In industrial drives, shaft bending is a serious flaw that shortens machinery lifespan and reduces performance. To avoid failures and maximize maintenance, shaft bending changes must be accurately predicted and detected early. The proposed work recommends a unique method for identifying shaft bend faults by incorporating Gradient Boosting Regression (GBR) with Adaptive Wavelet Transform (AWT). The flux signals are break down into frequency components, AWT is able to capture the characteristics related to shaft bent faults to increases the sensitivity of fault detection. A GBR model manage the non-linear relationships of the features produced by AWT. Shaft bent fault is an important defect in industrial drives, which reduce the performance and life time of machinery. Shaft bend faults must be accurately predicted and identified in the initial stage to avoid failures and optimize the motor maintenance. GBR and AWT-based technique extract the features from the flux signal and propose a novel methodology for predicting shaft bent faults. AWT produced the transient features related to shaft bent faults by separating flux signals into their frequency components. Its versatility makes defect detection more sensitive. The features collected from the flux signals by AWT are used by the GBR model to handle non-linear data relationships to predict the shaft bent faults with high accuracy.