Condition monitoring (CM) is widely used in process industry to maximise equipment availability, uniform product characteristics, safety in the work environment, and minimise production losses and waste of materials. In particular, this is the case for the production of kraft liner due to the high number of rotating components of a paper machine. With a state-of- the art CM system, maintenance engineers can identify and study the signatures of most bearing faults in a paper machine several months before the fault severity motivate replacement. However, several years of training are required to perform such analyses and decisions, and human resources are a bottleneck in preventive maintenance development required to avoid unplanned stops and maximise equipment performance. In this collaboration between process industry (Smurfit Kappa Piteå, SCA Munksund), SKF and machine learning researchers (LTU and RISE) we aim to develop and demonstrate concept solutions for automated fault severity estimation and classification in a CM system using machine-learning tools for integration of documented fault analyses and condition indicators. This is to automate frequently performed tasks, so that the trained engineers can focus on more complex CM tasks and preventive maintenance development. Furthermore, we aim to improve the human- machine interface for efficient knowledge management and training. This project is based on a joint pre-study funded by Process IT Innovations. The proposed development of machine learning tools for knowledge integration and digitalisation is motivated also by a general need for scalable CM and decision support innovations, for example in domains like remote CM and equipment performance services.
In addition to the Luleå University of Technology, the project also includes the research institute RISE, SCA Munksund, Smurfit Kappa Piteå and SKF. The project has a turnover of more than SEK 9 million and is financed by the participating parties and Vinnova through the PiiA program.