Gear fault detection utilizing adaptive multi-scale morphological gradient transform

Vibration signals which carry the dynamic information of the machines are frequently used for mechanical fault diagnosis. Impulsive modulated signals often generated by the defected gear and how to extract the impulsive components from the raw vibration signal with strong background noise has become the most important tasks for gear fault diagnosis. An adaptive multi-scale morphological gradient (AMMG) filter, which can depress the noise at large scale and preserve the impulsive details at small scale, was presented in this work for extracting the impulsive characteristics from the vibration signals generated by defected gear. Both simulated and gear fault vibration signals were employed to evaluate the performance of the proposed technique. Results revealed that the AMMG method has demonstrated a more effective tool for feature extraction of gear compared with the traditional envelope analysis and the morphological close approach.