为保证旋转机械安全稳定运行,在实现轴承早期疲劳损伤阶段故障诊断,提出了改进自适应白噪声平均总体经验模态分解与卷积神经网络融合的故障诊断方法。通过该方法分解原始故障信号,联合核主成分分析降维与分形盒维数筛选最佳重构分量,输入卷积神经网络实现非线性故障特征提取。通过不同信噪比下与现有方法进行对比,表明本方法具有较强抗噪性和较高识别准确率,可有效解决传统轴承故障诊断方法信号非线性特征提取不充分,识别精度不足的局限性。

In order to ensure the safe and stable operation of rotating machinery and realize fault diagnosis in the early fatigue damage stage of bearings, a fault diagnosis method based on improved adaptive white noise average total empirical mode decomposition and convolutional neural network fusion is proposed. The original fault signal is decomposed by this method, and the optimal reconstruction component is selected by combining the dimension reduction of kernel principal component analysis and fractal box dimension, and the nonlinear fault feature extraction is realized by inputting the convolutional neural network. By comparing with the existing methods under different signal-to-noise ratios, it is shown that this method has strong anti-noise performance and high recognition accuracy, which can effectively solve the limitations of insufficient nonlinear feature extraction and insufficient recognition accuracy of traditional bearing fault diagnosis methods.

TH133

国家自然科学基金资助项目(51976131,52006148,52106262)；上海市Ⅳ类高峰学科-能源科学与技术-上海非碳基能源转换与利用研究院建设项目资助Fund-support Project: National Natural Science Foundation of China ( 51976131,52006148,52106262 ) ; shanghai IV peak discipline-Energy Science and Technology-Shanghai Non-carbon-based Energy Conversion and Utilization Research Institute Construction Project Funding