Affiliations: Department of Mechanical Engineering of Agricultural
Machinery, University of Tehran, Karaj, Iran
Note: [] Corresponding author: A. Moosavian, MSc Student, Department of
Mechanical Engineering of Agricultural Machinery, University of Tehran, Karaj
31587-77871, Iran. Tel.: +98 912 796 0368; Fax: +98 263 280 8138; E-mail:
a.moosavian@ut.ac.ir
Abstract: Vibration analysis is an accepted method in condition monitoring of
machines, since it can provide useful and reliable information about machine
working condition. This paper surveys a new scheme for fault diagnosis of main
journal-bearings of internal combustion (IC) engine based on power spectral
density (PSD) technique and two classifiers, namely, K-nearest neighbor (KNN)
and artificial neural network (ANN). Vibration signals for three different
conditions of journal-bearing; normal, with oil starvation condition and
extreme wear fault were acquired from an IC engine. PSD was applied to process
the vibration signals. Thirty features were extracted from the PSD values of
signals as a feature source for fault diagnosis. KNN and ANN were trained by
training data set and then used as diagnostic classifiers. Variable K value and
hidden neuron count (N) were used in the range of 1 to 20, with a step size of
1 for KNN and ANN to gain the best classification results. The roles of PSD,
KNN and ANN techniques were studied. From the results, it is shown that the
performance of ANN is better than KNN. The experimental results
dèmonstrate that the proposed diagnostic method can reliably separate
different fault conditions in main journal-bearings of IC engine.
Keywords: Fault diagnosis, main journal-bearing of IC engine, power spectral density, K-nearest neighbor, artificial neural network
