An Edge-Cloud Collaboration Framework Based on Federated Feature Separation Autoencoding for Milling Cutter Remaining Useful Life Prediction

Substantial progress has been made in remaining useful life prediction methods for mechanical equipment and milling tools. Nevertheless, the large-scale practical application of these methods still confronts two major challenges. Firstly, high-performance prediction models require high-compute devices to operate, resulting in a significant increase in costs. Secondly, there are issues related to data silos and privacy protection. We propose an edge-cloud collaboration framework based on federated feature separation autoencoding to address these challenges. The training process of the proposed framework is divided into two stages. In the first stage, multiple low-compute clients collaboratively train a feature separation autoencoder through Federated training. This FSAE is designed to extract public features with similar distributions from the data on different clients. In the second stage, the clients upload public features and corresponding labels to a trusted server, where a global prediction model is trained. The trained prediction model can be deployed locally or in the cloud based on the requirement. Finally, the effectiveness and superiority of the proposed framework are verified through the full-life experimental data of milling cutters.