Informatics Publishing Limited

Sangeeta Yadav ¹, K. K. Pathak ², Rajesh Shrivastava ³

Affiliations
1 Barkatullah University, Bhopal - 462 026, IN
2 Advanced Materials and Processes Research Institute (CSIR) Bhopal - 462 026, IN
3 Government Science & Commerce College Benazir, Bhopal - 462 003, IN

Abstract

In this study, shape optimisation of steel pedestals are attempted using neural network. Considering different geometrical parameters, finite element analyses of pedestals are carried out. Using these results, a back propagation neural network is trained. Successfully trained networks is further used for shape optimisation of newer problems. Thus optimised pedestals are further validated with finite element analyses counterparts and found to be in close match.

Keywords

Stress, Displacement, Finite Element, Pedestals, Neural Network.

Full Text

   

Sangeeta Yadav ¹, K. K. Pathak ², Rajesh Shrivastava ³

Affiliations
1 Barkatullah University, Bhopal - 462 003, IN
2 Advanced Material and Processes Reasearch Institue (CSIR), Bhopal - 462 026, IN
3 Government Science and Arts College, Benazir, Bhopal - 462 003, IN

Abstract

Numerical analysis of Partial Differential Equations (PDE's) of the solid mechanics using Finite Element Method (FEM) is very popular. One issue which is haunting the finite element solution is the computer time. Finite Element Analysis (FEA) using fine mesh and large number of nodes consumes lot of solution time. To overcome this difficulty, a hybrid Neuro-FEM is proposed. FEM solutions considering coarse mesh is used for training neural networks which is further employed for finer predictions. The proposed methodology is successfully employed on a cantilever beam problem.

Keywords

Stress, Displacement, Finite Element, Neuro-Finite Element.

Full Text