80 J. Koutsoupakis et al. The effects of the TVMS model are shown in Figure 8 where comparison between the measured and simulated responses is shown for the Y axis of Accelerometer A1. Fig. 8 Comparison between the experimental and the constant and varying stiffness MBD models of the system in its healthy state at 775 RPM. In this work, two defects were examined to validate the modeling and CM framework. The first case, which for the rest of this text is denoted as Damage 1 (D1), is that of a missing tooth on the inlet gear. The second case, henceforth denoted as Damage 2 (D2) is that of a surface defect on one of the gear teeth. The respective fault cases are shown in Figure 9. Fig. 9 Experimental drivetrain defective gear cases. The results from analysis of the system through the MBD simulations of the two damaged gear cases are shown in Figure 10. As indicated by the figure, the damages result in changes in the magnitudes of the GMF and its harmonics, a change in the features of the signal that the DL classifier is expected to be able to identify with great accuracy. While the results from the optimal MBD simulations in the healthy state of the system indicate a high fidelity between the simulated and measured system responses, validating the proposed numerically trained – experimentally validated CM
RkJQdWJsaXNoZXIy MTMzNzEzMQ==