DS-Evidence-Theory-Based Order Spectrum Sparse Representation Classification for Drivetrain Fault Diagnosis Under Variable Working Conditions

To address the challenges of fault diagnosis in wind turbine drivetrains under variable speed conditions, this paper proposes a novel method called Dempster-Shafer (DS) evidence theory-based order spectrum sparse representation classification (DS-OSSRC). By integrating multi-sensor data, the proposed approach combines order spectrum analysis and sparse representation classification to extract discriminative speed-invariant features for classifier-free intelligent diagnosis. A decision-level fusion strategy based on DS evidence theory is proposed to effectively resolve the conflicts among individual channel outputs, enhancing diagnostic accuracy and robustness. Experimental validation on a wind turbine drivetrain dataset demonstrates that the proposed method achieves 99.52% accuracy under varying working conditions and significantly outperforms single-sensor-based models and two other fusion strategies, especially in noisy environments. The proposed DS-OSSRC method offers a computationally efficient and reliable solution for cross-condition transfer fault diagnosis.