Convolutional Probabilistic Neural Network Approach for Fault Diagnosis in Wind Turbine Gearbox with Symmetrized Dot Patter

Failure of the wind turbine gearbox can result in decreased operational efficiency and downtime. To enhance the reliability of wind turbine operation, this paper proposes the utilization of convolutional probabilistic neural network (CPNN) for recognizing gearbox audio signals. Initially, three common types of gearbox faults are established, including gear tooth breakage, gear wear, and gear rust. Utilizing measured gearbox audio signals, a symmetrized dot pattern (SDP) is generated to form symmetric snowflake patterns, serving as feature images for fault diagnosis. Subsequently, CPNN is employed to extract crucial feature information from the feature images for training and recognition purposes. Experimental results indicate that the proposed method achieves an accuracy of over 98% in identifying wind turbine gearbox fault types.