Neural network approach for the selection of processing parameters of aluminium—iron composite preforms during cold upsetting

Abstract

The powder forging process of die forging, sintering, and upsetting is a convenient way of reducing or eliminating the porosity from traditional powder metallurgy products. Forging of metal powder enhances the demanding high-tensile, impact, and fatigue strength of powder metallurgy products. In this research, a demonstration system has been developed that employs a neural network for advising aluminium—iron composite compositions and optimum process settings with desired properties at an early stage in the design of the component. The input comprises the desired mechanical properties, such as formability index, and the system employs these data as inputs in order to recommend suitable metal powder compositions and process settings such as the particle size, percentage of iron content, preform density, aspect ratio, and compact load. The training data were collected by the experimental set-up in the laboratory for the sintered aluminium—iron composite preforms. Comparison of predicted and experimental data has confirmed the accuracy of the neural network approach; therefore, a new way for recommending suitable metal powder compositions and process settings is explored.