Multilayer feedforward networks are universal approximators
Neural Networks
Approximation capabilities of multilayer feedforward networks
Neural Networks
Finite Elements in Analysis and Design
| Hi-index | 0.00 |
This paper describes a method for statistical analysis effects of material parameters variation on the damage evolution within the workpiece during metal extrusion processes. The proposed approach includes finite element method (FEM) and neural networks analysis. The finite element simulation can predict crack initiation and propagation within the workpiece based on Rice and Tracey fracture criterion. A sensitivity analysis was carried out with respect to the material parameters values in order to identify those parameters to which the risk of failure was distinctly sensitive. Because Monte Carlo simulation is a time consuming repeated analysis, the neural networks are employed in this investigation as numerical devices for substituting the finite element code needed for the workpiece defect prediction. The input data for the artificial neural network are a set of material parameters generated randomly according to a normal distribution to represent the parameters uncertainties. The output data is the maximum damage evolution within the workpiece.