Optimization of the location of multiple discrete heat sources in a ventilated cavity using artificial neural networks and micro genetic algorithm

This work is aimed at evaluating the optimal location of three discrete heat sources which could be placed anywhere inside a ventilated cavity and cooled by forced convection. The computational domain involves a square cavity with adiabatic walls, diagonally opposite inlet and outlet, with a heat flux of 1000 W/m2 on the heat sources and constant velocity of 4 m/s at the inlet. The two dimensional flow and temperature fields are obtained by performing simulations on FLUENT 6.3. The micro genetic algorithm (MGA) using the six coordinates of the heat sources as input parameters and 5 individuals in a population is used for the optimization, with the objective function as minimizing the maximum temperature on any of the heat sources. Initially for 66 generations, simulations were repeatedly done to evaluate the objective function. This data was used to train a back-propagation artificial neural network (ANN) using the Bayesian regularization algorithm to predict the fitness from the six inputs. This trained ANN was integrated with the micro genetic algorithm to evolve the population for 1000 generations to arrive at the global optimum. Sensitivity studies have been carried out on the optimal solution by varying the Reynolds number. This study shows that by integrating ANN with GA, the computational time can be reduced substantially in problems of this class.