Inverse Design of Digital Materials Using Corrected Generative Deep Neural Network and Generative Deep Convolutional Neural Network

Generative networks are effective tools for digital materials (DM) inverse design. However, the optimization performance of generative is restricted by increasing discrepancy between optimized input and prescribed domain as design loop increases. Herein, a correction technique incorporated into deep neural network (GDNN) convolutional (GDCNN). The performed pulling machine learning (ML)-optimized inputs back to at certain interval during process instead only postprocessing end. A DM system with two phases, i.e., matrix phase pore phase, used structural datasets produced using numerical model describe relationship material structure elastic modulus tensile strength. results show that effectiveness corrected GDNN/GDCNN significantly improves given fact more structures converge best fewer nonrepetitive left after optimization, which helps search decreases computational burden when verifying ML-recommended structures. GDNN GDCNN also manage find higher strength in new