Using artificial intelligence and machine learning, IoT platforms can do a better job of monitoring and securing networks.

April 26, 2021

7 Min Read
Industrial IoT platforms
Getty Images

By Rich Castagna

 The Internet of Things’ killer app might be artificial intelligence.

While it may be a stretch to classify  artificial intelligence (AI) and its multifaceted offshoot machine learning as true applications, these techs can profoundly change IoT operations. AI makes IoT networks smarter and able to scale as needed without the risk of uncontrollable growth.

IoT operations is an ongoing struggle to try to ensure that the thousands or more devices run properly and safely on an enterprise network and that the data that’s being collected is both accurate and timely. While the sophisticated back-end analytics engines do the heavy lifting of processing the steady stream of data, ensuring the quality of the data itself is often left to somewhat archaic methodologies.

To help rein in sprawling IoT infrastructures, some IoT platform vendors are baking in AI/ML technology to boost their operations management capabilities. Some notable platform vendors, such as IBM and Schneider Electric, have already logged years of experience integrating AI/ML into their products, but the use of AI/ML is far from universal among all IoT platform purveyors.

“I would say across the hundreds of IoT platform vendors out there, it’s still a fairly rare phenomenon,” noted Sam Lucero, chief analyst, IoT services and technologies, at analyst firm Omdia. “It’s still a developing feature in the solution sets.”

Why IoT Platforms Need AI/ML

Despite the limited product rollouts to date, there’s ample evidence that AI/ML will be a necessary ingredient in most IoT platforms. Traditional management tools can meet the demands of larger IoT environments, as they are unable to keep up with the sheer size of the networks and the growing number of devices they link.

Current tools like SCADA systems may be able to provide basic monitoring of sensors, actuators and other connected devices, but the information they receive is basic at best. Typically the data is based on predetermined thresholds, with little or no qualitative distinctions.

Joe Berti, vice president for AI applications at IBM, sees aging SCADA environments as a key motivation for upgrading to AI-infused IoT management.

“Just because there’s this massive infrastructure of SCADA systems that collect data for utilities, oil and gas, and manufacturing, and they’ve been collecting data for 10 to 15 years,” said Berti, “but they’re based on set points.”

Such manual processes—specifically establishing the points at which data collection operations turn from “good” to “bad”—is one of the key issues that contributes to inefficient and often inaccurate management methods.

Another contributing factor that adds urgency to the adoption of AI is a dwindling workforce across many industries that relies on their IoT environments. The contracting labor force—shrinking as a result of retirements, layoffs and shifting operations overseas—is leaving an expertise gap that can be mitigated with the help of smarter management systems.

<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

For more on IoT platforms, check out the Omdia report “Connectivity Management Platforms – 2021 Analysis.”

What AI Can Do for IoT

Platform-based AI is focused on the data that flows through the operational plane to ensure that data collection and other devices are operating efficiently. Platform-based AI doesn’t affect the data that’s collected for analysis.

It’s an important “distinction between the data about how your system’s operating and the data that your system is providing,” said Omdia’s Lucero.

On the analytics side, some applications–typically cloud-based—have also integrated AI technologies, but those are distinct from the operationally oriented platform implementations.

With AI—particularly machine learning—the operational health of network devices can be monitored based on real-time data and tracked over a period of time so that a range of parameters can be analyzed. This approach offers more and more specific information about how the devices are operating compared to less informative performance measured against preset benchmarks. In some cases, feeding already captured operational data into a machine learning engine will increase its breadth of experience and allow it to provide even more granular information.

The real-rime aspect is critical as well. Today, many IoT administrators are overwhelmed with the sheer amount of information that their networks are yielding. IBM’s Berti said customers are clamoring for help, and noted that many of them say, “We’re getting thousands of alerts and so we just we can’t pay attention to them–this is noise and it’s too many for us to deal with.”

IBM’s solution, said Berti, can handle the onslaught of information and parse it for the truly meaningful data points: “It’s basically AI-based anomaly detection,” said Berti, “and really what we’re finding is what’s really operating differently here?”

That level of data collection and analysis provides considerably more insight into network performance.  “What we’re talking about is trying to, for instance, detect anomalies or detect usage patterns and then be able to say, OK, let’s operate differently,” said Lucero. “Let’s change these operating Instructions because we’re getting this data that we’re processing automatically and we can operate more efficiently as a result.”

Schneider Electric provides AI capabilities “fully integrated as an option” accordingly to Martin Bauer, Schneider’s EcoStruxure marketing manager, who responded to IoT World Today’s questions via email. “Customers have the full flexibility to run EcoStruxure Machine Advisor to collect and display data [collected from] machines or to add the analytics option for predictive maintenance.”

IBM’s implementation doesn’t use AI to just detect anomalies, it can also initiate activities based on that detection. “We actually close the loop,” said Berti. “We can create a work order inside Maximo and then have a technician go look at the equipment.” The technician can use a mobile device to see the information along with suggested remediations.

AI Aids IoT Security Too

With better data received and analyzed faster, security systems and system operators can react more quickly when a perceived threat appears.

Without AI, a security or management system might generate only an alert if a device fails to continue to operate and collect and transmit data. But AI/ML can detect the subtleties of device operation which may indicate that a device that is apparently operating properly is acting doing so in an anomalous manner—perhaps collecting data when it’s not expected to or operating outside of its temperature range.

“On the control plane, the use of ML is a type of anomaly detection, improving security as a result,” Lucero said.

IBM’s Berti noted that the information gathered and acted on by AI-aided management, can help to isolate segments of the IoT network and thus reduce vulnerabilities and potential attach surfaces for interlopers.

Schneider’s EcoStruxure platform also taps into its AI expertise to bolster network security. “Cyber security is one of the most relevant aspects in the development of our offering,” wrote Schneider’s Bauer.

Little Accommodation Required to Add AI to IoT

Some users might balk at implementing or upgrading to an AI-enhanced IoT platform, assuming that such state-of-the-art software technology will require equally sophisticated hardware, which would mean extensive—and expensive—device upgrades.

But that’s not necessarily the case.

“I have not heard of any special modifications needing to be integrated or developed on the device itself,” said Lucero, “and really if there was for the vast majority of IoT devices that would be sort of a deal-breaker right off the start.”

The same goes for the format of the data the devices transmit and the protocols they use to move the data long. Most AI-capable platforms can collectg and interpret data in a variety of familiar formats using tried-and-true transmission protocols.

“We can actually accept any type of data,” Berti said. “What we’ve done is we’ve written connectors to the major SCADA systems.”

Getting up and running generally isn’t all that difficult either. As noted earlier, some AI/ML systems benefit from being able to ingest and analyze historical data, but there’s typically little training required for the systems or the operators.

AI Accelerates IoT Market

There’s little question that AI has become an integral part of IoT operations management. Larger IoT installations will see the benefits of AI sooner than smaller installations simply because of the scope and challenges of operating a large and complex IoT environment. And while today the array of AI-enabled platforms is limited, that will soon change.

“We already see a consolidation of the vendor landscape under way,” Lucero said. “I suspect that AI/ML is going to be one of those things that helps to speed that process along.”

It’s also possible—although not happening today—that vendors of AI-enhanced platforms will make some of those AI capabilities available to other applications via APIs or other integrations.

“I’m sure that would be exposed along with other features and functionality,” Lucero said, “but I think that that is again a bit further down the field in terms of direct integration with the IoT platform.”

 

 

Sign Up for the Newsletter
The most up-to-date news and insights into the latest emerging technologies ... delivered right to your inbox!

You May Also Like