BP neural network based flexural strength prediction of open-porous Cu-Sn-Ti composites

Abstract

Open-porous Cu-Sn-Ti composites are fabricated by the space holder sintering technique using carbamide particles as space-holder material. Generally, the mechanical properties of open-porous sintered composites, especially the flexural strength affect the machine tools wear significantly. In this paper, a back-propagation (BP) artificial neural network with genetic algorithm (GA) and particle swarm optimization algorithm (PSOA) was then employed to relate the composition parameters (pore size, porosity and concentration of molybdenum disulfide particles) to the flexural strength. Furthermore, a comparison of predicted and experimental results using GA-BP and PSOA-BP models was conducted and good prediction accuracy was obtained. The study showed that PSOA-BP models could achieve better prediction results in aspects of the higher convergence velocity, lower relative errors of the flexure strength utilizing GA-BP models. Finally, the high porosity and desired flexural strength were achieved by optimizing the input parameters of open-porous Cu-Sn-Ti composites.