Rainer Haas, Kurt Pichler,
"Fault Diagnosis in a Hydraulic Circuit Using a Support Vector Machine Trained by a Digital Twin"
: Dynamics and Control of Advanced Structures and Machines, Vol. 156, Springer, Cham, Seite(n) 47-60, 2022, ISBN: 978-3-030-79324-1
Fault Diagnosis in a Hydraulic Circuit Using a Support Vector Machine Trained by a Digital Twin
Sprache des Titels:
Dynamics and Control of Advanced Structures and Machines
This paper presents a novel approach for detecting failures in a hydraulic accumulator loading circuit. By measuring only the accumulator pressure, pump leakage and changes in the accumulator?s pre-fill pressure can be detected. A hydraulic circuit model, which is part of the digital twin, is used to acquire simulated data for the development and training of the condition monitoring method. Especially, it is used to generate data containing different system failures. In a feature engineering step, these data are used to extract meaningful features from the pressure signal. Then an SVM classifier is applied to the feature space to classify the different failure modes. For evaluation, the classifier is applied to different failure cases, and the proposed approach is compared to a commonly used approach that observes the loading time. The results show that the proposed approach is significantly better than the commonly used one especially in the case of multiple failures.