University of Twente Student Theses

Login

Automatic failure analysis for big data-driven industry

Verkuil, B. (2021) Automatic failure analysis for big data-driven industry.

[img] PDF
5MB
Abstract:Many industries have been increasing their interest in data collection and storage over the past decades. With recent technologies allowing huge amounts of data to be processed with relative ease, companies are constantly looking at how their data may help them find and prevent failures. In this work we propose a combination of existing methods for failure analysis which we adjusted to work on continuous big data to automatically learn fault trees from observational data. Our solution scales well and is tested on a real-world dataset of domestic boiler usage. Our approach is based on two previous methods, the C4.5 algorithm for converting continuous to Boolean data, and the LIFT algorithm for learning fault trees from the Boolean data. We use the C4.5 algorithm to automatically split variables on a optimal threshold, replacing the continuous values with a 0 or 1 depending on whether the value is below or above the threshold. This Boolean split allows us to distinguish between faulty and normal operation of the variable and makes it suitable as input for the LIFT algorithm. Besides evaluation of our results we also provide critical feedback on potential problems with previous approaches.
Item Type:Essay (Master)
Clients:
Intergas Verwarming B.V.
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:http://purl.utwente.nl/essays/88759
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page