Fault diagnostics of spur gear using decision tree and fuzzy classifier

Gears are one of the most widely used elements in rotary machines for transmitting power and torque. The system is subjected to variable speed and torque which lead to faults in gears. This paper presents condition monitoring and fault diagnosis of spur gear conceived as pattern recognition problem. Pattern recognition has the following two main phases: feature extraction and feature classification. Under feature extraction, statistical features like skewness, standard deviation, variance, root-mean-square (RMS) value, kurtosis, range, minimum value, maximum value, sum, median, and crest factor are considered as features of the signal in the fault diagnostics. These features are extracted from vibration signals obtained from the experimental setup through a piezoelectric sensor. The vibration signals from the sensor are captured for normal tooth, wear tooth, broken tooth, and broken tooth under load. The feature extraction is done and the best features are selected using decision tree (J48 algorithm). The selected best features are used to train the fuzzy classifier for the fault diagnosis. A fuzzy classifier is built and tested with representative data.