A Federated Dictionary Learning Method for Process Monitoring With Industrial Applications
Keke Huang, Xinyi Liu, Fanbiao Li, Chunhua Yang, Okyay Kaynak, Tingwen Huang
Abstract
Edge computing is an indispensable technology in the Industry 4.0 era. The data are stored at the edge and transmission is prohibited due to concerns about data privacy. Thus, the traditional centralized data-driven modeling methods will face many difficulties and can hardly be used in practice. Federated learning aims to let local nodes learn a global model cooperatively while keeping their data localized. Motivated by the pioneering framework of federated learning, a federated dictionary learning method is proposed for process monitoring. In detail, local nodes train local data by the K-SVD method and learn local dictionaries. Then, local dictionaries, instead of local data, are transferred to the fusion center for calculation of the global dictionary. In order to guarantee an optimal global dictionary, a novel federated dictionary average strategy is introduced. Finally, the reconstruction errors of local data and the control limit are calculated based on the global dictionary for process monitoring. The performance of the proposed method is verified on a numerical simulation, the continuous stirred tank heater benchmark process, and the industrial aluminum electrolysis process. Process monitoring can help industrial enterprises monitor the operating status of the industrial process in real time based on the collected process data. However, industrial data are stored at the edge and enterprises pay more attention to the privacy of industrial data. The prohibition of external transmission of raw data leads to challenges for traditional centralized process monitoring methods. The federated dictionary learning method proposed in this article can have the good process monitoring performance under the premise of protecting data privacy, which is superior to some state-of-the-art methods. For example, in the aluminum electrolysis industrial process, the false alarm rate is only 1.91%, and the false discovery rate is 86.23%. The method has high data security and low model complexity, which can be quickly and conveniently promoted in other industrial processes, helping enterprises to ensure the safe and stable operation of industrial processes.