algorithmic modeling for Rhino
This optimisation component, produced by the Eckersley O'Callaghan's Digital Design Group, is focused on producing optimised solutions with as few function calls as possible, leading to time-efficient optimisation. In addition we have endeavoured to produce useful live feedback during the optimisation and we have allowed for optimisation constraints to be dealt with in a user-friendly way.
The algorithm is based on the Nelder-Mead method, a local search-based optimisation algorithm. Compared to genetic algorithms this method typically has fewer function calls making it efficient for computationally expensive processes such as FEA.
Our implementation allows for multiple constraints to be added and automatically handles these constraints during optimisation (constraint aggregation uses the Kreisselmeier–Steinhauser function). Constraints are added as utilisations where numbers above 1.0 represents an infeasible solution and 1.0 and below is feasible.
Live feedback during the optimisation is provided which gives useful insight into the fitness values, the parameter search space and the feasibility of the solutions.
This is the first release of this component and therefore we would appreciate any feedback or reports of bugs.
We hope the community will find this useful,
EOC's Digital Design Group
Since the first release of the Nelder-Mead Optimisation plug in there have been a few new features:- Automatic restart: Nelder-Mead is a local search-based optimisation algorithm, in order to prevent…Continue
Started by Sam Gregson Aug 26, 2017.