Tampere University of Technology

TUTCRIS Research Portal

Fault detection of elevator systems using automated feature extraction and classification

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Details

Original languageEnglish
Title of host publicationElevator Technology 22, Proceedings of Elevcon 2018, 22nd International Congress on Vertical Transportation Technologies
Subtitle of host publication22-24 May 2018, Berlin, Germany.
Place of PublicationBerlin
PublisherThe International Association of Elevator Engineers
Pages116-122
Number of pages7
Volume22
Edition2018
ISBN (Print)978-965-572-261-1
Publication statusPublished - 21 May 2018
Publication typeA4 Article in a conference publication
EventInternational Congress on Vertical Transportation Technologies -
Duration: 27 Jun 2018 → …

Conference

ConferenceInternational Congress on Vertical Transportation Technologies
Period27/06/18 → …

Abstract

In this research, we study an automated feature extraction technique to calculate new features from raw sensor data provided by an elevator data recording system and to create a more generic machine learning model for fault detection. Another data set called maintenance data is used to find the time period for creating class variables. The calculated features attached to
class variables are classified as healthy or faulty using random forest algorithm. The time period starts from a fault reported by the customer and ends when maintenance is finished and reported. We use accuracy, sensitivity and specificity as evaluation parameters for this research.

Keywords

  • Machine learning, elevator system, deep learning, classification, fault detection, feature extraction

Publication forum classification

Field of science, Statistics Finland