Misalignment Fault Simulation in Rotating Machinery Induced by Stiffness Anisotropy of Rotational Components based on Flexible Multibody Dynamics for Machine Learning-oriented Vibration Data Generation

Condition-based maintenance (CBM) techniques based on machine learning (ML) require large and diverse datasets with high reliability. However, in rotating machinery, intentional fault induction under normal operating conditions is practically impossible owing to safety and operational constraints. Accordingly, securing sufficient fault data for CBM development remains a significant challenge. To address this issue, this study developed a computational model based on flexible multibody dynamics (MBD) to simulate misalignment faults in rotating machinery, induced by stiffness anisotropy of the rotational components. Using the constructed model, time-series vibration data corresponding to misalignment conditions were generated, providing physically consistent datasets that could be utilized for training ML algorithms. The proposed approach demonstrates the potential of flexible MBD modeling as an effective tool for fault data generation, offering a viable solution to the sparsity of fault datasets in CBM research and contributing to the advancement of intelligent fault diagnosis technologies for rotating equipment.