Intelligent fault identification of planet bearings using discriminative dictionary learning based sparse representation classification framework

Planet bearing fault identification is an attractive but challenging task in numerous engineering applications, such as wind turbine and helicopter transmission systems. However, traditional fault characteristic frequency identification and impulsive feature extraction based diagnosis strategies are not sufficient to resolve the problem of planet bearing fault detection, due to complex physical configurations and modulation characteristics in planetary gearboxes. In this paper, a novel discriminative dictionary learning based sparse representation classification (SRC) framework is proposed for intelligent planet bearing fault identification. Within our approach, the optimization objective for discriminative dictionary learning introduces a label consistent constraint called ‘discriminative sparse code error’ and incorporates it with the reconstruction error and classification error to bridge the gap between the classical dictionary learning and classifier training. Therefore, not only the reconstructive and discriminative dictionary for signal sparse representation but also an optimal universal multiclass classifier for classification tasks could be simultaneously learnt in the proposed framework. The optimization formulation could be efficiently solved using the well-known K-SVD dictionary learning algorithm. The effectiveness of the proposed framework has been validated using experimental planet bearing vibration signals. Comparative results demonstrate that our framework outperforms the state-of-the-art K-SVD based SRC method in terms of classification accuracy for intelligent planet bearing fault identification.