Enhanced sparse representation-based intelligent recognition framework for fault diagnosis of wind turbine drive trains

Condition monitoring and fault diagnosis techniques can effectively enable smart maintenance of wind turbines. Due to complex kinematic mechanisms and strong modulation characteristics, planetary transmissions have remained the most challenging unit for fault diagnosis in wind turbine drive trains. To tackle this challenge, this chapter develops an enhanced sparse representation-based intelligent recognition (ESRIR) framework consisting of structured dictionary design and intelligent fault recognition stages. In detail, structured dictionary designs are implemented using an overlapping segmentation strategy, which incorporates two crucial physical prior knowledge including the self-similarity and shift-invariance property of planetary transmission vibration signals, to greatly promote the representation ability of ESRIR. Then, intelligent fault recognition is implemented with a sparsity-based diagnostic strategy utilizing a minimal sparse reconstruction error-based discrimination criterion. Finally, the applicability and advantage of ESRIR for wind turbine planetary transmission fault diagnosis have been experimentally validated, showing that ESRIR obtains superior diagnostic performances and efficient computational costs.