Fault diagnosis of motorized spindle based on lumped parameter model and Wasserstein generative adversarial network

Fault diagnosis of high-speed motorized spindles, crucial for ensuring the reliability and productivity of mechanical systems, faces significant challenges due to sample scarcity and limitations of traditional methods reliant on physical principles. These limitations hinder accurate and reliable diagnoses, particularly given the intricate design and high cost of motorized spindles, which restrict the availability of real-world fault samples. To address these issues, this paper proposes a method based on Lumped Parameter Model (LPM) and Wasserstein Generative Adversarial Network (WGAN) to improve the accuracy of fault classification. Firstly, a 14 Degrees-of-Freedoms (DOFs) for the motorized spindle is created, and its parameters are optimized using Genetic Algorithm (GA). Subsequently, a comprehensive set of simulated fault samples is generated by introducing fault excitations. Finally, WGAN is employed to augment the simulated dataset with additional fault samples, thereby enhancing the accuracy and robustness of AI-driven fault classification models. The results demonstrate that the proposed integrated method effectively mitigates the limitations of traditional methods, yielding accurate diagnosis results.