A Novelty Detection Technique for Machine Condition Monitoring using S.O.M

This paper presents a novelty detection based method for machine condition monitoring (MCM) using Kohonen s self-organising map (S.O.M.). As the fault data set is difficult to acquire in MCM problems, the method requires only the knowledge of normal condition data set. By exploiting S.O.M. s ability of multi-dimensional mapping, the Euclidean distance between the S.O.M. and the data under test is used to discriminate anomaly from normal condition. A set of real world condition monitoring data is used to evaluate the method presented. Experimental result shows high accuracy and reliability of this method.