Optimization of concrete hollow brick using hybrid genetic algorithm combining with artificial neural networks

A structure optimization of concrete hollow brick with four rectangle enclosures is carried out to minimize the equivalent thermal conductivity (ETC) in the constraint of variable shape and position parameters. During the optimization hybrid genetic algorithm (HGA) is developed combining with artificial neural networks (ANN). The modified Latin hypercube sampling (i.e. the maximum minimum distance criterion) is employed to make a robust decision. The ETC of the samples is computed using the finite volume method (FVM) on the basis of 3D multi-mode heat transfer simulation. It indicates that the well-trained ANN can accurately predict the ETC of the concrete hollow brick which matches very well with data obtained from the FVM simulation. The optimization obtains 21.69% improvement on the ETC for the given range of design parameters. The optimized concrete hollow brick owns the largest void volume fraction, the minimum rid and wall thickness, same width of the enclosure, and the optimum staggered arrangement with two same large enclosures and two same small enclosures, which is resulted by the multi-mode heat transfer characteristic of the concrete hollow brick. A novel method of the optimum concrete hollow is proposed to construct new concrete hollow brick with many rows of enclosures. Relative Staggered Ratio (RSR) is used to discuss the effect of the staggered form. By combining two or more rows of the optimized enclosures to one brick with the same size the efficiency to block heat transfer is evidently improved. It is concluded by the present work that the combination of ANN and HGA and the popularizing method are powerful to the optimization of the concrete hollow brick.