Fault diagnosis technology supports the healthy operation of mechanical equipment. However, the variations conditions during the operation of mechanical equipment lead to significant disparities in data distribution, posing challenges to fault diagnosis. Furthermore, when deploying applications, traditional methods often encounter issues such as latency and data security. Therefore, conducting fault diagnosis and deploying application methods under cross-operating conditions holds significant value. This paper proposes a domain adaptation-based lightweight fault diagnosis framework for edge computing scenarios. Incorporating the local maximum mean discrepancy into knowledge transfer aligns the feature distributions of different domains in a high-dimensional feature space, to discover a common feature space across domains. The acquired fault diagnosis expertise from the cloud-model is transferred to the lightweight edge-model using adaptation knowledge transfer methods. While ensuring real-time diagnostic capabilities, accurate fault diagnosis is achieved across working conditions. We conducted validation experiments on the NVIDIA Jetson Xavier NX kit. In terms of diagnostic performance, the proposed method significantly improved diagnostic accuracy, with average increases of 34.44% and 17.33% compared to the comparison method, respectively. Regarding lightweight effectiveness, proposed method achieved an average inference speed increase of 80.47%. Additionally, compared to the cloud-model, the parameter count of the edge-model decreased by 96.37%, while the Flops decreased by 83.08%.

翻译：故障诊断技术支撑机械设备的健康运行。然而，机械设备运行过程中的工况变化导致数据分布存在显著差异，给故障诊断带来挑战。此外，在部署应用时，传统方法常面临延迟、数据安全等问题。因此，在跨工况条件下进行故障诊断并部署应用方法具有重要价值。本文提出一种面向边缘计算场景的、基于域自适应的轻量化故障诊断框架。通过将局部最大均值差异引入知识迁移，在高维特征空间中对齐不同域的特征分布，以发现跨域的共同特征空间。利用自适应知识迁移方法，将云端模型获取的故障诊断知识迁移至轻量化边缘模型。在保证实时诊断能力的同时，实现了跨工况的准确故障诊断。我们在NVIDIA Jetson Xavier NX开发套件上进行了验证实验。在诊断性能方面，所提方法显著提升了诊断准确率，相较于对比方法平均分别提高了34.44%和17.33%。在轻量化效果方面，所提方法的平均推理速度提升了80.47%。此外，与云端模型相比，边缘模型的参数量减少了96.37%，计算量（Flops）降低了83.08%。