Making Sense of IIoT Analytics

Jan. 18, 2017
Capitalizing on the growing desire for actionable insights, startups and mainstream automation vendors are flooding the Industrial Internet of Things market with new analytics offerings.

As the Industrial Internet of Things (IIoT) picks up steam, attention is pivoting from connectivity to analytics, flooding manufacturers with a wave of new offerings that all promise to facilitate real business change.

Startups as well as familiar automation providers are pulling together new platforms and tools designed to spin the treasure trove of data collected from plant floor equipment and industrial assets into nuggets of actionable insights that can help optimize decision-making. Much of this data has existed in some form for decades, but it’s primarily been locked away in siloed and incompatible plant floor systems. As a result, the data has never been fully utilized as part of a broader analytics effort to foster predictive maintenance, optimize energy usage of plant floor assets, or to initiate a response to critical events like a water leak or pump failure to minimize lost production.

“It’s really easy to capture data, but to then make that data actionable is where companies are really struggling,” notes Ryan Lester, director of IoT strategy for Xively, an IoT platform provider. “Companies don’t have the right analytics tools to parse through the data and they don’t have access to good algorithms to get insights.”

In fact, according to research by Forrester and Xively, 51 percent of companies are collecting data from connected products, but only 33 percent are leveraging the intelligence to create actionable insights.

That gap explains the recent flurry of new analytics offerings specifically aimed at IIoT and manufacturing use cases. Dozens of startup companies are now pitching IIoT analytics solutions for specific verticals like the oil and gas sector or the energy market while others are even more granular, applying analytics models and machine learning to solve very specific problems—for example, well pump maintenance or wind turbine energy efficiency. At the same time, industrial automation behemoths like GE Digital and Siemens are also expanding into analytics as part of their IIoT efforts, delivering a core platform for collecting, integrating and securely managing Big Data at scale along with limited analytics capabilities.

What’s driving all the activity is that industrial companies see a real upside to marrying IIoT initiatives with Big Data analytics. Forecasts from Accenture and General Electric place a $500 billion price tag on activity in this area by 2020. At the same time, industry consultants like Bain & Co. say analytics solutions that are too generic could miss the mark and not be specific enough to promote early user adoption. In a report on how providers can succeed in IoT, Bain consultants make the case for analytics solutions to target just a few (four or so) industries and for horizontal solutions to align with industry-specific partners to ensure analytics are mapped to real business needs.

“Many times companies are so neck deep in what they’re doing, they miss a lot of what’s happening around them,” notes Nav Dhunay, president and CEO of Ambyint, a provider of specialized IIoT analytics for the oil and gas sector. “They’re starting to understand the benefits of looking at business challenges from a data perspective rather than an intuitive or process perspective, and they’re bringing in external entities to help with that.”

Deep domain expertise
Companies like Ambyint and Maana are among the analytics upstarts targeting particular vertical industries and specific problems within those sectors. Other newcomers like Bit Stew Systems (now owned by GE), Seeq and Sight Machine are touting their ability to integrate, investigate and gain real-time visibility into data from a complex environment, including the plant floor.

Smaller, more specialized companies have a leg up delivering analytics, in part because of their ability to innovate quickly in response to market changes, Dhunay contends. “It’s really the speed of technology that we are talking about,” he explains. “With everything accelerating so quickly, it’s the more nimble companies that can come up with solutions quickly. Small companies may also have a deep analytics understanding that large companies could take years to get.”

Consider Ambyint’s area of focus: An intelligent, end-to-end solution designed specifically to monitor and optimize oil well performance and tackle problems associated with artificial lift. Through a combination of sensors, wireless communications and data analytics, including machine learning, the Ambyint solution is deployed on a pump’s hydraulic lift system and is primed to tackle 10 specific problems related to artificial lift, including leak detection.

Think about a torque problem on an industrial motor—the problem might happen on the same motor used in a different industry use case, but the root cause could vary. “In our space, you’d look at torque data to determine if you’re getting paraffin buildup in a well,” Dhunay says. “You need to have a deep domain understanding to correlate what’s happening in the real world. Our focus on oil and gas gives us a significant advantage over generic analytics platforms.”

The vision for GE’s Predix platform and ecosystem is to go beyond predictive analytics to prescriptive insights that determine the likelihood of future outcomes.

Beyond specific domain expertise, newcomers might also have an edge working with Big Data technologies that are essential for applying analytics to IIoT data at scale, says Tara Prakriya, chief product officer at Maana, an IIoT analytics company focused primarily on the industrial and oil and gas sectors. “The larger providers in this space have built their empires on the analysis of structured data,” she says. “Maana has cracked the problem on how to represent knowledge, whether it’s coming from data or domain expertise or some other type of sources.”

Maana employs patented semantic search capabilities, advanced algorithms, deep learning and something it calls a knowledge graph to extract information from time series data silos as well as domain experts, applications, data warehouses and Big Data stores to deliver predictive—and, more significantly, prescriptive—insights that can help manufacturers maximize plant floor productivity or stoke profitability. “Most companies attack it solely from time series data, which in the case of a four-stroke diesel engine, might give them a window of five minutes before it fails,” Prakriya explains. “The real savings is if you can prescribe a maintenance schedule six months out that is tuned to the situation.”

An open approach
The big automation leaders are taking a different approach, cultivating an ecosystem of small upstarts for vertical or domain analytics and positioning their offerings as an open integration platform for IIoT analytics. That’s what GE Digital is doing with its Predix platform, which offers access to a growing catalog of analytics building blocks—some like generic anomaly detection provided by GE, and other, more specialized tools served up by third-party partners and customers. GE is also adding to its analytics coffers through acquisition—its most recent being Bit Stew, which employs machine intelligence and Big Data technologies like NoSQL to tackle the IIoT data integration problem at scale.

“There are many devices relevant to IIoT applications that GE hasn’t built, and we just don’t have the in-house expertise to cover all of those machines at a deeper level,” says Marc-Thomas Schmidt, GE Digital’s chief architect of the Predix cloud platform and the former distinguished engineer and chief architect of IBM Watson. “We want to tap into the rich ecosystem of vendors who are quite specialized in particular areas of math, who have an understanding of deep learning or neural networks, and who know an industry very well. The whole idea around Predix is to make it the center of a really lively ecosystem, and analytics is the most popular space at the moment.”

Siemens’ MindSphere platform has an open application interface so third-party vendors and manufacturers can create their own individual analytics applications.

A platform approach adds value in other ways, says Jagannath Rao, senior vice president for Siemens’ data services business. He says Siemens’ MindSphere platform as a service (PaaS) provides the device management, connectivity, data storage and infrastructure capabilities that will enable manufacturers to scale IIoT analytics beyond manufacturing use cases to next-generation business.

Siemens is offering capabilities in areas like drive train analytics, energy analytics and machine tool analytics for the MindSphere platform, but it’s also created an open API so third parties can create their own, specialized analytics.

Security is another upside to an IIoT platform, Rao says. “There is no customer out there today that’s not been hacked,” he says. “Security is baked into the MindSphere platform in all of its layers.”

Integration is key
While the big automation vendors push platforms, another category of upstarts is focusing on integration, yet another piece of the IIoT analytics puzzle. One such product is Seeq, which is billed as an application dedicated to investigating time series data. By leveraging Big Data innovations like NoSQL and Hadoop, along with machine learning, Seeq allows for Google-like searching on operational data collected in process historians as well as the contextualization of that data through integration with other manufacturing, asset management and transactional systems.

Seeq integrates monitoring data with other data sets while taking advantage of Big Data technologies that don’t require data science expertise.

Sight Machine is another company tackling the actionable insight problem by integrating, contextualizing and visualizing data throughout the manufacturing enterprise. With velocity and variety two hallmarks of the IIoT era, Sight Machine has come up with an automated intake process that ingests data collected by incompatible factory floor systems while applying expert systems and machine learning classifiers to refine and clean data and contextualize it for subsequent analytics, explains Jon Sobel, Sight Machine’s CEO and co-founder. The company’s visualization, dashboard and KPI capabilities report on asset performance throughout the plant, but also highlight anomalies and show root causes for problems in real time, he says. The insights are also presented through contextualized dashboards that reflect the needs of different stakeholders, including data scientists, plant and corporate managers, and machine operators.

Knowing what specific business problems you are trying to solve before investing in any kind of IIoT analytics tool or platform, staying mindful of long-term architectural needs and strategic objectives, and starting with small targeted projects are the best way to advance IIoT analytics without biting off more than you can chew, Sobel says.

In many cases, it won’t be an either/or choice between highly specialized analytics or a horizontal platform that can support a broader effort at scale. “There’s a place for general-purpose reporting tools that IT can use to help serve various businesses, but when you get into the more operational use cases, general-purpose tools can fall short,” says Franco Castaldini, vice president of Bit Stew. “There is a place for both.”