11. Kolloquium Mobilhydraulik - Machine Learning for failure mode detection in mobile machinery

  Test bench and feature relevance diagram

A paper on "Machine Learning for failure mode detection in mobile machinery" was published at the 11th Mobile Hydraulics Colloquium in Karlsruhe 2020.

09/10/2020

Contact

Phone

work
+49 241 80 47742

Email

E-Mail
 

Machine learning algorithms (MLA) can be of great assistance to analyze sig-nals of complex systems. By building models based on training data, MLAs learn from examples and detect patterns between inputs (features) and out-puts (labels) to later classify new, unseen data. Prior to working with MLAs, specific steps, like algorithm selection, signal preprocessing and feature ex-traction are required. After model training and validation, further optimiza-tion can increase the models performance and robustness.

The scope of this paper is to give an overview of how MLAs can be used in hydraulic systems. It will serve as a guideline, explaining methods of feature extraction and model building. Using MLA minimizes the need of expert knowledge to detect failure modes from pressure signals resulting in reduced maintenance and down time costs. As an example, a cleaning loader with three bearing shaft failure modes is presented. Pressure signals of the hydrau-lic system are the only available signals that can be used to analyze the cur-rent health status of the machinery and are therefore the focus of this paper.