Fault Identification of Multi-level Gear Defects Using Adaptive Noise Control and a Genetic Algorithm

Abstract

This paper proposes a reliable fault identification model of multi-level gearbox defects by applying adaptive noise control and a genetic algorithm-based feature selection for extracting the most related fault components of the gear vibration characteristic. The adaptive noise control analyzes the gearbox vibration signals to remove multiple noise components with their frequency spectrums for selecting fault-informative components of the vibration signal on its output. The genetic algorithm-based feature selection obtains the most distinguishable fault features from the originally extracted feature pool. By applying the denoising during signal processing and feature extraction, the output components which mostly reflect the vibration characteristic of each multi-level gear tooth cut fault types allows for the efficient fault classification. Due to this, the simple k-nearest neighbor algorithm is applied for classifying those gear defect types based on the selected most distinguishable fault features. The experimental result indicates the effectiveness of the proposed approach in this study.