Transfer Learning Strategies for Anomaly Detection in IoT Vibration Data

Abstract

An increasing number of industrial assets are equipped with IoT sensor platforms and the industry now expects data-driven maintenance strategies with minimal deployment costs. However, gathering labeled training data for supervised tasks such as anomaly detection is costly and often difficult to implement in operational environments. Therefore, this work aims to design and implement a solution that reduces the required amount of data for training anomaly classification models on time series sensor data and thereby brings down the overall deployment effort of IoT anomaly detection sensors. We set up several in-lab experiments using three peristaltic pumps and investigated approaches for transferring trained anomaly detection models across assets of the same type. Our experiments achieved promising effectiveness and provide initial evidence that transfer learning could be a suitable strategy for using pretrained anomaly classification models across industrial assets of the same type with minimal prior labeling and training effort. This work could serve as a starting point for more general, pretrained sensor data embeddings, applicable to a wide range of assets.

Industrial Transfer Learning for Multivariate Time Series Segmentation: A Case Study on Hydraulic Pump Testing Cycles

“…The research field is still relatively young, with most contributions being published after 2018. However, despite the novelty of the research field, time series-based TL has already been applied in various fields like manufacturing [ 15 ], finance [ 16 ], geoscience [ 17 ], mobility [ 18 ], and medicine [ 19 ]. Successfully solved tasks include time series imaging [ 20 ], anomaly detection [ 21 ], classification [ 22 ], and forecasting [ 23 ].…”

“…Ref. [ 15 ] compared different TL approaches for anomaly detection based on pump vibration data. In turn, ref.…”

Autoencoder Based Anomaly Detection and Explained Fault Localization in Industrial Cooling Systems

“…Wang, Yan, & Oates, 2017;Karim, Majumdar, Darabi, & Chen, 2018) and required a large number of ground truth data, often lacking in industrial applications. Furthermore, these supervised approaches are often difficult to scale and difficult to transfer to other, similar systems (Kemnitz, Bierweiler, Grieb, von Dosky, & Schall, 2021;Heistracher, Jalali, Strobl, et al, 2021). As the affected system is very complex including several hundreds of sensors, the maintenance employee can save valuable time and effort if the affected sensors and thereby the affected subsystem can be localized.…”