基于注意力BiGRU的机械故障诊断方法研究

为了解决机械故障诊断领域传统方法自适应性差、参数选择过于依赖人工的问题,提出了一种基于循环神经网络的机械故障诊断算法。该方法利用预处理后的机械振动信号,搭建了双向门控循环单元的故障诊断模型,并进行了基于注意力机制的模型优化,提高了特征提取效率。经过美国凯斯西储大学轴承数据集以及自采集的柴油机故障实验数据验证,相比于传统神经网络算法提升了计算效率和诊断准确率,并表现出了良好的抗噪能力。结果表明,该方法可以有效适用于基于机械振动信号的故障诊断,具有一定的工程应用价值。

Mechanical fault diagnosis method based on attention BiGRU

Here,in order to solve problems of poor adaptability of traditional methods in mechanical fault diagnosis field and parameter choosing being too artificial,a mechanical fault diagnosis algorithm based on the recurrent neural network(RNN)was proposed.Using the preprocessed mechanical vibration signal,the fault diagnosis model of the bidirectional gated recurrent unit(BiGRU)was built,and the model optimization based on attention mechanism was performed to improve the efficiency of feature extraction.After verification using US Case Western Reserve University bearing data set and self-collected diesel engine fault test data,compared with the traditional neural network algorithm,it was shown that the proposed algorithm improves calculation efficiency and diagnostic accuracy,and reveal a good anti-noise ability.The example results showed that the proposed method can be effectively applied in fault diagnosis based on mechanical vibration signals,and has a certain engineering application value.