Author's: N. Tontchev, S. Popov, P. Koprinkova-Hristova, S. Popova and Y. Lukarski
Pages: [69] - [91]
Received Date: July 29, 2011
Submitted by:
The paper considers two alternative approaches to modelling of dependence between steels alloying components quantities and their obtained after thermal treatment characteristics- nonlinear regression models and neural network models. The obtained two kinds of models are further applied for optimization of the steels compositions aimed at obtaining better mechanical characteristics. The optimization procedures were multiple objective mathematical programming (MOMP) approach using nonlinear regression model and gradient descent optimization procedure using neural network model. The obtained by both approaches results are compared with respect to the quality of models and characteristics of theoretically obtained steel compositions.
steel alloys, nonlinear regression, neural networks, multiple objective mathematical programming approach, gradient descent, modelling, optimization.