In the aviation industry, health monitoring of asynchronous motor bearings is critical for ensuring operational safety and reliability. However, the characteristics of fault features in the frequency domain of stator current signals are often unclear, posing challenges for effective fault diagnosis. This study addresses these issues by proposing a bearing fault identification method based on a stacking ensemble learning classification model. A fault experimentation platform was established to collect stator data, which was subsequently divided into samples. Feature extraction and dimensionality reduction techniques were applied to each sample, encompassing statistical, temporal, and frequency domains, including fault characteristic frequencies and entropy values. The reduced dataset was subsequently input into a stacking model that combined six robust classifiers as base models and various meta-classifiers. The model's performance was evaluated in terms of accuracy and classification efficiency across different meta-classifiers. The findings indicate that the stacking model utilizing the K-Nearest Neighbors classifier as a meta-classifier achieved the best training speed of 0.08 s, with an accuracy rate of 97.67%. This comprehensive approach significantly enhances fault detection capabilities in aviation applications, demonstrating its unique contribution to the improvement of bearing health monitoring practices.