基于可解释时空图卷积网络的多传感数据融合诊断方法

Multi-sensor Data Fusion Diagnosis Method Based on Interpretable Spatial-temporal Graph Convolutional Network

随着大数据和人工智能技术的发展,深度学习及其衍化算法在故障诊断领域取得了丰硕成果并得到了广泛应用。然而,智能故障诊断方法也存在着诸多挑战:(1)现有的多传感数据融合方法难以充分挖掘多传感器间的时空信息来优化诊断性能;(2)智能诊断模型的决策过程可解释性弱,无法满足实际工业场景下的可靠性要求。因此,提出一种基于可解释时空图卷积网络的多传感数据融合诊断方法。首先,通过构建门控卷积层以学习并增强时间特征;其次,结合传感器空间布局关系和图卷积网络多传感信息融合能力,学习并提取多传感器数据的空间特征,并通过行星齿轮箱故障诊断任务验证了模型有效性;最后,利用改进梯度类激活映射算法分析各传感器数据对模型诊断决策过程的重要度,从而提升模型更新过程的可解释性。试验结果表明,所提方法不仅具有较好的诊断性能,而且为多源传感融合诊断提供了一种有效的事后解释方法。

With the development of big data and artificial intelligence technologies,deep learning and its derived algorithms have achieved fruitful results and have been widely used in the field of fault diagnosis.However,intelligent fault diagnosis methods also face many challenges:1)existing multi-sensor data fusion methods have difficulty fully exploring the spatial-temporal information between multiple sensors to optimize diagnostic performance;2)the decision-making process of intelligent diagnostic models has weak interpretability and cannot meet the reliability requirements in actual industrial scenarios.Therefore,we propose a multi-sensor data fusion diagnostic method based on an interpretable spatial-temporal graph convolutional network(ISTGCN).Firstly,a gate convolution layer is constructed to learn and enhance time features.Secondly,by combining the spatial layout relationship of sensors and the multi-sensor information fusion ability of graph convolutional networks,the spatial features of multi-sensor data are learned and extracted,and the effectiveness of the model is verified through the planetary gearbox fault diagnosis task.Finally,the improved gradient-based activation mapping algorithm is used to analyze the importance of each sensor data to the model diagnostic decision-making process,thereby improving the interpretability of the model update process.The experimental results show that the proposed method not only has good diagnostic performance but also provides an effective post-explanation method for multi-source sensor fusion diagnosis.