82 J. Koutsoupakis et al. Fig. 12 Health states classification results for the experimental single stage gear drivetrain. Conclusion In this work, an augmented optimal MBD model of a single stage gear drivetrain is developed in order to create highfidelity data for training an ensemble of DL classifiers for CM on the corresponding physical system. This augmented model incorporates an advanced contact normal force module, dedicated to gear meshing, where the effects of the TVMS are taken into account during simulation. This formulation, which is proven to yield improved results compared to the constant stiffness modeling of the contact forces between gears, results in highly accurate numerical data, whose features indicate increased similarity with the physical system’s measured response. The developed CNN ensemble which is trained on numerical data alone is proven capable of making accurate predictions of the system’s health state, while its robustness is also validated through its application on three different operating velocity conditions. The results derived from this work indicate that modeling mechanisms such as contacts, which are dominant in gear-related applications, with greater detail can significantly improve the simulation accuracy, aiding in the development and improvement of simulation-based DL methods. Acknowledgments This project is carried out within the framework of the National Recovery and Resilience Plan Greece 2.0, funded by the European Union – NextGenerationEU (Implementation body: HFRI). Project Code: 16381. References 1. Yuanning Mao, Jun Tong, Zhan Yie Chin, Pietro Borghesani, Zhongxiao Peng, Transmission-error- and vibration-based condition monitoring of gear wear with contaminated lubricant, Wear, Volume 523, 2023, 204760, https://doi.org/10.1016/j.wear.2023.204760.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==