optimization of concrete mix design under the aggressive environmentwith gmdh-type neural networks

one of important causes for failure of concrete structures particular in persian gulf region is diffusion of chloride into concrete. prediction of concrete diffusion factor is an important issue as a key parameter in the being cycle of concrete structures. in addition concrete diffusion factor, increasing in compressive strength and reduction in initial cost is inevitable. the important conflicting objectives that have been considered in this paper are, namely diffusion factor and 28 days- compressive strength. these objective functions have been selected for two objective optimization process. group method of data handing (gmdh) algorithm is self-organizing approach by which gradually complicated models are generated based on the evaluation of their performances on asset of multi-input-single-output data pairs .the gmdh was firstly developed by ivakhenko as a multivariate analysis method for complex system modeling and identification. in this way, gmdh was used to circumvent the difficulty of knowing prior knowledge of mathematical model of the process being considered. in other word, gmdh can be used to model complex system without having specific knowledge of the systems .the main idea of gmdh is to build an analytical function in a feed-forward network based on a quadratic node transfer function whose coefficient are obtained using regression technique. in fact, real gmdh algorithm in which model coefficient are estimated by means of the least squares method has been classified in two complete induction and incomplete induction, which represent the combinational and multilayered iterative algorithms, respectively. nsgaii algorithm is used for multi-objective optimization, this algorithm has some problem in the crowding distance subroutine, therefore a new diversity preserving algorithm, named å_elimination, is proposed to enhanced the performance of multi-objective evolutionary algorithms inoptimization problems.