noun_Email_707352 noun_917542_cc Map point Play Untitled Retweet Group 3 Fill 1

Artificial intelligence and machine learning enable major savings in the pulp and paper industry

Future AI-based support for operators will drive production consistently in an optimal way, which calls for much deeper integration between automation and business systems.

Antti Blomqvist / March 16, 2020

Mills need to move away from running production business processes almost fully based on human decision-making. Future AI-based support for operators will drive production consistently in an optimal way, which calls for much deeper integration between automation and business systems.

Last year, we wrote in our blog how the pulp & paper industry needs to renew and look for major savings potential. This will happen by automating mainly manual business processes in the offices and enabling operator mobility at shop-floor level. We assumed that Artificial Intelligence (AI) and Machine Learning (ML) are the drivers of the change. Let’s take a quick look where we are today in this inevitable change.

Can AI run a paper machine better than humans?

Many companies are developing data lake-based solutions, where they collect data from automation, mill execution and ERP systems. Data maintenance and modelling, however, need to be solved before the data becomes practically useful. Obstacles include unharmonised master data or missing domain knowledge. How to cope with huge amounts of process data from multiple automation systems is another issue.

Analytics scalability needs looking at. What works on one machine may not work on another. Therefore, models need to be self-learning, using long-term historical data to learn correctly. If the model, for example, does not know the quality of raw materials used, self-learning algorithms are missing essential information leading to wrong decisions.

How to make top-down analytics and bottom-up automation meet?

While analytics services progress as IT initiatives from the top down, automation suppliers are approaching the issue from the bottom up. They typically focus on a specific machine or machines and usually do not cover the complete production line.

Quality controllers today are not designed to optimise both quality and cost, as they do not have the information to do it. This is where major cost savings potential is hiding: optimisers on different levels feed their information to automation to enable holistically optimised production.

As AI and ML go forward and start to provide reliable results, the next step is to hook algorithms into active process control. This requires standard API interfaces and services across multivendor environments.

Towards continuous planning

Many companies today focus on their future digital core – “ERP version 2” – struggling with two topics:

  1. Which business processes should be standard and which ones are industry or company specific?
  2. Which services are run in the cloud, preferably as a service, and which ones are needed on-premise to provide business continuity?

There is no single correct answer. Every company needs to figure out what the best solution is for their strategy and operational efficiency.

As business processes will be more and more automated, we are moving towards “continuous planning” instead of the present “plan and execute” model. Continuous and automated planning enable production in future to accept new orders in the middle of a production run or automatically re-optimise the run with the latest orders.

This is important for architecture since continuous planning requires complex interfaces in an area where no integration standards exist. The optimisation potential can be realised by combining information from several systems and integrating this into active production control.

Local MES (Edge) enables automation to optimise production with full visibility into incoming materials and optimised settings to fulfil business targets. It then transfers the production outcome to the next production step with re-optimised targets.

Human involvement only where needed

The long-term target, besides continuous process optimisation, now seems to be to enable operator mobility. In future, the systems take care of standard events without operator involvement.

Incoming raw materials are automatically identified, produced units are automatically labelled, optimal production sizes are automatically calculated and sent to automation, produced waste is automatically detected and connected to produced units.

Do we in future need full-time winder operators or can operators do something else while the machine does the routine work? Raw materials can be automatically transferred to the correct machines at the correct time. When something goes wrong or operator decisions are needed, the system will ask the operator for support through mobile.

It’s a long race, but the prize is worth it

The road to fully automated, self-learning production chains is still long and somewhat winding. There are no shortcuts. TietoEVRY’s industry expertise, integration competencies and process management knowledge help you create a platform on top of which specific solutions can be built to make use of data.

In 2020, we continue working towards major costs savings, providing value and even better customer service for the complete pulp & paper industry value chain.

Read more of our solutions 

Antti Blomqvist
Head of Offering Development, Pulp, Paper and Fibre

Antti has strong professional experience in the paper industry, from automation systems to order-to-cash processes. He has had various roles in product management, business consultancy and business process management, and background in the chemical, steel and manufacturing industries. 


Antti Blomqvist

Head of Offering Development, Pulp, Paper and Fibre

Share on Facebook Tweet Share on LinkedIn