基于坐标注意力机制的旋转机械故障诊断

Rotation Machinery Fault Diagnosis Based on Coordinate Attention Mechanism

旋转机械故障诊断在工业领域具有重要意义。本研究提出了一种基于坐标注意力机制与迁移学习的旋转机械故障诊断方法。为了捕捉旋转机械的故障信号在时频域的特征,运用连续小波变换将原始信号转换为小波时频图。然后,引入基于坐标注意力机制的模型,该机制能够自适应地学习不同位置的特征权重,提升了故障特征的辨别能力。通过在预训练阶段和微调阶段对网络进行训练,实现了模型在不同工况下的迁移学习,提高了模型的泛化能力。实验结果表明,该方法在旋转机械故障诊断中取得了显著的性能提升。相较于传统故障诊断方法,基于坐标注意力机制的模型在故障识别准确率方面取得了明显的提高。同时,通过迁移学习,该模型在不同工况下均表现出较好的性能,证明了其泛化能力和适应性。

Rotating machinery fault diagnosis has significant importance in the industrial domain.This research presents a novel approach for diagnosing faults in rotating machinery using a combination of coordinate attention mechanism and transfer learning.The method involves converting the raw signals from the rotating machinery into time-frequency representations using continuous wavelet transform,enabling the identification of fault patterns in the time-frequency domain.Then,a model based on the coordinate attention mechanism is introduced,which adaptively learns the feature weights at different positions,enhancing the discriminative power of fault features.By training the network in both pre-training and fine-tuning stages,the model achieves transfer learning across different operating conditions,improving its generalization capability.Experimental results demonstrate a significant performance improvement of this method in rotating machinery fault diagnosis.Compared to traditional fault diagnosis methods,the model based on the coordinate attention mechanism achieves notable improvements in fault recognition accuracy.Moreover,through transfer learning,the model exhibits good performance across different operating conditions,demonstrating its generalization ability and adaptability.