Image Deep Learning Assisted Prediction of Mechanical and Corrosion Behavior for Al-Zn-Mg Alloys

The use of metallographic images to predict the mechanical properties materials and their corrosion behavior is helpful in achieving nondestructive detection quality control. However, after a long-term attempt, traditional methods cannot accurately correlate with corresponding microstructure images. In this study, we propose deep learning strategy property large-scale extruded aluminum profiles using surface optical proposed models remarkable were established through experimental dataset collection, preprocessing, network modification, key parameter screening. Taking Al-Zn-Mg alloys different microstructures as example materials, 4,800 sets “metallographic image – hardness (HV) potential (E corr )” data experimentally collected establish HV E prediction accuracies 90% 82%, respectively. exhibit great generalization ability mean average errors 1.8 7.0 mV on validation sets, model can images, property, behavior, which provide theoretical support for intelligent testing further prevent unexpected material failure.