Online Adaptive Modeling for Fault Detection of Underwater Thrusters

Samy Marcelo Nascimento, Su-Kyoung Kim, Frank Kirchner

In: Abdel Aitouche (Hrsg.). Proceedings of the 14th International Workshop on Advanced Control and Diagnosis. International Workshop on Advanced Control and Diagnosis (ACD-2017) 14th November 16-17 Bucharest Romania IEEE 2017.


Fault diagnostics and prognostics of thrusters are essential to ensure reliable operation of autonomous underwater vehicles (AUVs). In the case of model-based approaches, it is required to use methods capable of estimating the thruster behavior including non-linearities and adapting online to different operational conditions. We present a data-driven online adaptive modeling pipeline for fault detection of underwater thrusters based on the combination of nominal and adaptive regression models for analytical redundancy and degradation estimation. We evaluate static and adaptive non-linear regression methods for novelty detection applied to the rotational speed and current signals of thrusters.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence