College of Control Science and Engineering, Zhejiang University
The National Science Fund for Distinguished Young Scholars (No. 62125306),The National Natural Science Foundation of China (No. 62133003),The Research Project of the State Key Laboratory of Industrial Control Technology, Zhejiang University, China (ICT2021A15)
旋转机械设备是工业生产中的关键性设备,对其进行高效地故障诊断对保障工业安全生产具有重要意义.传统的旋转机械设备智能故障诊断方法采取人工特征提取策略,存在依赖专家经验知识、特征泛化性差、特征完备性不足等局限性,导致故障诊断模型精度差,特别是在噪声环境下性能下降明显. 针对上述问题,本文提出了一种用于旋转机械故障诊断的多模态耦合输入神经网络模型. 首先,利用信号分解方法将原始输入信号分解为多个子信号,并将子信号与原始信号成对组成二维矩阵输入到神经网络中,使得网络能够提取其间重要的相关特征;此外,利用双通道并行的卷积神经网络与长短期记忆网络分别提取信号中的时空间特征并融合,大大提高了网络模型的特征表达完备性,实现了对旋转机械设备的高精度故障分类.本文通过实验验证了该模型相较于传统故障模型具有更高的准确率,并且对于噪声干扰也有较好的适应性.
Rotating machinery equipment is the key equipment in industrial production . It is of great significance to carry out efficient fault diagnosis for industrial safety production. The traditional intelligent fault diagnosis method of rotating machinery adopts the strategy of artificial feature extraction, which has the limitations of relying on expert experience knowledge, poor feature generalization and insufficient feature completeness, leading to poor precision of fault diagnosis model, especially in the noisy environment. To solve the above problems, a multimode coupled input neural network model for rotating machinery fault diagnosis is proposed in this paper. Firstly, the raw input signal is decomposed into several sub-signals by signal decomposition method, and the sub-signals are paired with the raw signal to form a two-dimensional matrix and input into the network, so that the network can extract important related features between the raw signal and sub-signals. In addition, the two-channel parallel convolutional neural network and long and short-term memory network were used to extract and fuse the time-space features of signals, which greatly improved the feature expression completeness of the network model and realized the high-precision fault classification of rotating machinery.The results of the experiments show that the proposed model has higher accuracy and better adaptability to noise disturbance than traditional fault models.