The Use of Machine Learning for TSA Improvement Implemented on Rotor Balancing Using Low SNR Vibration Signal

Precision rotor balancing is crucial for the optimal performance of rotating machinery. High-precision balancing is achieved by accurately determining the rotor's balancing mass, which relies on the signal-to-noise ratio (SNR) of the vibration measurements. Currently, the time synchronous averaging (TSA) technique is widely used to increase the SNR by reducing asynchronous noise. The effectiveness of TSA depends on the number of averaging-the greater the number of averaging, the higher the SNR vibration. Consequently, the TSA requires more data to achieve optimal results. To minimize the number of averaging needed in TSA, this research proposes a novel technique using a machine learning (ML) algorithm to improve TSA for low-SNR signal. A multistage rotor system is used to validate the proposed technique. This approach successfully achieves high SNR vibrations with good accuracy and reduces the required data for rotor balancing.