Ball bearings are probably the most used components in mechanics. Since they usually connect mechanical parts with relative speed – like the rotor and stator in an electrical motor - they are at the core of the machine functionality. Damage in these components quickly lead to sudden and unexpected stop of machineries with a loss of production for industries. In a packaging machine, for example, an unexpected stop of a couple of hours may cause costs of loss-production which are several time the cost of the single broken component. The need to avoid unexpected stop becomes mandatory for Industry, which asked Academia ideas, algorithms and procedures to monitor the health of the bearings and predict any incipient fault. In the last decades a huge number of publications covered analysis of vibration data of monitored bearing. A massive number of signal processing techniques have been suggested both on a physical model of the component, or pure blind data analysis, such as the so-called artificial intelligent systems (Artificial Neural Networks, Support Vector Machines, etc…). Most of these intelligent systems require two steps: a training step that “teaches” the system about the correct classification of the incoming data (e.g. into “health bearing” class or “damaged bearing”), and a test step when the inner rules build in the training step are tested on unknown data. There’s a lot interest on intelligent system approaches, since they promise to automatically build the classification rules and they could be applied to different components, not only on the ball bearing. Unfortunately there is a hidden trouble: the intelligent systems work well if the incoming data vectors work well, i.e. they properly describe the signal changes related to an incipient damage. The aim of this paper is to prove that the RMS and Kurtosis values of the vibration data are good parameters that allow a proper classification of the bearing. Moreover the variability of these parameters is close related to the evolution of the damage, suggesting a simple procedure to make the bearings diagnostics
MULTIVARIATE ANALYSIS FOR BEARING CLASSIFICATION / Giuseppe, Curcurù; Cocconcelli, Marco; Rubini, Riccardo. - STAMPA. - (2014), pp. 303-312.
|Data di pubblicazione:||2014|
|Titolo:||MULTIVARIATE ANALYSIS FOR BEARING CLASSIFICATION|
|Autore/i:||Giuseppe, Curcurù; Cocconcelli, Marco; Rubini, Riccardo|
|Titolo del libro:||Settima Giornata di studio Ettore Funaioli|
|A cura di:||http://amsacta.unibo.it/4064/|
|Editore:||Società Editrice ESCULAPIO|
|Citazione:||MULTIVARIATE ANALYSIS FOR BEARING CLASSIFICATION / Giuseppe, Curcurù; Cocconcelli, Marco; Rubini, Riccardo. - STAMPA. - (2014), pp. 303-312.|
File in questo prodotto:
I documenti presenti in Iris Unimore sono rilasciati con licenza Creative Commons Attribuzione - Non commerciale - Non opere derivate 3.0 Italia, salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris