Decision-making based on decision tree for ball bearing monitoring

Abstract

The vibrations produced by rotating machines affect people and the environment in many ways. They affect comfort, work capacity, health, and safety. For this, condition monitoring is an indispensable tool to track the evolution of vibrations. In the condition monitoring process of rotating machines, maintenance decision-making is subject to constraints relating to the lack of performance of the material resources and the unavailability of experts in the field on the sites, and sometimes the inability of these experts to make a decision. For this reason, a methodology based on two approaches: the vibration analysis and the decision tree to model of the decision-making is proposed. A set of data is collected from vibration signals analysis taken from a set of experiments performed on ball bearings. Then a classification algorithm is applied to build the decision tree. Finally, expert rules are extracted. These rules will be used in the development of a decision-making system, called an expert system. The effectiveness of the proposed methodology is demonstrated in this study.