Algorithm 1. The proposed scheme: The combination of the SVAL algorithm and adaptive neural-fuzzy structure with CNN.
1:	Signal Modeling Approximate the bearing function from normal vibration signal using autoregressive technique. (1,2)
2:	Improve the robustness of autoregressive technique by combining autoregressive algorithm with the Laguerre filter. (3,4)
3:	Increase the accuracy and nonlinearity of Equation (3) using support vector autoregressive–Laguerre (SVAL). (13,14)
4:	Signal Estimation ASignal estimation using the combination of the SVAL algorithm and variable structure technique. (15,16)
5:	Reduce the chattering of variable structure technique using the combination of the SVAL algorithm and high-order variable structure technique. (19,20)
6:	Increase the stability and accuracy of high-order variable structure technique using the combination of the SVAL algorithm and adaptive neural-fuzzy structure (proposed method). (27,28)
	Fault Decision
7:	Generate the residual signal. (31)
8:	Resample the residual signals.
9:	Fault pattern recognition and crack size identification using 1D-CNN. (32)