Development and validation of ANN for friction riveting of polycarbonate and AA2024-T351 hybrid joints

Abstract

The main objective of this abroad internship project is to evaluate the efficiency of artificial neural networks (ANN) models developed in Brazil/UFSCar during the main project (FAPESP master scholarship in Brazil). These models establish a correlation between friction riveting joining parameters and general properties (ultimate tensile force and aspect of ratio anchoring) of PC/AA2024-T351 joints. The ANN models have been developed during the master's project, from the experimental data obtained in the Full Factorial Design of Experiments (DOE), using the joints produced at the Institute for Research Helmholtz Zentrum Geesthacht (HZG), in Germany from 2010-2011, during the undergraduate years of the applicant. For the validation of the models, random joining conditions in the joinability system PC/AA2024 -T351 interval will be selected. These conditions are different from those previously tested in the design of experiments. The parameters considered for the models are the rotation speed vary between 18000 and 21000 rpm, the joining time between 3 and 4 s and the joining pressure between 0.75 MPa and 1.10 MPa. The joints manufactured will be analyzed by macrostructure, microstructure and mechanical (tensile test) properties. Neural networks have been implemented using software FFNN neural network, developed by Professor Norbert Huber from the Helmholtz-Zentrum Geesthacht Research Centre. (AU)