Spatter feature-based in-situ identification of disturbances in process parameters during laser powder bed fusion

Abstract

Laser powder bed fusion (LPBF) is one of the most precise and optimal technologies for achieving superior comprehensive performance in metal additive manufacturing. However, the inherent complexity of the laser and material interaction process poses significant challenges to achieving high reliability and precision in part production. Fluctuations in process parameters are one of the main causes of instability and part quality during the LPBF process. Existing research indicates that spatters are closely related to process parameters and the stability of the molten pool. In this paper, we design a high-speed camera system to monitor the spatters, enabling the detection of disturbances in process parameters. A series of experiments with 10 sets of different process parameters are conducted. To effectively extract representative spatter features from captured images, a DeepSort algorithm-based method is proposed. 7-dimensional features of spatters, including area, perimeter, height, width, speed, angle, and distance from the melt pool to spatters, are extracted to represent spatter characteristics. We further explore the relationships between extracted spatter features and process parameters to verify their rationality in identifying disturbances in process parameters. The results demonstrate that extracted spatter features can effectively represent spatter characteristics and have a high sensitivity in identifying and tracking small spatters. The extracted features are utilized as inputs for a Convolutional Neural Network (CNN), achieving a promising accuracy rate of 96.58% for classifying process parameters in the LPBF process. This work demonstrates the feasibility of utilizing spatter features to monitor process parameter fluctuations, enhancing process control and quality assurance in LPBF.