—in wind turbines, solar farms, hydropower plants and geothermal wells. And we’ve written about DAS’s ability to identify problems in electrical grids before they lead to failures or outages.

The impact of DAS could expand further in coming years if industries implement standardized methods, procedures and protocols to collect and analyze data. Doing so would open the door to greater interoperability and cost efficiency. Several challenges remain here, though, including industries’ different immediate objectives, variations in the types and sources of the data they collect and the technical requirements they face. 

Whatever the future holds for DAS, it’s encouraging to watch advances in AI/ML allow for greater exploitation of an existing resource—that’s right under our feet.