Research on a Spindle Speed-Based Periodic Segmentation Regression Method for Rotor Vibration Data in Aircraft Engines

In the operation of rotating machinery, the analysis and evaluation of vibration signals are crucial for condition monitoring and fault diagnosis. Due to the complex structure and harsh working environment of aero-engines, challenges such as insufficient fault identification samples and vibration data loss are common. Traditional methods for handling missing data often fail to fully restore the hidden information in the original signals, resulting in suboptimal signal reconstruction. In recent years, deep learning techniques, especially Transformer models, have shown good performance in time-series data analysis. However, they still face limitations such as fixed input lengths and high computational complexity. To address these issues, we proposes an RPMTST network based on spindle speed periodic segmentation, aiming to effectively predict and reconstruct missing vibration signals, particularly capturing local variations in the signal. Experimental validation using public datasets demonstrates that the proposed model significantly improves the accuracy of signal reconstruction and fault prediction. This provides more reliable support for condition monitoring and fault detection in aero-engines, contributing to the safe and stable operation of mechanical systems.