Parametric analysis is a well known technique for searching for 'optimal' designs. A parametric analysis would usually consist of 1, 2 or 3 design variables being adjusted in a systematic way to find designs with the most favourable characteristics (e.g. low energy consumption, best comfort etc). With parametric analysis, a maximum of 3 variables is normally used because a) the results of more than 3 dimensions to a design problem are difficult to visualize and b) the large number of simulations required with 4 or more design variables would take too long to complete. For example a designer might want to investigate the carbon impact of variable levels and types of glazing. The results would be displayed as a series of parametric design curves. This is a fairly crude way to find optimal designs as only a few variables can practically be included.
Genetic or evolutionary optimisation algorithms can be used to explore ‘parameter space’ for optimal design solutions much more efficiently when more variables are involved. This process allows typically up to 10 variables to be analysed and often with multiple conflicting objectives (e.g. minimize carbon emissions while also minimizing life-cycle costs). For example a base design might be optimised for orientation, wall and roof construction, glazing amount and type, degrees of shading and HVAC system type. The results might be displayed graphically with carbon on one axis and cost on the other and the performance of each design permutation plotted on the graph. The minimum values for cost/carbon form a Pareto front along the bottom of the data points.
DesignBuilder are developing optimisation functionality to allow this powerful design process to be carried out with the minimum of previous technical knowledge in future versions of the software.