| Abstract Scope |
In the present work, an image-based deep convolutional neural network (DCNN) transfer learning (TL) workflow was developed to predict the strength and elongation of AlSi10Mg produced by laser powder bed fusion additive manufacturing based on in process data. Samples were fabricated using a wide range of processing parameters, with resultant variation in mechanical properties. During fabrication, photodiode sensor signals were collected, processed into an image dataset, and fed into a DCNN-TL model, that resulted in over 97% and 95% accuracy in predicting ultimate tensile strength and elongation to fracture with model inputs of only photodiode data. Additionally, the model was used to determine both the optimal processing parameters, and number of build layers, required for training/testing of the model |