This paper presents the use of Decision Tree for selecting best statistical features that will discriminate the fault conditions of the hydraulic pump from the signals extracted. In this study J48 algorithm was used as a Decision Tree for detection and classification of those faults that are caused by gear tooth with face wear (GTFW) in MF 285 tractor steering hydraulic pump. The diagnostic features were extracted from frequency spectra of the 20 samples of vibration signals for both good and GTFW conditions. Then the statistical characteristics such as average, standard deviation, kurtosis, median, variance, and skewness of vibration signals were determined. These characteristics were used as input vector to the J48 algorithm. The outcome of J48 algorithm shows that the Rate of accuracy of this method for classification of good and GTFW conditions at speeds 1000, 1500, and 2000 rpm, respectively were 100, 95, and 100 percent. Results show that this method can reliably classify the faults of hydraulic pumps and can be used in future tractors as a part of diagnosis system.
