Federated Learning as a Privacy-Providing Machine Learning for Defect Predictions in Smart Manufacturing

In this work, the federated learning methodology is applied to predict defects in sheet metal forming processes exposed to sources of scatter in the material properties and process parameters. Numerical simulations of the U-channel forming process were performed to analyze springback for three types of sheet steel materials. The datasets of different clients are used to train a single machine learning model. With this approach, multiple parties would simultaneously train a machine learning model on their combined data by training the models locally on the client nodes and progressively improving the learning model through interaction with the central server. This way the industrial peers have no access to the others local data in a centralized server. The predictive performance achieved is similar to a standard centralized learning method, offering competitive results of collaborative machine learning in industrial environment.