Reliable Characterization of Bearing Rings Using Eddy Current and Barkhausen Noise Data Fusion

Abstract A nondestructive testing (NDT) data fusion method for evaluation of bearing rings quality was proposed to reliably and efficiently recognize the qualified bearing rings in terms of hardness. A softmax classifier programed by python was used to classify bearing rings. Bearing rings are nondestructively inspected using Magnetic Barkhausen Noise (MBN) and Eddy Current (EC). Following these nondestructive examinations, the information gathered from these two NDT methods has been fused. The results obtained with these processes were presented in this paper and the classification results were discussed. It was observed that more robust and better prediction performance on bearing ring classification was achieved by using the fused information. Moreover, the fused Root Mean Square (RMS) and reactance features outperform the others in terms of accuracy and loss value indicating least misclassification probability.