Hyperspectral imaging for prediction of surface roughness in laser powder bed fusion

Abstract This article discusses the relevance of in situ quality assurance metal additive manufacturing for cost-efficient product qualification. It presents an approach monitoring laser powder bed fusion (LPBF) process using area-scan hyperspectral camera to predict surface roughness R z with help a convolutional neural network. These investigations were carried out during LPBF processing magnesium alloy WE43 that, due its bioresorbability and compatibility, holds significant potential biomedical implants. A data acquisition methodology has been set up enable efficient management data. The images obtained from labeled as determined by confocal microscope. was used train network whose hyperparameters optimized hyperparameter tuning process. resulting able within mean absolute error (MAE) 4.1 μm over samples three different parameter sets. Since this is significantly smaller than spread actual measured (MAE = 14.3 μm), it indicates that identified features linking roughness. results provide basis future research aiming link further part properties relevant assurance.