研究目的

To propose a system-wide platform involving various machine learning principles and leveraging production data stored in the cloud to identify process parameters that may affect print quality in Laser Powder Bed Fusion (LPBF) additive manufacturing process.

研究成果

This letter explores a cybermanufacturing and AI framework of LPBF processes. In short, this framework allows data sharing from multiple printers and training of ML models. The proposed framework would enable various strategies for process monitoring, diagnosis, and prognosis in metal 3D printing farms. This framework incorporates HDFS and DDS for data sharing and Hadoop ecosystem tools for ML model training based on the shared data. In addition to the LPBF application outlined in this letter, this proposed framework has potential applications for the other multiple-stage processes such as semi-conductor manufacturing and supply chains.

研究不足

The current methods use the local data and adopt a closed-loop feedback control system to adjust a small number of process variables such as laser power in a single AM printer. Due to the lack of deep knowledge of the process parameters and their relationship to print quality, existing closed-loop feedback control systems are not capable of fully controlling the process to prevent defective parts. It is not humanly possible to screen and model all these variables using a statistical method such as the design of experiments or response surface methodology.