Feature-Engineering für die Zustandsüberwachung von Wälzlagern mittels maschinellen Lernens

Abstract

In rotating machinery, rolling bearings are often the components limiting service life. To avoid unforeseen downtimes, they have to be maintained. For reasons of safety and cost optimization, condition-based maintenance is increasingly being used. Knowing the condition of the components that are critical to wear is essential for this maintenance approach. The insight about the condition is achieved by means of suitable measurement variables, which can be used to automatically detect the condition of the components using machine learning. The quality of the condition monitoring is strongly dependent on the available measurement data and its preprocessing. For condition monitoring of rolling bearings, structure-borne sound signals can be used. The decisive factor here is to determine so-called features from the high-frequency sampled structure-borne sound signals. These features are supposed to reflect the characteristic properties of the measured signals. At the same time, the amount of data is considerably reduced. In this article, different methods of feature engineering based on structure-borne sound are investigated. For this purpose, the wear of rolling bearings is considered in the context of endurance tests. A new feature generation method is presented and compared to common methods from literature.

On the Importance of Temporal Information for Remaining Useful Life Prediction of Rolling Bearings Using a Random Forest Regressor

“…In a previous paper presented by the present authors, the influence of feature engineering on condition monitoring of rolling bearings was shown using a random forest regressor [20]. A feature engineering approach is presented in the previous work, which, compared to features from Lei et al [21], achieves particularly good results in structureborne sound-based condition detection.…”

“…Calculating rolling means with progressively staggered window widths also adds value in terms of predictive accuracy, although the results are slightly worse than those obtained with the cumulative approach. In the case presented here, the base features are formed from the so-called averaged-frequency-bands, which have already been shown to perform particularly well on the data used in [20]. The authors assume that the methodology presented here will lead to improved RUL predictions for other base features in an analogous manner.…”

“…The focus of the research conducted here is on the comparison of different feature-engineering methods that consider the temporal past in the context of feature generation. As a basis, the so-called averaged-frequency-band (afb) features are used, which have already been shown in [20] to be particularly performant and computationally efficient compared to the features proposed by Lei et al [21]. The studies in [20] were based on the same data set also used for the present work.…”

“…As a basis, the so-called averaged-frequency-band (afb) features are used, which have already been shown in [20] to be particularly performant and computationally efficient compared to the features proposed by Lei et al [21]. The studies in [20] were based on the same data set also used for the present work. Starting from the afb features, additional features are now to be generated, which contain the information of the temporal past.…”

“…The label must represent the progressive bearing damage. As already shown in [20], a label that linearly increases from 0 to 1 is used for this purpose. A similar labeling approach has also been presented in [23].…”