<![CDATA[A fan is a mechanical device that produces airflow in a particular area. To achieve sufficient torque and speed, an industrial fan often uses a gear transmission. During its operation, the gears may experience damage. The vibration spectrum is a common method to detect a faulty gear. However, the spectrum often produces a graph that is hard to understand. Moreover, the spectrum sometimes fails to show a clear and high amplitude for small gear faults. The study aims to detect faulty gear based on a classification approach using the Support Vector Machine (SVM) algorithm. It is one of the most robust and accurate algorithms among the other classification algorithms especially for cases with a large number of features. The SVM needs statistical parameters as predictors but the decision to choose the parameters seems arbitrary. This research proposes a simple method to select the parameters using a combination of visual inspection and relief feature algorithm. Twelve statistical parameters are introduced and evaluated for potential input for SVM. The statistical parameters are extracted from the time domain of the vibration signal. The experiment is carried out on an industrial fan test rig and introduces 3 carbon steel spur gear conditions i.e. normal, fault 1, fault 2, and records vibration signal using an accelerometer located near the gear transmission system. Â The SVM classifier is built using the RBF kernel function and the classification is carried out by one vs one and one vs all methods. The result shows that classification accuracy for both methods achieves 100%.]]>
Copyrights © 2020
