Mr. Tariq Khamlaj 0 Report post Posted November 7, 2017 Greeting everybody, I am new to the CAESES forum. I am currently using DAKOTA with OpenFoam to optimize ducted turbines. I am trying to maximize the power output and minimize drag on the duct. Since it is a multiobjective problem, I have chosen the geneatic alogathim for this purpose. I have read the user guide on their website for version 6.6 and decided to use the setup on page 138 or as below # Dakota Input File: mogatest1.in environment graphics tabular_data tabular_data_file = ’mogatest1.dat’ method moga seed = 10983 max_function_evaluations = 2500 initialization_type unique_random crossover_type shuffle_random num_offspring = 2 num_parents = 2 crossover_rate = 0.8 mutation_type replace_uniform mutation_rate = 0.1 fitness_type domination_count replacement_type below_limit = 6 shrinkage_fraction = 0.9 convergence_type metric_tracker percent_change = 0.05 num_generations = 40 final_solutions = 3 output silent variables continuous_design = 3 initial_point 0 0 0 upper_bounds 4 4 4 lower_bounds -4 -4 -4 descriptors ’x1’ ’x2’ ’x3’ interface analysis_drivers = ’mogatest1’ direct responses objective_functions = 2 no_gradients no_hessians Furthermore, I have added the following line "print_each_pop" under the method section to print out each population. I have noticed that the number of generations in each printed population varies. Some populations have 50 generations some have 224. Shouldn't we get 40 for each population based on number of generations? I have noticed in this example that the maximum function evaluations are 2500, does this mean that we have 62.5 population size? Another important question regarding the inverse design, would be a good optimization practice to use this moga setup for the inverse design? I have attached the results as well as the input files of this example. I would really appreicate any help. Thank you very much for your time.3.zip Share this post Link to post Share on other sites
Jörg 29 Report post Posted November 8, 2017 Hi Tariq, as stated in their documentation you can only control the initial population size: The number of designs in the initial population is specified by the population_size. Note that the population_size only sets the size of the initial population. The population size may vary in the JEGA methods according to the type of operators chosen for a particular optimization run. As far as I know, the guys from Dakota offer a mailing list where you can put your questions. They can probably tell you in more detail what's happening if you need to know. There are also a couple of references given in their documentation Hope this helps,Joerg Share this post Link to post Share on other sites
Mr. Tariq Khamlaj 0 Report post Posted November 8, 2017 Hi Joerg, Thank you very much for the reply. I will contact the authors of DAKOTA. In the meantime, I would like to thank you again for the support. Best regards, Tariq Share this post Link to post Share on other sites
Jörg 29 Report post Posted November 9, 2017 Perfect! Share this post Link to post Share on other sites
