Autoencoder-based Representation Learning from Heterogeneous Multivariate Time Series Data of Mechatronic Systems

Research output: Working paper/Preprint › Preprint

Authors

  • Karl-Philipp Kortmann
  • Moritz Fehsenfeld
  • Mark Wielitzka

Research Organisations

View graph of relations

Details

Original languageEnglish
Number of pages7
Publication statusE-pub ahead of print - 8 Apr 2021

Abstract

Sensor and control data of modern mechatronic systems are often available as heterogeneous time series with different sampling rates and value ranges. Suitable classification and regression methods from the field of supervised machine learning already exist for predictive tasks, for example in the context of condition monitoring, but their performance scales strongly with the number of labeled training data. Their provision is often associated with high effort in the form of person-hours or additional sensors. In this paper, we present a method for unsupervised feature extraction using autoencoder networks that specifically addresses the heterogeneous nature of the database and reduces the amount of labeled training data required compared to existing methods. Three public datasets of mechatronic systems from different application domains are used to validate the results.

Keywords

    cs.LG, cs.AI, eess.SP, 68T05 (Primary) 62H12, 68T07 (Secondary), J.2

Cite this

Autoencoder-based Representation Learning from Heterogeneous Multivariate Time Series Data of Mechatronic Systems. / Kortmann, Karl-Philipp; Fehsenfeld, Moritz; Wielitzka, Mark.
2021.

Research output: Working paper/Preprint › Preprint

Download
@techreport{eb7272cbd9da4b02aaa120e2b5694bc0,
title = "Autoencoder-based Representation Learning from Heterogeneous Multivariate Time Series Data of Mechatronic Systems",
abstract = "Sensor and control data of modern mechatronic systems are often available as heterogeneous time series with different sampling rates and value ranges. Suitable classification and regression methods from the field of supervised machine learning already exist for predictive tasks, for example in the context of condition monitoring, but their performance scales strongly with the number of labeled training data. Their provision is often associated with high effort in the form of person-hours or additional sensors. In this paper, we present a method for unsupervised feature extraction using autoencoder networks that specifically addresses the heterogeneous nature of the database and reduces the amount of labeled training data required compared to existing methods. Three public datasets of mechatronic systems from different application domains are used to validate the results. ",
keywords = "cs.LG, cs.AI, eess.SP, 68T05 (Primary) 62H12, 68T07 (Secondary), J.2",
author = "Karl-Philipp Kortmann and Moritz Fehsenfeld and Mark Wielitzka",
note = "A later version of this paper in German language was submitted to VDI Mechatronic Tagung 2021 and will be published in the conference proceedings",
year = "2021",
month = apr,
day = "8",
language = "English",
type = "WorkingPaper",

}

Download

TY - UNPB

T1 - Autoencoder-based Representation Learning from Heterogeneous Multivariate Time Series Data of Mechatronic Systems

AU - Kortmann, Karl-Philipp

AU - Fehsenfeld, Moritz

AU - Wielitzka, Mark

N1 - A later version of this paper in German language was submitted to VDI Mechatronic Tagung 2021 and will be published in the conference proceedings

PY - 2021/4/8

Y1 - 2021/4/8

N2 - Sensor and control data of modern mechatronic systems are often available as heterogeneous time series with different sampling rates and value ranges. Suitable classification and regression methods from the field of supervised machine learning already exist for predictive tasks, for example in the context of condition monitoring, but their performance scales strongly with the number of labeled training data. Their provision is often associated with high effort in the form of person-hours or additional sensors. In this paper, we present a method for unsupervised feature extraction using autoencoder networks that specifically addresses the heterogeneous nature of the database and reduces the amount of labeled training data required compared to existing methods. Three public datasets of mechatronic systems from different application domains are used to validate the results.

AB - Sensor and control data of modern mechatronic systems are often available as heterogeneous time series with different sampling rates and value ranges. Suitable classification and regression methods from the field of supervised machine learning already exist for predictive tasks, for example in the context of condition monitoring, but their performance scales strongly with the number of labeled training data. Their provision is often associated with high effort in the form of person-hours or additional sensors. In this paper, we present a method for unsupervised feature extraction using autoencoder networks that specifically addresses the heterogeneous nature of the database and reduces the amount of labeled training data required compared to existing methods. Three public datasets of mechatronic systems from different application domains are used to validate the results.

KW - cs.LG

KW - cs.AI

KW - eess.SP

KW - 68T05 (Primary) 62H12, 68T07 (Secondary)

KW - J.2

M3 - Preprint

BT - Autoencoder-based Representation Learning from Heterogeneous Multivariate Time Series Data of Mechatronic Systems

ER -