The finite element analysis–based simulation and artificial neural network–based prediction for milling processes of aluminum alloy 7050

Abstract

The cutting forces will generally suffer massive complex factors, such as material deformation, tool eccentricity and system vibration, which will inevitably induce many great difficulties in accurately modeling the cutting force predictions that are very significant to investigate cutting processes. Therefore, the genetic algorithm optimized back-propagation and particle swarm optimization neural networks will be adopted to effectively construct cutting force prediction models. In these two back-propagation prediction models, the main milling parameters will be defined into their input vectors, and the transient milling forces along three different directions will be selected as their output vectors, then the implicit relationships between input and output vectors can be directly generated through practically training and learning these two built back-propagation models with a set of experimental milling force data. Meanwhile, the finite element analysis method will be also used to predict milling forces through programming two easy-to-operate plug-ins that can efficiently construct finite element analysis models, conveniently define processing parameters, and automatically perform mesh generation. Subsequently, the milling forces predicted by the established genetic algorithm optimized back-propagation and particle swarm optimization back-propagation models will be analytically compared with finite element analysis simulations and experiments; also the stress distribution and chip formations of finite element analysis and experiments will be comparatively investigated. Finally, the obtained results clearly indicate that these two back-propagation models built by artificial neural networks can well agree with finite element analysis simulations and experiments, but the particle swarm optimization back-propagation model is superior to the genetic algorithm optimized back-propagation model, which clearly demonstrate the particle swarm optimization back-propagation model has higher efficiencies and accuracies in predicting the average and transient cutting forces for different milling processes on aluminum alloy 7050.