Sheet metal forming in a virtual reality environment using LS-Dyna and neural networks

Manufacturing processes can be simulated and their performance can be optimized by use of finite element techniques. However, finite element techniques cannot typically be used for real time prediction of the effect of variation of geometric and process parameters on the process due to large run times. The aim of this research was to develop and implement a robust methodology to provide real time prediction of process sensitivity to changes in inputs. Neural network models were developed based on finite element simulations of sheet metal forming processes, to achieve real time prediction. Agile and interactive visualization of the processes was accomplished by using the predictions of the neural network to update the model for visualization using virtual reality. The virtual reality environment was developed in JAVA. The research has resulted in the development of 'VRForm' an artificial intelligence based software for real time visualization and optimization of forming processes. VRForm can be used to study the effect of various forming parameters on the process and to optimize blank and tool design to improve process efficiency and robustness. Hydro-forming of channel shaped sheet metal components was used as a case study to illustrate the methodology.