Applying multiple regression analysis to adjust operational limits in condition-based maintenance

Abstract

Condition-based maintenance (CBM) seeks to implement a policy wherein maintenance management decisions are based on the identification of the current condition of monitored machinery. It involves not only collecting data but also comparing them with reference values and, if necessary generating alerts based on preset operational limits. This approach is adopted by a system responsible for monitoring turbomachinery plants in oil platforms, to identify when a machine deserves special attention. With the purpose of extending the functionalities of such system for dynamically adjusting the detection limits and thus improving the precision in setting the appropriate time for maintenance, we proposed an approach based on the identification of clusters of correlated variables and multiple regression analysis. In this paper, we describe our approach and discuss our experience in implementing such functionalities.