Constructal Design of Isothermal Shaped Cavity Double-T by Means of Simulated Annealing
Gill Velleda Gonzales1; Elizaldo Domingues dos Santos2; Liércio André Isoldi2; Emanuel da Silva Diaz Estrada2; Luiz Alberto Oliveira Rocha3
1 Instituto Federal de Educação, Ciência e Tecnologia Sul-Rio-Grandense (IFSUL); 2 Universidade Federal do Rio Grande (FURG); 3 Universidade Federal do Rio Grande do Sul (UFRGS)
doi:10.20906/CPS/CILAMCE2015-0582
Resumo
In this work we employed the Constructal Design (CD) and the Simulated Annealing (SA) algorithm to the geometric optimization of a isothermal shaped cavity double-T. The problem consist in a solid with internal uniform heat generation and its outer surfaces are in adiabatic conditions. Then, the heat only can be removed by the cavity that is maintained at a minimal temperature. The objective is minimized the maximum excess of temperature (θmax) in the solid domain through geometric optimization of the cavity. The cavity geometry is optimized varying the three degrees of freedom (H2/L2, H1/L1 and S1/H0), but the reasons H2/L2 and H1/L1 are varied with the same values, so maintaining equal the branches of the shaped cavity double-T. According to CD the optimization process can be subjected to four constraints that are fixed (A = 1,c=0.1,1=0.015 and 2=0.015 ). The others degrees of freedom of the cavity geometry are keep in H/L =1 and H0/L0=6. The seek by the optimal geometry is performed by Exhaustive Search (ES) and the results in this process validate the results of SA algorithm. In this study was also investigated the parameter of Cooling Schedule (CS) of the SA algorithm in the results of optimization process. The CS is the main parameter of the algorithm, that determine the temperature decay during the algorithm execution. This parameter influences the robustness of the algorithm because the temperature of algorithm also influence the probability of convergence to local optimal solutions. Then, the results of five different versions of SA was observed, each on with a different parameter of CS. The parameters used are the traditional CS, present in the MATLAB environment, known as Exponential and Boltz, more three hybrids parameters purposed here and called of BoltzExp, ConstExp1 and ConstExp2. The aim to compare the results of SA versions is to improve the algorithm results mainly when is not possible to use the ES method to validation, such as the complete geometry optimization or
Palavras-chave: Constructal Design; Exhaustive Search; Heat Transfer; Simulated Annealing