The development of an effective time-series based fault identification technique using parametric-distance method
Nurprasetio P.a, Bagiasna K.a, Suharto D.a, Tjahjowidodo T.
a Faculty of Mechanical and Aerospace Engineering, Institut Teknologi Bandung, Indonesia
b Division of Mechatronics and Design, School of Mechanical and Aerospace Engineering, College of Engineering, Nanyang Technological University, Singapore
Abstract
This paper presents the implementation of the combination of time-series modeling and nearest neighbor classification method in detecting common faults in rotating machineries. In this paper we propose the utilization of parametric distance as an instrument to diagnose faults. The parametric distance is defined as the Euclidean distance between the vector of parameters of an unknown fault and the vector of parameters of known faults obtained from the learning stage. Since the vectors are defined in a hyperspace spanned by the parameters of the identified time-series model, the parametric distance is definitely metric. The method has been successfully implemented in the laboratory using a simple vibration test rig. © October 2011 IJENS.
Author keywords
Indexed keywords
Euclidean distance,Fault identification,Rotating machinery,Time series modeling