How to Make Sense Out of Difficult Data

Most manufacturers take sustainability efforts seriously, but when the company makes paper, it is more than an honorable act for the environment, it’s a business imperative.

As the only paper mill in Denmark, Skjern Paper, which started production in 1967 with the idea of manufacturing paper exclusively from old newspaper, has had sustainability built-in to its production process from the beginning. Today, the company manufactures 75,000 tons of paper and board products each year from 100% recycled fiber.

The ability to provide high quality and just-in-time delivery to customers requires Skjern to use technology. Paper plants have hundreds of Proportional-Integral-Derivative (PID ) control loops that can cause process variations. In addition, Skjern’s quality assurance team did not have many real-time ways to measure paper quality, making real-time quality control difficult.

With the goal of reducing quality rejects, Skjern Paper turned to GE Digital’s Proficy CSense, an industrial advanced analytics software package that can predict future asset and process performance. The software uses artificial intelligence (AI) and machine learning (ML), and what was compelling for Skjern is the software’s ability to do real-time closed loop control actions.

The other thing that made the software so valuable to the organization had nothing to do with the technology, but rather the training that GE Digital provided. Understanding that learning and using analytics isn’t easy, and that most analytic tools are designed for data scientists, GE Digital offers its customers six hours of free consulting to get engineers up to speed on how to use the software. GE calls the program “jump start” as it is a way for customers to quickly understand the tools in order to gain immediate value.

At Skjern, which was looking to reduce quality rejects by 5%, the consulting time with GE Digital experts enabled them to capture insights from the initial project and to then be able to build models to analyze data and discover quality variations. The company was also able to monitor the health of the PID control loops to reduce process variation and to predict product quality in real-time.

“The interesting thing about jump start is that we make sure they don’t get stuck in simple thinks like installation,” said Cobus van Heerden, senior product manager for analytics and machine learning at GE Digital. “We also get to know their teams and what the software was bought for to understand the highest value use case.”

The Skjern team built a model in CSense around a quality parameter related to the paper strength and the fiber bond. By examining that data the paper company was able to discover possible causes of errors in the production line. For example, when producing paper they add dewatering chemicals, and they found that they were producing scrap due to adding too much chemical to the pulp mix. 

Proficy CSense 8.5 Upgrade

GE Digital, on the other hand, found out that it needed to add some more functionality to CSense based on feedback from Skjern. And that has resulted in the roll out of Proficy CSense 8.5 that adds the ability for manufacturers to improve equipment operations by up to 10% through mathematical algorithms that recommend the best tuning parameters or configuration settings to optimize the real-time closed loop control actions. While CSense has always been able to monitor PID control loops, the new tuning assistant can calculate recommended PID loop tuning settings. In addition, there is a system identification assistant to estimate process transfer function models from historical data for process simulation or control optimization.