DCAE-enabled Thermal History Compression for Laser-based Additive Manufacturing Processes

Additive Manufacturing (AM) process monitoring continues to improve for mission critical applications, such as aerospace, biomedical, and defense industries, to minimize process or component defects and meet strict tolerances. A major driving force that advances AM process monitoring is the advanced sensing modalities that collect more data of diverse formats and at faster sampling rates. A drawback to integrating more sensing modalities is the significant rise of the resulting data volumes, making it challenging to store, transmit, and share these overwhelmingly large datasets. Therefore, there is an urgent need for effective management of the big data generated from AM processes. One promising strategy to alleviate storage space and boost AM process data transmission speeds is data compression. This study proposes a new framework that leverages deep convolutional autoencoders (DCAEs) to compress, share, and restore thermal history data, which is one of the most critical sensing modalities in the quality control of laser-based AM processes. A case study based on a Laser Powder Bed Fusion (L-PBF) process is used to evaluate the proposed framework using multiple compression performance metrics, including compression ratio and reconstruction accuracy. The proposed DCAE-enabled framework allows AM users to preserve the most significant data characteristics in their compressed forms, leading to significantly reduced data volumes for more efficient data storage, transmission, and sharing.