Optimization of the irregular shape rolling process with an artificial neural network

This paper presents a design method for a shape roller that combines the numerical analysis and optimization of process parameters with an artificial neural network. Rolling processes are to reduce the cross-section of the incoming material while improving its properties and to obtain the desired section at the exit from the rolls. Shape rolling is one of the most complex deformation processes in comparison with that of the plates, the sheets, and the bars because the workpiece is non-uniformly reduced by grooved rolls. In the case of a product that has an irregular cross-section, the rolled product is severely bent because of the non-uniform deformation of the material in the cross-section. The product analyzed in this study is a steel cutter to cut the desired shape of leather or rubber. A neural network was used to optimize the process parameter. Finite element analyses are done by the pre-defined process parameter which represents the roller geometry. The optimization of the neural network and experimental results of shape rolling are compared. The proposed scheme has successfully optimized the process parameters.