Predicting of crack spacing for concrete by using neural networks

The current building codes provide calculation techniques to estimate crack spacing for regular building members (beams and slabs). Thick members are commonly used for offshore platforms and containment structures for nuclear power structures. The results have proven that it is necessary to modify the equations used for the crack spacing prediction of thick members. The neural network concept is thus introduced as a tool to estimate crack spacing. Two kinds of neural networks are used: the radial basis and the feed forward back propagation neural networks. In general, both networks show better estimates compared to other available tools. This paper also presents a simplified practical equation for the estimation of crack spacing. The proposed equation is shown to have very good potential in preliminary estimations of crack spacing. Important parameters that control crack spacing are included in the equation, such as rebar diameter, rebar spacing and concrete cover. The results show that other parameters, such as concrete compressive strength and element thickness have minimal effect on crack spacing.

► Building codes use building members to estimate the crack spacing.
► Equations for thick members used for offshore platforms structures are needed.
► Neural networks are introduced as a powerful tool for estimating crack spacing.
► A simplified equation is presented for the estimation of crack spacing.