A Novel Baseline-Differencing Approach for Creating Generalizable Reliability Models of Ocean Turbine Behavior  
Author Randall Wald

 

Co-Author(s) Taghi M. Khoshgoftaar; Bassem Alhalabi

 

Abstract Machine condition monitoring/prognostic health monitoring (MCM/PHM) is an important aspect of understanding and predicting the behavior of complex systems such as remote ocean turbines. Frequently MCM/PHM systems will examine vibration data from rotating machinery and define models describing the system’s current state. However, vibrations can be affected by many different forces, and state detection generally only seeks to understand some of these. To help eliminate the effects of irrelevant forces, we propose generating a baseline vibration pattern while the machine is operating under stable environmental conditions, and then differencing out this baseline from any data gathered afterwards. In this way, abnormal states will be identified by how they modify baseline behavior. We present a case study demonstrating this approach using data from a dynamometer test bed, and show that the modified data leads to more general models (models which can be built from one environmental condition and applied to others) than using the data without modification.

 

Keywords Condition monitoring, data mining, data preprocessing, baseline data, environmental conditions
   
    Article #:  1811
 
Proceedings of the 18th ISSAT International Conference on Reliability and Quality in Design
July 26-28, 2012 - Boston, Massachusetts, U.S.A.