Prime Polymer Breaks Down Data Silos with a Smart Factory Solution Powered by Cognite Data Fusion®

Summary

Prime Polymer, a chemical manufacturer, wanted to reduce the cost and time of on-site inspections of large-scale facilities by implementing on-site autonomous inspections via quadruped walking robots. However, the company encountered a data silo challenge early in the process and turned to Cognite Data Fusion® to address this issue and enable their smart factory solution.

The Challenge

Prime Polymer’s Anegasaki Plant has two manufacturing facilities with a total installation area as large as one Tokyo Dome. The manufacturing facilities are equipped with more than 2,000 machines - from pumps, compressors, and blowers to reactors and silos - and more than 7,000 instruments. All of which must be inspected visually by specialized operators every day.

According to Yutaro Sotozaki and Shuntaro Sato of the company's Production and Technology Department, "Four operators spend about 3 to 4 hours a day checking whether the instruments are within the normal range and whether there are any leaks or damage to the various equipment."

The burden of this on-site inspection is extremely heavy. Yutaro Sotozaki and Shuntaro Sato, leaders in the Polypropylene Section of the Polypropylene Department at Prime Polymer's Anegasaki Plant were tasked with devising a plan to reduce this burden. The two quickly arrived at an answer: robots. Autonomous robot inspectors could take pictures of the equipment's appearance and instrument displays, and AI would determine if there were any abnormalities.

They first came to Cognite to learn more about its robot control and management capabilities. They were impressed by Cognite’s high operability for controlling robots, its ability to perform automatic inspections with the scheduling function, and its ability to make abnormality judgments using AI, which was the most important requirement. Furthermore, the photos taken by the robot could be analyzed by AI and converted into numerical data, and the data could be accumulated, so it was expected to be used for predictive maintenance.

However, as they proceeded with discussions with Cognite, they learned more about the importance of Industrial DataOps. Even if on-site inspections by robots were realized in this state, a new siloed data group of on-site inspection data would be created. In other words, they faced the problem that in order to utilize the data collected by the robots, they would have to access a new server, making data utilization even more difficult.

In fact, at Prime Polymer, operation data from distributed control systems, electronic documents such as P&ID, technical documents, and repair history were managed separately. Of course, operation data was stored on a dedicated server and could be searched and viewed in chronological order, but since they were all managed on different servers, operators had to access different servers and collect the necessary data on their own computers when conducting operation analysis, troubleshooting, or examination work.

"Siloing had been a problem for a long time, but we had given up on solving it," said Yutaro Sotozaki.

With Cognite Data Fusion®, all data, including on-site inspection data collected by inspection robots, equipment operation data handled by distributed control systems, electronic documents such as piping and instrumentation diagrams (P&ID), digitized past repair history, and technical documents, can be handled on a single platform.

Solving the Silo Problem First

Cognite Data Fusion® uses machine learning and AI-powered data contextualization to automatically recognize data and tag it. This allows various data to be linked together, organized and stored in an easy-to-navigate industrial knowledge graph. Now, when various data that were previously managed on different servers is consolidated into Cognite Data Fusion®, it can all be linked to one asset. As a result, searching for an asset on Cognite Data Fusion® will easily provide all information, such as operation data, technical documents, past repair history, and P&IDs.

The first step was to organize the various data to be input into Cognite Data Fusion®. As mentioned earlier, the manufacturing equipment and instrumentation are vast, and they had to collect technical documents, repair history, and past operation data for each piece of equipment and instrument, and if they could not find them, they created new ones. After all the data was gathered and they started inputting data into Cognite Data Fusion®, the work was extremely fast. "I felt that the work of inputting and contextualizing the data was almost instantaneous," said Shuntaro Sato. In fact, the data input work was completed in just a few months.

The Value of Data Contextualization

The impact of rolling out Cognite Data Fusion® was significant and exceeded expectations, but there are four main areas of improvement worth noting:

First, the "democratization of data." Traditionally, there was data and information that only equipment managers and veteran employees knew. Therefore, when dealing with problems or updating equipment and instruments, there were cases where work was hindered because the necessary information and data could not be found. Now, everyone has access to everything they need to do their jobs effectively.

Second, digitizing and creating interactive P&ID. Engineers can now click on the equipment or instrument that needs to be investigated and Cognite Data Fusion displays a list of technical documents, repair history, and operation data associated with it. This has made it easy for anyone to obtain the necessary data to proceed with their work.

Third, search time for equipment and instruments has been greatly reduced. Although various data such as equipment and instruments had been digitized before, they were not managed in an integrated manner. For example, facility information and operation information were managed on separate servers, so they could not be searched together. In addition, "The server itself was slow, so it took a very long time to search and display the results, but after the introduction of Cognite Data Fusion®, it can now be shortened to a few seconds," said Shuntaro Sato.

Fourth, dashboarding. Cognite Data Fusion® makes it easy to process the input data and display/output useful information. Prime Polymer created multiple dashboards with the support of Cognite, including:

• An equipment management dashboard, which displays the location of equipment and instruments within the large manufacturing facility on a digitized P&ID by entering the four-digit number assigned to them. It can also display technical documents and repair history related to each piece of equipment and instrument. "Traditionally, it was necessary to flip through the paper P&ID drawings to find the target equipment or instrument, and then follow the piping line by hand when you got close to it," said Yutaro Sotozaki.
• A KPI dashboard that allows real-time monitoring of the operating status of heat exchangers by enhancing access to heat efficiency, fouling factor, and electricity/water/steam unit consumption on a single screen.

Towards the Realization of a "Smart Factory"

Initially, the introduction of Cognite Data Fusion® was positioned as a preliminary step to adopting autonomous on-site inspections. However, it unexpectedly contributed greatly to reducing the burden on operators working in manufacturing facilities, and became a major first step towards Prime Polymer's goal of "smart factory."

The company aims to further reduce the burden on operators by making more sophisticated use of Cognite Data Fusion®. In addition to accelerating efforts to realize on-site inspections by robots, they plan to utilize Cognite's Industrial Canvas to enable AI-powered analysis and collaborative work by multiple people.