Abstract
The objective of the paper was increasing the efficiency of the material model identification. An attempt to determine the coefficients of the model on the basis of industrial data was made. It is suggested that substituting the FE simulation of the process by a fast metamodel in the inverse analysis should make this analysis much faster. Continuous hot strip rolling process was selected as an object. Artificial neural network was used as a metamodel of this process and good results were obtained in testing the network. Possibility of application of the metamodel of the rolling process for identification of the material model on the basis of measurements of loads during rolling process was the main objective of the paper. The inverse problem was formulated. Due to the lack of the industrial data in a wide range of temperatures and strain rates, in the first approach the experimental data (forces in rolling) were generated by the FE model. The tests of the inverse analysis for the rolling process were performed and good results were obtained. It is shown in the paper that the application of the metamodelling allows for making inverse analysis very fast and, in consequence, any optimization technique can be applied. It is also confirmed that the flow stress model can be identified for the data obtained directly from the industrial rolling process. Possibility of application of the metamodel to the on-line control of the rolling process is demonstrated as well.