In a paper mill, data from thousands of sensors installed in paper machines are analyzed daily. The analyzes are carried out by maintenance engineers or analysis technicians and the purpose is to detect operating malfunctions and determine when worn rotating components need to be replaced. Simply put, it is a matter of anticipating when machine parts need to be replaced, thereby avoiding expensive unplanned downtime.
Analyzing data and assessing when components need to be replaced requires extensive experience and is a relatively time-consuming routine for maintenance engineers.
– If we can automate this routine work, the maintenance engineers can focus on more complex tasks and preventive maintenance, says Fredrik Sandin, Associate Professor of Industrial Electronics and coordinator of the project.
– The purpose of the project is therefore to develop and demonstrate conceptual AI solutions. We do this by combining machine learning for language analysis and condition monitoring. You could say that we digitize knowledge about machine damage and maintenance needs.HIT
More stable production
In order to develop AI solutions, measurement data and documentation from the paper mill's monitoring systems will be analyzed and supplemented with in-depth investigations of machine damage. The researchers will use machine learning methods and data and document analysis tools to automatically identify damage and anticipate maintenance needs on the paper machines. Hopefully, the project's results will contribute to reducing the paper mill's unplanned downtime by a quarter, which corresponds to a production loss of approximately SEK 4 million per paper machine and year. The number of large rolling bearings that can be re-manufactured will also increase, which contributes to more efficient production and reduced environmental impact.
– We look forward to working with Luleå University of Technology and the other companies in the project, says Jonas Snäll, maintenance engineer at Smurfit Kappa in Piteå.
– With future solutions on how and when production machines are to be maintained, we can have a smoother and more stable production. If the decision-making process for when maintenance is to be performed, becomes less dependent on people, we can work more efficiently.
AI as enabler
Fredrik Sandin also points out that there are more benefits to the project.
– We hope that the development of AI tools will strengthen companies' development and innovation capacity. AI tools can be used to improve both products and production, says Fredrik Sandin.
Marcus Liwicki, Professor of Machine Learning at Luleå University of Technology agrees:
– Introducing machine learning in paper production is a first step towards more sustainable resource management in our region. In addition, similar applications of AI are possible in almost all industrial production processes, says Marcus Liwicki.