According to myriad articles, studies and white papers, analytics is now “the topic” for any business application, from supply chain optimization to pricing and distribution. But while this firehose of news and information has eased the way for applications in industrial operations by gaining the attention of the C-suite, it has made it more difficult to navigate the path to a successful implementation.
“It presents a major opportunity, but also major confusion. Emerson favors a practical, pragmatic approach,” said Peter Zornio, chief technology officer of Emerson’s Automation Solutions business, in a press conference at the 2019 Emerson Global Users Exchange in Nashville, Tennessee.
Analytics has the potential to deliver more than $4 trillion of growth in industrial manufacturing, according to Gartner Group. But potential users are confused, asking questions like where to start, what supplier to use, what types of models to apply where, what types of problems it can solve, and how OT analytics fits into IT.
“On the supplier question, one of our customers identified more than 900 resources,” Zornio said. Fewer are involved in industrial IT and OT, focused on plant-level benefits. For a manufacturer, those are the biggest opportunity, with potentially high-return applications in productivity, reliability and energy efficiency.
Emerson’s portfolio of operational analytics focuses on the greatest source of value for industrial manufacturers—the production process itself. Operational analytics with embedded domain knowledge can impact and improve performance of simple equipment, complex assets and process units, and entire production plants.
“We recommend addressing the high impact, known problems first,” Zornio said. “By using proven models that make analytics accessible to the personnel responsible for the performance of assets, our customers can act quickly to solve problems faster. For example, Emerson’s solutions can detect and address 80% of the equipment failure modes contributing to production loss in a plant in real time.”
Analytics can be broken down into two classes: traditional and data-driven. “Traditional analytics are principle-driven, where you know the mechanisms—the mechanistic models,” Zornio said. You know that equipment and units are designed a certain way, so these analytics can be rule-based: if something goes wrong, you probably know the cause, for example, by failure modes and effects analysis (FMEA).
Data-driven analytics build a model from analysis without knowing the physics, using standard statistical analysis, Zornio said. “Here is where advances in computing have driven excitement, with machine learning (ML), enhanced pattern recognition and mathematical correlations.”
Emerson’s enhanced portfolio includes ML and artificial intelligence (AI) that can be used to identify new discoveries and deepen insight to impact business performance. These tools provide perspective previously unattainable with traditional analytics.
Deriving the benefits
Plants are complex systems, with components that roll up into assets that become process units, a whole plant, and often, a fleet of plants. Users ask, “Where is the opportunity? How do we apply analytics?” Zornio said.
“We already have analytics we can apply to lower-level assets. We need to do more at the plant level,” Zornio said. “If we have knowledge about the plant—as it is, not just as-built—we can use first principles. Then we need to get the analytics output to a person who will implement change based on the results—people who will actually do something.”
Where first principles are not understood, or not enough, it makes sense to turn to data-driven analytics. “The question about data-driven analytics is, why use which where?” Zornio said. Using a car as an example, “It takes until the third time you run out of gas for machine learning to learn cars need gas to run,” he said. “We have more than 6,200 equipment models with 500 FMEAs. Some 80% of equipment can be done using existing first-principles analytics.”
Before turning to data-driven analytics for a product or equipment class, decide if you can use packaged analytics. “Some engineers want to develop their own analytics, but a known answer to a known problem is probably a better solution,” Zornio said. You can hire a data scientist, but it’s probably more important to have someone familiar with the equipment.
“We know our devices and their operational analytics,” Zornio said. “We started building solutions into DeltaV in the 1990s, with fuzzy logic and neural nets. We’ve added simulation and digital twins, and now we have a generic toolbox for data-driven analytics and AI.”
With the company’s acquisition of KNet and its integration into the company’s Plantweb Optics asset performance platform, “Emerson can provide not only some of the most advanced machine learning and AI tools in the industry, but also the connection to people and workflows, which are critical to digital transformation success,” Zornio said.
Emerson’s portfolio now provides both pre-packaged analytics solutions as well as a complete analytics toolbox for users to develop their own applications. This portfolio is supported by Emerson’s Operational Certainty consulting practice and robust data management capabilities that provide a foundation for analytics success.