Finally, the roof works!

The massive 'light' concrete roof finally performs well enough to be built!! After much iteration of curvature, height and sectional sizes, it was wonderful to see all the hoope forces acting in the structure. The maximum deflection achieved now, 316mm, is within the permissible limits for such a large span structure. The stresses are also well under control.

As u can see all the compressive stresses gradually increase towards a maximum towards the 'inner hoope'

The first image above shows the roof with a more elliptical inner ring, which resulted in a deflection of 850mm, after raising the height of the roof from 7m to 12m. A slight change in geometry, as suggested by Andrew, to make the inner ring as close to a circle as possible, and also a slight change in curvature, reduced the deflection to just 316mm. What a drastic change in deflection by only a small modification in geometry!!

Daylight on ground floor and floorplans

As you can see, the different dwellings on the ground floor have different lightsituations. It is important that the dwellings make as much use as possible of the potential daylight in their area. In order to achieve that, our strategy is to cluster the dark functions in a box. This box can be placed in the dark spots. The living room will then always be in the lighter area.

pseudo optimizing

After deciding on doing a more ´geometrical´ optimization, it was more easy to get some geometry with pretty good results in the daylight analysis. The first and second floors are added by trying to achieve as much overlap on the floor beneath as possible. In this image you see the projection of the 3rd floor on the 2nd floor. It is optimized, so there is as little overhang as possible with the given grids.


The geometry (including windows to the southside):

Leaving Ecotect

Another strategy would be to seek for configurations that have as much connection with open space as possible. In other words: minimize the amount of blind walls.
For one island, this goes like this:


Without the streets taken into account (I acknowledge this is an err, but i need to figure out a way to 'discount' the walls facing a street, because the streets are so small that they block out a significant amount of daylight) it looks like this:


As you can see, Galapagos is more or less able to seek in the good direction. It generates better options as you let it run longer. This is the merit of the 'progressive seek algorithm'. Its very satisfying to see that it actually works.


And with the streets added and a quick daylightfactor calculation:



the result is not optimal, in a sense that there are other configurations that have a better ecotect daylight result. But still it is a good option, that follows from common sense.

Galapagos-Ecotect

I have a Celeron Dual Core T3000 1,8 GHz and 3 gig ram. This is way too slow too handle the optimisation of the flat part. The calculation of ecotect is quite heavy and requires more power. Our ´progressive seek algorithm´ does not have much impact on the results, because the population-size is too small. But we cant make it higher because the calculation-time would become too high. Days or even weeks.

Here is an movie of the seekingprocess of only 4 of the 16 'islands'. It is speedened up 9 times (!). so the real proces took half an hour.

As you can see the population-size is low, and the optimum it finds is a clustering of dwellings. This is not the optimal solution, we can tell by common sense. Also earlier calculation with a higher ecotect-grid resolution, show that the optimal configuration for 2 of the blocks is this:

Compared to the optimum of the movie, this looks like its more optimal: more light enters the built space, because there are more surfaces facing open space. This shows that, with a high resolution grid and a bigger population size, the optimization works. But we lack the computational power to do the calculation...which is a shame. The bottleneck is the slow calculation of ecotect. Which is really frustrating.

Daylight analysis

To analyze the amount of daylight in the cavity, we calculated the model in DIALux. These results however were very much too high, when some research didn't solve the problem we decided to do the calculations in Ecotect. The first result looks realistic and promising. More results will follow.

daylightfactor in Ecotect