An Evolving Fuzzy Classifier for Gear Fault Diagnosis

Gearboxes are commonly used in rotating machinery. Fault in a gearbox will degrade transmission accuracy and may also result in sudden failure of machinery systems. Effective fault diagnosis of gear systems has been a subject of intensive research. In this work, a new evolving fuzzy (EF) classifier is proposed for real-time health monitoring of gear systems. The input parameters to the classifier are features extracted from the energy, amplitude and phase domains, respectively. The features of gear health are used to constrain partition of the output space to prevent possible misleading results and improve diagnostic reliability. The effectiveness of the proposed EF classifier is tested experimentally under different gear conditions.