基于仿真数据与深度学习驱动的液压多路阀故障诊断

Fault Diagnosis of Hydraulic Multiway Valve Driven by Simulation Data and Deep Learning

摘要: 多路阀是工程机械液压系统中至关重要的控制元件，为保证工程机械的安全稳定运行，需要对多路阀进行准确的故障诊断，因此本文提出了一种基于一维卷积门控循环网络(1D-CNN GRU)的多路阀故障诊断方法。首先通过建立多路阀的AMESim仿真模型得到大量的故障样本数据，解决了深度学习需要大量数据的问题，并且在样本数据中加入随机噪声模拟实际工作环境的数据；接着用1D-CNN提取样本数据的空间特征；然后将提取后的空间特征输入GRU进一步提取时序特征；最后采用Softmax分类器输出故障诊断结果。实验结果表明，所提出方法的准确率高于其他故障诊断算法，故障诊断准确率高于98%。

Abstract: The multi-way valve is a crucial control component in the hydraulic systems of engineering machinery. To ensure the safe and stable operation of the machinery, accurate fault diagnosis of the multi-way valve is necessary. Therefore, this paper proposes a multi-way valve fault diagnosis method based on a one-dimensional convolutional gated recurrent unit network (1D-CGRU). Firstly, a large amount of fault sample data is obtained by establishing an AMESim simulation model of the multi-way valve. This resolves the issue of deep learning requiring a substantial amount of data. Additionally, random noise is added to the sample data to simulate real working conditions. Next, the spatial features of the sample data are extracted using 1D-CNN. Subsequently, the extracted spatial features are input into the GRU to further extract temporal features. Finally, the Softmax classifier is employed to output the fault diagnosis results. Experimental results demonstrate that the proposed method achieves a higher accuracy compared to other fault diagnosis algorithms, with a fault diagnosis accuracy exceeding 98%.