Prediction of Shaft Bend Variation in Synchronous Motor Using Modulated Lapped Transform and Support Vector Regression

The combination of Support Vector Regression (SVR) and Modulated Lapped Transform (MLT) methods, offers a novel method for forecasting shaft bend variation in synchronous motors. The objective is to improve shaft bend prediction accuracy, which is important because of less maintenance and performance optimization of synchronous motor. This approach use MLT to extract the features from vibration data and to identify the shaft bend. SVR is a strong regression model to predict the shaft bend variables from the energy band variables obtained from the MLT. The strengths of machine learning (SVR) and wavelet transform are combined to create a predictive model that is more dependable. Actual data from synchronous motors, the proposed work assesses the effectiveness of the suggested approach and shows how well it can predict shaft bend variations. According to the results, synchronous motor health monitoring predictive maintenance strategies should benefit from the integration of MLT and SVR, providing a useful tool for preventive maintenance and reducing the likelihood of unanticipated failures. The MLT & SVR-based SBV prediction accuracy is about 92.9%.