Abstract: Bearing fault diagnosis is an important research topic in aviation engine prediction and health management. Signal processing algorithms and deep learning models in this field rely on datasets. However, publicly available datasets generally cover narrow speed ranges, large speed intervals, single loads, and a lack of composite fault data, making it difficult to support the practical development of fault diagnosis methods. This article discloses a vibration dataset of aircraft main shaft bearings with a wide speed range. In addition to providing single fault data, this dataset also provides multiple composite bearing fault data, covering multi-channel bearing vibration signals with a wide speed range under different loads. The dataset well supports the research of classic fault diagnosis algorithms, and due to the large speed range covered by the data and high-speed sampling rate, it is more conducive to training deep learning fault diagnosis models.