The purpose of my thesis is to examine various existing methods of data science in detail, and beside studying their benefits, to develop data-based proactive maintenance methods based on them for industrial devices, tools, and machines.
This thesis presents my work based on two datasets. In order to reach my goals, the datasets have to be preprocessed, analysed, and with that information, two proactive maintenance methods ought to be designed, implemented, and verified.
For educational and research purposes I have used two technologies (R, Python) to reach these ambitious objectives, to implement various data processing, and analysing methods, borrowed from statistical or data science field. Armed with that knowledge, I came up with special processes helping to indicate various types of failures in the examined machines and applications.
The difficulty what I had to overcome with in the "CAN fingerprint" dataset is the low rate (2-3) measurement point of the target value, and with the "Electrical failure" dataset is usage of the graph-like data representation. For more detail, please have a peak into my thesis.