Just like in virtually every other industrial segment, artificial intelligence (AI) techniques are helping paperboard producers like Metsä Board to crunch data and identify how to optimise production processes to achieve the best possible result.
Technical Development Manager Tomi Vähä-Ruohola discusses the company's AI journey so far and what it takes to get from idea to production-ready software.
We began our journey back in 2018 when we started to map the areas where we saw the greatest potential for AI to help improve production processes. Paperboard production is already a highly technical process, and our mills have very advanced automation systems. With AI we can take things one step further as it allows us to optimise based on data gathered from thousands of points in the production process. Further improving the quality of our paperboards means less waste, more efficient raw material use and happy customers.
Speeding up data gathering
With sensors gathering data and AI crunching the numbers we can skip the time-consuming and laborious process of taking board samples, analysing them in the lab and waiting for the results. The virtual sensor we developed in collaboration with the IT partner who was supporting us in this project predicts the quality value based on the lab analysis results we input at the start; the more data it processes the more accurate its predictions become.
Building a bridge to the future
As well as developing the AI analytics model itself, we had to work out how to transfer the huge volumes of data out of the mill environment to our partner. Designing this kind of data architecture was something we had never done before, and we had to ensure that we developed a watertight, secure system. The Metsä Group ICT team were hugely valuable partners in this process. The end result is that we now have a 'data bridge' from our mills that we can use for all kinds of analytical applications in the future.
Plenty of applications in the pipeline
One concrete example of a new application idea can be found in the tail threading process, which involves using ropes and other equipment to rethread the sheet when the board machine is started up after a shutdown or following wet-end sheet breaks.
Tail threading ropes suffer wear over time, and they are located in areas that are not accessible while the board machine is running. Staff at our Äänekoski mill proposed a way to use machine vision and AI analytics to monitor the condition of the ropes and predict the optimal time to replace them, avoiding unnecessary shutdowns. This is the way we like to work – we use our internal knowledge of the paperboard production process to identify applications and our data bridge to get the data out from the mills, and then we trust third-party IT experts to build the algorithms for us.
Now that we have the technical infrastructure in place, we are hard at work developing plenty more new ideas about how we can use AI to further improve the quality and quality consistency of our paperboards for the benefit of our customers.
TAPPI
http://www.tappi.org/