proof

the final design of the airport maintains a sense of wayfinding by providing clear views towards the planes, as well as orienting the ceiling treatment so that clear views of the sky and/or planes are

the final airport design is comprised of the selected pareto designs

once the automated test generated several thousand designs, designs along the pareto frontier - the best performing designs (meaning, the ones that best met the different objectives) - were selected f

animation of high performing (pareto) designs with initial random designs ghosted underneath

animation of designs with a structural objective

animation of initial random designs

final presentation

final design map

hi ian,

midreview was partly good, partly not-so-good.

midreview presentation.  admittedly lacking an experiment's results and important issues/ideas gleaned from the results.  i do believe, however,

draft of mid-review presentation.  please download attached pdf file to view.

please download attached pdf to view

design map c
first revision

design map a
first revision

comparison between predictions and results in the SOBOL + MOSA experiment

percentage of designs on the Pareto frontier using SOBOL (to generate the initial generation's designs) + MOSA (simulated annealing) algorithm.  a larger population size seems to increase successful

basic Pareto designs generated with SOBOL + MOSA (simulated annealing) algorithm

correlation analysis between all three objectives of the SOBOL + MOSA experiment

analysis of SOBOL + MOSA experiment's Pareto frontier for objective 1 +objective 3

analysis of SOBOL + MOSA experiment's Pareto frontier for only objective 3

objective 3 - minimum surface area of the structural cells

run log showing statistics on the total designs generated, average time spent generating each design, and percentage of designs with failed geometries

predictions for experiment using SOBOL + MOSA

predictions for experiment using SOBOL + MOSA

setup of the revised experiment's workflow in mode Frontier

. addition of structural cells in between the linear elements

despite being designs on the Pareto frontier within the first generation, using MOSA seemed to quickly hone in to an optimal input range.  This range experience further refinement as new designs (tha

first three occurrences of most fit designs in respect to only one of the two experiment objectives