In a recent demonstration of the vendor-agnostic Smart Manufacturing Innovation Platform (SMIP) from CESMII, project partners first helped managers of North Carolina State University’s water purification plant get off the dime and analyze the data they were collecting with smart instruments.
“Automation has been going on for more than 100 years,” said Niels Andersen, co-founder of project partner ThinkIQ. “We have a lot of data collecting, but very little data being used. Data does not have meaning, does not have context and is not being put together in ways that we can do analysis.”
Other key parts of the project include ensuring all data-collection equipment has fully qualified names, using graph databases to map diverse relationships, focusing on one standard (ideally OPC UA), ensuring that different systems and protocols can communicate with each other, and developing better virtual models of equipment.
The end result will be a SMIP that standardizes smart manufacturing deployments and replaces custom jobs with repeatable, manufacturer-agnostic, easy-to-use platforms that will enable manufacturing plants of any size to leverage all of their data.
CESMII (the Clean Energy Smart Manufacturing Innovation Institute) intends to have a strong version 1.0 of the SMIP by the end of this year, said Jonathan Wise, VP of technology at CESMII.
To get to predictive analytics at the NC State plant—considered a manufacturing-like environment— project partners ThinkIQ, Savigent, Seeq, Syspro, Semiotic Labs and Microsoft focused on integrating multiple technologies from separate silos and upgrading a PC that was monitoring water quality, said Tim Shope, director of digital transformation at systems integrator Avid.
Before the upgrade, basic tasks like pulling up a week’s worth of historical data took too long.
Next, the ThinkIQ team and others deployed an analytics engine from Seeq to pull real-time data from the programmable infrastructure environment in OSISoft, analyze the membranes on reverse osmosis filters, combine all the data using CESMII’s SMIP, and set a condition level for when membranes need to be replaced—an action that would automatically kick off a work order through the Savigent software, Shope said.
The next step was to develop a model for the high-purity water pump, a pump that is so mission critical that a spare pump sits beside it for a quick, NASCAR-style switch out in the event an alarm indicates a pressure problem or other issue, he said.
An IoT sensor monitors the pump remotely in a machine learning system. After a couple of weeks, it learns the signature wave form of the pump so it can detect if the pump is performing as designed or if it is starting to experience mechanical or electrical problems.
The new platform can predict a motor failure almost four months in advance, he said.
The partners also gave each pump a separate name. “If all the pumps are labeled Pump_01, we have no context,” Andersen said. “It’s critical to give complete names and to have a wide variety of tools to provide meaning.”
Ideally, each pump would be named in a way that identified the name of the machine, the name of the line the pump is on, the name of the plant, the supplier who made the pump, Wise said. A few manufacturers offer a fully qualified naming system but many do not.
In a later phase of the CESMII project, this mapping will be automatic, he said.
CESMII in April launched a new Smart Manufacturing Innovation Center (SMIC) at NC State. It is the first of CESMII’s four SMICs to have the SMIP layered on. The other three centers are at UCLA, Texas A&M and Rensselaer Polytechnic Institute.
At CESMII’s recent virtual annual meeting, CEO John Dyck said he is looking for more SMICs in, for example, the areas of robotics, automotive, chemical, steel and AI/machine learning.
The SMICs’ aim is to link manufacturers, tech vendors, systems integrators and equipment providers with academia, “demonstrating and driving research and innovation that scales to all of US manufacturing,” the public-private organization said.
CESMII will publish the SMIP as open spec—so that any vendor can use it, Wise said.
“We win if these ideas are adopted elsewhere within the industry. This requires a level of altruism that some manufactures are hesitant on because they’re not sure it will work,” he said.
“The time has come for one level of solutions to be open to posterity so we can focus on other problems,” Wise added.
One of the biggest challenges is that “every facility has done an amalgam of infrastructure, systems, equipment and tools over the course of time,” Mark Besser, senior VP for customer success at Savigent, said. “All of those investments were made very specifically because they were solving a problem or challenge at one point in time. Many of those technologies were custom built to solve those problems. But when you take a broader look across the landscape, you start to notice all the white space that exists in between those technologies.”
In other cases, a manufacturer might want to integrate technologies from different vendors but can’t because these systems aren’t designed to mesh with other proprietary platforms.
The goal is to get all of the available data in play to improve efficiency and reliability without breaking the budget or forcing manufacturers to switch vendors, Wise said.
For example, if a manufacturer is running software from Company A and wants an energy-monitoring solution from Company B, often the only recourse is to take out Company A’s system and install Company B’s.
Added Shope: “We can’t go rip all the vendor platforms out of our facilities today. We need to get to that vendor-agnostic place.”
About 70 percent of the cost of a digital factory project is hardwiring the automation structure, connecting data, and creating infrastructure, Wise said. The remaining 30 percent is creating new value from that data, often by developing apps.
But many times, the entire digital factory budget is spent on setting up infrastructure well before a project to develop applications to analyze data can begin.
With the SMIP, CESMII aims to lower that 70 percent cost so manufacturers can focus budgets on applications to analyze the data, Wise said.
This transformation won’t happen all at once. “It won’t be a light switch,” Shope said. “We have a large manufacturing base in the U.S. This system will need to be migrated and adopted over a period of time.”
Connect With Us
