مشخصات پژوهش

Obtaining reliable fault diagnosis performance for rotating machinery in noisy industrial environments remains a challenging task. Lightweight convolutional neural networks often suffer from limited robustness under severe noise conditions. To address this issue, a lightweight convolutional framework, termed Global Aggregation Convolutional Neural Network (GACNN), is proposed. The model incorporates a global aggregation block to enhance feature representation by integrating global contextual information while maintaining low computational complexity. Experimental results on bearing vibration data demonstrate that GACNN improves diagnostic accuracy and noise robustness compared with a baseline CNN and several existing methods, making it suitable for practical fault diagnosis applications.