The recent and remarkable use of Artificial Intelligence (AI) techniques, and particularly, of data mining, allows the improvement of industrial processes through pattern analysis. These tools become very useful when considering condition-based maintenance (CBM) processes, where it is necessary to detect the inflection point in normal operation conditions. In this paper, a novel methodology for CBM is proposed, consisting of 3 data mining techniques: Artificial Neural Networks (ANN), Support Vector Machines (SVM) and Random Forests (RF). Initial analysis of the experiment outcomes suggests that it is recommended to continue the researching efforts in this field because of the improvement obtained in predictive maintenance.
European standard EN13306-2011 defines maintenance as the “combination of technical, administrative and management actions, throughout an asset lifecycle, intended to keep or restore to a required operational state”.
The goal of maintenance process is that the asset performs its required function without losses, which is a rather complicated mission.
Therefore, it is necessary to develop techniques that support assets life extension with optimum performance, mitigating failure rate, or, at least, reducing failure impact. Moreover, it is needed the deployment of techniques that are able to anticipate failure occurrence.