5TH International Congress on Technology - Engineering & Science - Kuala Lumpur - Malaysia (2018-02-01)

Gear Fault Analysis Based On Acoustic Signal And Least Square Support Vector Machine

Condition based maintenance is the use of the machinery run time data to determine the machinery condition and hence its current fault condition which can be used to schedule required repair and maintenance prior to breakdown. Gear damage is one of the fault that is often encountered in a transmission system of a machine or mechanical equipment. This problem can lead to a big loss of production. To prevent the occurrence of a catastrophic failure, it requires early fault detection to determine the existence of gear damage. This study aims to analyze gear damage based on the acoustic signal. The gear data was obtained from the test rig with different gear conditions. Five gear conditions, i.e. normal gear, broken 25%, 50%, 75%, and 100% were used in the experiment. The acoustic signals were acquired using PCB 426E01 microphone and DEWESoft acquisition data with a sampling frequency of 20 kHz. The gear data were divided into training and testing data. The training data was used developed the classification model and the testing data was used to evaluate the model performance. Gear fault analysis was done in three steps, feature extraction, feature selection, and gear fault classification. Feature extraction was performed on acoustic signals at time domain, frequency domain, and time-frequency domains. Feature selection was done using Distance Evaluation Technique to obtain salient features that sensitive to the gear damage. The gear fault classification was done using machine learning technique, namely least square support vector machine (LS-SVM). Two multiclass LSSVM classification methods were applied in this study. They are One-Against-One (OAO) and One Against-All (OAA) methods. The LSSVM classification was also performed using different kernel function to obtain best classification result. The results showed that the LSSVM classification using the One-Against-One method and Radial Basis Function (RBF) kernel function give the best performance in analyzing the condition of the gear damage based on the acoustic signal with the accuracy above 95 %.
Didik Djoko Susilo, Achmad Widodo, Toni Prahasto, Muhammad Nizam