KIT | KIT-Bibliothek | Impressum | Datenschutz

Machine learning based activity recognition to identify wasteful activities in production

Hofmann, C.; Patschkowski, C.; Häfner, B.; Lanza, G.

Lean Management focusses on the elimination of wasteful activities in production. Whilst numerous methods such as value stream analysis or spaghetti diagrams exist to identify transport, inventory, defects, overproduction or waiting, the waste of human motion is difficult to detect. Activity recognition attempts to categorize human activities using sensor data. Human activity recognition (HAR) is already used in the consumer domain to detect human activities such as walking, climbing stairs or running. This paper presents an approach to transfer the human activity recognition methods to production in order to detect wasteful motion in production processes and to evaluate workplaces. Using sensor data from ordinary smartphones, long-term short-term memory networks (LSTM) are used to classify human activities. Additional to the LSTM-network, the paper contributes a labeled data set for supervised learning. The paper demonstrates how activity recognition can be included in learning factory training starting from the generation of training data to the analysis of the results.

Open Access Logo

Verlagsausgabe §
DOI: 10.5445/IR/1000120485
Veröffentlicht am 24.06.2020
DOI: 10.1016/j.promfg.2020.04.090
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Produktionstechnik (WBK)
Publikationstyp Zeitschriftenaufsatz
Publikationsjahr 2020
Sprache Englisch
Identifikator ISSN: 2351-9789
KITopen-ID: 1000120485
Erschienen in Procedia manufacturing
Band 45
Seiten 171-176
Bemerkung zur Veröffentlichung 10th Conference on Learning Factories, CLF 2020; Graz, Austria, 15 - 17 April 2020
Nachgewiesen in Scopus
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page