Applying the fuzzy lattice neurocomputing (FLN) classifier model to gear fault diagnosis

Gear faults are among the main causes of rotating machines breakdown in industrial applications. Intelligent condition monitoring for fault diagnosis can be helpful for detecting gear faults in an early stage so as to reduce production loss and, in addition, improve operation safety and reliability. In this work, we present an intelligent gear fault diagnosis scheme based on a novel classification model, namely the fuzzy lattice neurocomputing (FLN) classifier model. Five gear states including one healthy state and four defective states are tested in a two-stage gearbox. Statistical parameters in both the time domain and the frequency domain of vibration signals, acquired from gearbox, are used as features. We conducted experiments on a benchmark dataset as well as on a gear faults dataset to evaluate both the classification performance and the computational cost of the FLN classifier comparatively with alternative classification methods from the literature including artificial neural networks, support vector machines and decision trees. Our study has demonstrated that the FLN model yields better classification performance with smaller computational cost than the aforementioned alternative methods. The FLN classifier can further be used for condition monitoring and fault diagnosis in other mechanical systems.