Title: Pruned-bimodular neural networks for modelling of strength-ductility balance of HSLA steel plates

Authors: Prasun Das; Frank Pettersson; Shubhabrata Dutta

Addresses: Indian Statistical Institute, SQC and OR Unit, 203 B.T. Road, Kolkata 700 108, India ' Faculty of Technology, Heat Engineering Laboratory, Åbo Akademi University, Biskopsgatan 8, FIN-20500 Åbo, Finland ' Bankura Unnayani Institute of Engineering, Pohabaga, Bankura 722146, India

Abstract: In this paper, an attempt has been made in this study to grow the concept of modularity along with pruned networks for strength-ductility balance of high strength low alloy (HSLA) steel plates using lower and upperlayer pruning algorithms. Modelling of strength-ductility balance in case of high strength low alloy steel is a major concern in industrial research. In most cases, the cause of inferior mechanical properties of such steel products could not be clearly identified. The comparative analysis with standard fully-connected network and pruned network reveals an improved performance for pruned-modular architecture and explains the metallurgical phenomenon of HSLA steel in a better way.

Keywords: lower-layer pruning; upper-layer pruning; modular networks; concatenated networks; high strength low alloy; HSLA steel plates; strength; ductility; neural networks; modelling; modularity; pruned networks; metallurgy.

DOI: 10.1504/IJAISC.2014.065802

International Journal of Artificial Intelligence and Soft Computing, 2014 Vol.4 No.4, pp.354 - 372

Received: 07 Dec 2013
Accepted: 05 Jul 2014

Published online: 29 Nov 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article