一种基于联邦学习的云边协同工业安全模型进化方法
By employing techniques such as task similarity-aware federated grouping and differential privacy noise addition, the data privacy and heterogeneity issues in industrial safety systems are addressed, enabling high-precision, adaptive model evolution and improving the model's performance and robustness in complex industrial scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHONGDIAN XINGYUAN TECH CO LTD
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing industrial security systems face contradictions in terms of data privacy, compliance, and data heterogeneity. Traditional federated learning solutions have failed to effectively address data privacy issues in industrial scenarios and lack targeted model evolution strategies, leading to decreased model performance or negative transfer phenomena.
We adopt a cloud-edge collaboration approach based on federated learning. By using a task similarity-aware federated grouping mechanism, we aggregate edge nodes with similar data distributions into federated families. Combined with differential privacy noise addition, secure multi-party computation, and knowledge distillation assistance strategies, we achieve data privacy protection and high-precision model evolution.
It achieves high-precision, adaptive multi-factory collaborative model evolution that meets industrial data compliance requirements while protecting data privacy, improves the model's performance and robustness in complex industrial scenarios, and reduces the cold start time and annotation cost of sparse nodes.
Smart Images
Figure CN122420001A_ABST