Extending Statistical Process Control using quantitative methods in Industry 4.0 approaches
Master thesis
Permanent lenke
http://hdl.handle.net/11250/2619875Utgivelsesdato
2019Metadata
Vis full innførselSamlinger
Sammendrag
Informasjon er gitt om valgfrihet på dette punktet, og derfor er norsk versjon av sammendrag ikke inkludert her eller i selve oppgaven. This master thesis is written in cooperation with SINTEF Manufacturing AS and aims to answer the following research question; Can Statistical Process Control and be extended to function as a tool in Condition Based Maintenance for determining machine condition and Remaining Useful Life estimation, and how does this influence the usefulness of SPC in Industry 4.0? Through using methods including literature review and quantitative methods utilizing the widely used NASA C-MAPSS dataset to construct a model using SPC test data to determine machine condition and estimate RUL, the aim of the thesis is to answer this question in a satisfactory manner.X-Bar and R control charts are constructed and analyzed in a time-series using the complete Western Electric ruleset. A condition indicator score is constructed for every time-series chart by using a scoring algorithm. The distribution of indicator scores across all engines is tested for normal distribution, allowing for calculating probabilities on remaining useful life.The thesis concludes that the constructed model probably can be a useful manual alternative to Machine Learning for RUL estimation, and that SPC as a tool is likely to be increasingly useful in Industry 4.0. The thesis recommends further research into the viability of SPC as a tool for RUL estimation, but also focusing on the benefits of researching human-driven methods for condition monitoring and estimating remaining useful life. Creating real-life run-to-failure datasets should also be a priority.