Balanceamento de rotor em um plano de vibração utilizando técnicas de aprendizado de máquina

Abstract

This work investigates rotor balancing in a vibration plane using machine learning techniques. It proposes the development of an algorithm that performs the balancing based on two variables: vibration amplitude and phase angle. Data collection was carried out using the influence coefficient method and subsequently analyzed using descriptive, diagnostic, and predictive analysis methods. Two distinct approaches were implemented: a regression algorithm to determine the correction mass and a classification algorithm to identify the correction position. Various machine learning models, ranging from simpler methods to neural networks, were tested and compared to evaluate the effectiveness of each approach. The results demonstrated the feasibility of the process and the effectiveness of machine learning techniques applied to rotor balancing. After comparing the results obtained from the different models, the development of two neural network-based models was chosen.