(177a) Defect Detection Using Experimental and Simulated Dataset and Data Reduction for Direct Metal Laser Solidification (DMLS)

Direct metal laser solidification (DMLS) has been receiving increasing research interest in the additive manufacturing (AM) industry due to its outstanding performance in producing parts with ultra-high precision and variable geometries [1]. However, the lack of appropriate in-situ disturbance detection techniques specialized for DMLS makes real-time quality control extremely difficult. To help process engineers analyze sensor images and efficiently filter monitoring data for transport and storage, machine learning and data processing algorithms are often implemented [2]. These algorithms integrate the functionality of automated image processing, transferring and analytics. In particular, advanced image analytics using convolutional neural network (CNN) for printing error detection has been a really important approach [3]. Nevertheless, the industrial utilization of these deep learning methods usually encounters problems of the limited training dataset.

In this work, we are endeavoring to extend the CNN training to realistic thermal features from long wave infrared (LWIR) images. In particular, the training dataset combines a limited amount of experimental data with and without a variety of simulated data augmentation to examine its contribution to the CNN training efficiency and accuracy. The thermal feature extraction and the developed machine learning algorithm are then utilized for the data reduction purpose. Also, the transmission strategy is demonstrated to filter a significant amount of data while maintaining model prediction supremacy.

[1] Liu, R., Wang, Z., Sparks, T., Liou, F., Newkirk, J., 2017. Aerospace applications of laser additive manufacturing, in: Laser Additive Manufacturing. Elsevier, pp. 351–371.

[2] Zhang, Y., Hong, G.S., Ye, D., Zhu, K., Fuh, J.Y., 2018. Extraction and evaluation of melt pool, plume and spatter information for powder-bed fusion AM process monitoring. Materials & Design156, 458–469.

[3] Scime, L., Beuth, J., 2019. Using machine learning to identify in-situ melt pool signatures indicative of flaw formation in a laser powder bed fusion additive manufacturing process. Additive Manufacturing 25, 151–165