This is the first pass of my attempt to create a basic representation of the acoustic calculation of an auditorium or a concert hall to get a rough geometry that is based on reflection constraints.
The idea originated from a typical architectural bachelor course task focused on early acoustic reflection design. I used Galapagos with specific distance goals to achieve the correct sound delay time as the main objective to build the hall shape.
The process begins with a simple architectural setup: the desired number of seats and their organization, minimum hall volume based on code requirements per seat, stage size, ceiling limits, and row spacing and aisle parameters. These inputs generate the initial outline of the hall that becomes the base geometry for the calculation.
From this setup the algorithm performs an evolutionary search to generate ceiling reflection planes that achieve the desired early sound delay time. Ceiling and wall reference planes are calculated separately so the system can adapt to different architectural ideas, such as flat side walls or sloped acoustic walls.
The resulting ceiling planes can form either a continuous ceiling surface or a suspended reflector system derived from selected planes at a controlled distance. Small adjustment parameters are included at each stage to prevent invalid geometry and guide the solution toward a usable configuration.
Finally, the closed hall volume is checked against the required minimum acoustic volume. If the space needs to be increased, a single global width parameter expands the hall while preserving the originally calculated reflection planes and avoiding distortion of the acoustic geometry.
While working on this attempt, I already noticed a few points for improvement, so I want to list them here because I am planning to work further on this type of solution that creates truly elegant outputs with minimal manual tweaks and potential break points.
The major improvement point here is the inconsistency of ceiling planes with a target panels width. It’s 1.5 m minimum for a straight section and 2 m for the curved ones, so I decided to snap to the closest planes of a dense plane sequence spread over the ceiling by Galapagos. This is where manual adjustments would create jaggy outlines even within reasonable boundaries. It is manually manageable, but I would want to find another approach to transition from original boundary-shaping reference points to an actual buildable ceiling.
Another goal that I would want to achieve in the next study is smooth, shell-like, unified geometry for walls and ceiling. This is relatively easy to achieve out of context, but in this particular project it is linked to the previous weak point, so I have to solve the problem from its core.