Gear Fault Diagnosis and Classification Using Machine Learning Classifier

Abstract

In industry the condition monitoring of rotating machinery gear is very important. The defect in gear mesh may cause the failure in machinery and that causes a severe loss in industry. The failure in gear mesh reduces the efficiency and hence decreases the productivity in industrial operation. Therefore the health monitoring of gear mesh is very important. Proper health monitoring of gears can avoid the failure in machinery and can save money in industrial applications. The acoustic emission and vibration are the two widely used measuring parameters which is used for the condition monitoring of gear mesh. In this work the gear fault detection by using the acoustic emission monitoring technique is used. This experimentation is done by using an efficient instrumentation system. The experimental set-up is designed which consists of a gear mesh driving system and a hand-held sound analyzer. To carry out the experiment the measuring signals from the defective and healthy gears are captured and compared. In this work the measuring signal is the acoustic emission from the tested gears. Then for the fault detection, two signal processing techniques are followed. These are statistical analysis and adaptive wavelet transform (AWT) analysis. The comparison in statistical as well as in AWT analysis used to detect the fault present in gears. In AWT analysis the adaptive noise cancellation is used to enhance the signal to noise ratio (SNR). Finally faults in gears are classified using the machine learning classifier. The statistical parameter data are used as the input data for the classifiers to train the system to classify the fault.