Chen Lihai, Bai Xiaolong, He Yonghui, Jia Dong, Li Yican, Li Zhenshui
College of Mechatronics Engineering, Henan University of Science and Technology, Luoyang, 471003, Henan, China.
Postdoctoral Station, AECC Harbin Bearing Co., Ltd, Harbin, 150500, Heilongjiang, China.
Sci Rep. 2025 Aug 24;15(1):31121. doi: 10.1038/s41598-025-17177-w.
The development trend of high precision of mechanical equipment, the reliability of bearings work increasingly demanded. Therefore, it is crucial for the reliable operation of mechanical equipment to evaluate the health status of bearings. It combines IDBO (Improved Dung beetle optimizer) optimised VMD (Variational mode decomposition) and CNN-BiLSTM (convolutional neural network-Bi-directional Long Short-Term Memory) to achieve rolling bearing conformity fault diagnosis and damage assessment. Chaotic mapping, Golden sine algorithm and cosine iteration strategy are introduced to improve the performance of DBO, and the hyperparameters of VMD are optimised using IDBO to improve the signal pre-processing. Feature extraction and fault classification of signals using CNN-BiLSTM is used to compensate for the poor diagnosis of CNN timing signals by BiLSTM instead of Softmax classifier. The HUST dataset is used to discuss the application of signal processing methods and neural network models in bearing composite fault diagnosis. The advantages of the proposed scheme in bearing composite fault and damage assessment are verified, effectively solving the challenge of rolling bearing composite fault diagnosis.
机械设备高精度的发展趋势，对轴承工作的可靠性要求越来越高。因此，评估轴承的健康状态对于机械设备的可靠运行至关重要。它结合了改进的蜣螂优化算法（IDBO）优化的变分模态分解（VMD）和卷积神经网络-双向长短期记忆网络（CNN-BiLSTM），以实现滚动轴承一致性故障诊断和损伤评估。引入混沌映射、黄金正弦算法和余弦迭代策略来提高蜣螂优化算法（DBO）的性能，并使用IDBO优化VMD的超参数以改善信号预处理。利用CNN-BiLSTM进行信号的特征提取和故障分类，通过BiLSTM而非Softmax分类器来弥补CNN对定时信号诊断能力的不足。使用华中科技大学（HUST）数据集来探讨信号处理方法和神经网络模型在轴承复合故障诊断中的应用。验证了所提方案在轴承复合故障和损伤评估中的优势，有效解决了滚动轴承复合故障诊断的难题。
