Improving Machinery Safety : Modelling data to explain machine stops and developing a strategy on how to reduce them

Sammanfattning: The purpose of this thesis is to examine how machinery safety at Stora Enso can be increased, with the goal of reducing the amount of machine stops and improving the operational safety within the Packaging Solutions division. To do this, data from the one of the machines at the Jönköping mill has been used for classification and time series modelling. In addition to this, four different mills have been visited, several employees within the company have been interviewed, and a safety survey has been evaluated.  The data consist of information about stops in the machine, as well as information about air quality around the machine. The methods that have been used for classification are Least Absolute Shrinkage and Selection Operator, Random Forest, Extreme Gradient Boosting and Support Vector Machine, and the methods used for time series modelling are Vector Autoregression and Vector Error Correlation Model. The statistical part of the thesis resulted in the realisation that humidity and temperature is important for the board quality, and that thin board grades cause stops more frequently in some parts of the machine than other board grades due to jamming.  Mill visits, interviews and the safety survey gave the results that stress is a common cause for injuries, and that communication is an area of improvement within the division. Together with the results from the statistical modelling, these insights resulted in nine different strategies on how to increase the board grade quality, improve the air quality, reduce stress and improve the communication. For the communication part of the strategy, lean methodology has been used as a baseline.  It would be of great value for the division to improve the quality of machinery data, as well as to perform analysis on several production lines for future studies within the area of machinery safety.