پذیرفته شده برای ارائه شفاهی

کد مقاله : 1056-ISAV (R1)

¹دانشکده مهندسی مکانیک و انرژی

²دانشگاه شهید بهشتی

In this research, the efficiency of prognostic feature selection and artificial neural network and various neural network training algorithms in improving the remaining useful life of an-gular contact ball bearing based on acoustic emission signals are investigated. To capture the bearing acoustic emission signals, an appropriate laboratory setup and equipment is used. Acoustic emission signal processing is carried out in the time and frequency domain. 105 different features in the time and frequency domain extracted from raw data have been intro-duced and investigated in detecting the bearing remaining useful life. Prognostic feature se-lection method have been used to reduce the dimension of the extracted features. Also, corre-lation-based feature selection is used for the deletion of similar features. The remaining use-ful life of the bearing is introduced based on the appropriate time to start prediction (TSP) and threshold definitions. Artificial neural networks are used and applications of the different training algorithms are compared for angular contact bearing remaining useful life prediction. The results indicate that acoustic emission is a good method for bearing RUL prediction. Mo-bility, Square-mean-root, and Count are the best time domain features based on the used feature selection method. Also, the Frequency center, Signal power, and F60 are the best frequency domain features based on the used feature selection method. It was shown that be-tween different backpropagation training algorithms, Bayesian Regularization has the lowest SSE error of 3.22 for the prediction of bearing remaining useful life based on frequency do-main features.

Neural network؛ Remaining useful life؛ Angular contact bearing؛ Acoustic emission؛ Training algorithms

Monitoring and diagnostics of rotating machinery