Advanced Condition Monitoring framework for CFRP Gear Drivetrains Using Machine Learning and Multibody Dynamics Simulations 91 or degradation on the pinion gear teeth surface. This condition was simulated by reducing the contact stiffness of the gears by 15%, mirroring the reduction in stiffness typically observed in real-world surface damage scenarios. The third state represents a more severe fault, a missing tooth on the pinion gear, which drastically alters the dynamic behavior of the drivetrain due to the abrupt change in contact mechanics. Figure 5 illustrates the gear models used for each of these health states. By incorporating these different damage conditions into the MBD framework, the study provides a robust dataset for training the ML-based damage detection system, which is designed to classify the health states based on the system’s dynamic responses under varying operational conditions. Fig. 5 Gear geometry for different health states. Figure 6 presents the acceleration responses in the frequency domain for the optimal MBD models at rotational speed of 450RPM. The responses are shown for the first nine harmonics of the Gear Mesh Frequency (GMF) of the gear pair, providing insight into the system’s dynamic behavior at various speeds. Figure 7 compares the simulated responses of the drivetrain system in both its healthy and damaged states. This comparison highlights the differences in dynamic behavior and helps in identifying damage by observing the variations between the healthy and faulty conditions. Fig. 6 Frequency response comparison of the optimal MBD system in its healthy state at 450 RPM.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==