Prediction of Mechanical Properties of Aluminum Alloy by Integrating Digital Twins

The prediction of mechanical properties of aluminum alloys is of great significance for practical production. A prediction model that integrates digital twin and PSO-ELM (particle swarm optimization-extreme learning machine) algorithm was proposed. Firstly, a digital twin structure system was constructed, where the data perception layer received twin data from the physical layer and updated the historical database synchronously. The twin data were inputed into the PSO-ELM model to obtain initial prediction results. Then, researchers searched the historical database for production batch with the closest production conditions based on the twin data, and obtained the actual and predicted yield strength of the selected batch, and then revised the preliminary prediction results. The results indicate that the proposed method can effectively improve the accuracy of predicting of aluminum alloys yield strength. Compared with traditional ELM and PSO-ELM models, the digital twin model reduces the root mean square error by 29.6% and the mean absolute error by 22.9%, while the regression coefficient increases to 0.9629, demonstrating significant advantages