how to build a pavillion in 9 easy steps...

Here is the lastest plan of attack for the on-site assembly of the pavilion, based in large part by what we learned from our failures with the plywood mock-up, and on the phased production schedule of the Fibre-C profiles which we have agreed with Rieder.

1. Cross Members of Floor deck are layed in place

2. Primary Deck, Bench, and Wall profiles inserted into deck.

3. Lower wall Cross Profiles inserted.

4. Plywood framed jig support erected

5. Primary Global profiles are lifted into place via crane and positioned on jig support.

6. Top Cross Profiles are placed, locking entire roof structure together, and splicing the roof to the walls.

7. Horizontal Bench Profiles are inserted

8. Bench Cross profiles are placed forming the horizontal surfaces of the bench and counter areas.

9. COMPLETION!!!

Sure looks easy...doesn't it? We've still got a lot to figure out, but the wheels are in motion...as always, more to come...

structural bending tests...

as Fibre-C has been developed as a cladding material, not a structural material, we have had to have some tests done. In particular, because the tests that have been done by Rieder in the past, have only tested it's structural capacity perpendicular to the layout of the fibres, while we will be loading the forces in the axis of the fibres.

So today we began the structural bending tests at the labs in Aachen, Germany. According to AKT the tests have been going well...we have no bending failure and no problems with buckling or shear failure. AKT will provide some numbers on capacity of the profiles and further analysis.

In the meantime, here's a video from the testing labs posted for us on youtube...you'll see a big crack begin to appear eventually, but no structural failure!!

plywood mock-up v1.0

The first plywood mock-up test is complete, with a lot of help from Yusuke and some of the Phase One students at the AADRL. Many thanks to Brian Houghton, Abhishek Bij, and Rajat Sodhi for their assistance.

Below you will see some photos documenting the process. The fabrication of the profiles was done on the 3-axis CNC machine at the AA's Hooke Park facilities designed by Frei Otto in the 1970's. The assembly was done at the DRL's John St. studios.

While it wasn't successful in standing, it was definitely successful in teaching. This is what we learned:

• 1. Cutting. The translation of line types from rhino to the CAM cutting programme caused the shape of the profiles and notches to change in places. This must be carefully reviewed and solved before any final cutting.
• 2. The internal corners of the notches must be controlled – if the corners are rounded with a 5mm radius it will reduce the fit of the profiles and any silicone gasket. It a rounded corner is chosen here to reduce stress build up in the working joint then the gasket must be modified to deal with the curvature
• 3. Assembly. The cross profiles on the floor deck should be moved to the outside of the primary profiles. This will give the structure contact with the ground over the full width of the edge and more importantly it will provide a stable jig to fit to insert and align all of the tilted primary profiles. The cross profiles on the ground should be constructed as a mat that is fully bolted of fixed together to help provide full dimensional control of the assembly.
• 4. The assembly sequence should involve erecting a number of primary rings in the centre and either side of the shell and then infilling with the remaining profiles. This will help maintain overall dimensional control and ensure everything fits.
• 5. The tilting of the rear primary profiles in the opposite direction to the front will provide a more efficient structural solution and provide balance to allow easier assembly.
• 6. The joints at the splices on the side walls and floor should be much deeper – approximately 60-80mm
• 7. The numbering of pieces and referencing of specific joint correspondence needs to be further developed. At least one joint at the end of each profile needs to be marked to match a corresponding joint in the other direction.
• 8. Another plywood mock-up of the same area should be made before Christmas after the above changes are incorporated into the design. This should be made of high quality beech ply and assembled in Hooke park.
• 9. Wear gloves next time to avoid splinters!

FEM Analysis...

AKT has run an initial FEM analysis on the roof structure, and well...we've got a bit of work to do.
Though significantly lighter than conventional concrete, the Fibre-C is still concrete, and the self-load of the material is a bit more than our geoemtry can handle at the moment. Basically, with the tilt/lean of the roof structure, the self-load of the structural material, and the non-continouous nature of the structure....our notch joints are taking a hell of a load. The moment forces that these joints are taking is very high. We are also experiencing high stresses and deformation in the profiles due to bending in their weak axis. AKT has told us we need to reduce the stresses by a factor of about 6 (no kidding...my stress factor has got to be +100 these days).

We've got a few options for reducing these stresses, but it looks like it's going to be a combination of them all:

1. Decrease the tilt angle of the roof structure. By reducing the angle by 15 degrees we gain a factor of 2. We also lose some of the dynamic formal movement and drama of the structure.

2. Introduce bracing, especially at the sides of the pavilion. This could be done by infilling flat sheets of fibre-c stiffeners between profiles at various locations. This could also detract from the visual and structural purity of the pavilion.

3. Add more primary ribs at the side faces. This might alter the visual effect of the moire pattern, but could also be interesting if done as a gradient of spacing. I also find it interesting to architecturally express where the structure is performing the most.

4. Increase thickness of material at primary ribs to 13mm, and decrease thickness of material at secondary profiles to 10mm (reducing the self-load).

As I said, we've got some work to do. All of these decisions need to be looked at before we can make a call on them. We need to maintain our design concept, while finding a structural solution that suits. But in particular...we need to so so ASAP!

Here's a look at the revised wire model with spliced profiles and intersection nodes we sent over to AKT today for a quick analysis. In this version, we have reduced the tilt angle by 10 degrees on the front face, and 5 degress on the back.
On another note, once again we have also revised the bench geometry again. Don't worry, it won't be the last time. As you can see below, we are using the 3.6m x 1.2m standard sheet size of Fibre-C as the module for sizing our profiles.
In light of this, we have now lowered the height of the seatback/counter, so that it drops within the 1.2m width of the sheet material. this allows us to cut the entire bench profile in the primary direction, in one continous piece. this is a very minor change, but one which saves us huge amounts of hassle and detailing.

joint details

We've made some progress on the joint detials over the last week. The two models we were looking at use silicone gaskets to provide a cushion between the interlocking profiles and here are the two mock-ups back from the RP lab. Thanks to Jeroen for all his help.

After a design team meeting, we've decided to go with the second option because it can deal with a variation in angles in two planes; it acts as a rotational stiffener once in place, and it will provide a locking resistance against vertical shear.

The assembly process has been reduced to a minimum - a pair of gaskets will be bonded to the notches on the primary concrete profile in advance. On site, the cross profile will be inserted and then the remaining pair of gaskets can be mechanically fixed with one pin each to lock the joint together.

bench in progress...

more revisions to the bench profiles. the original competition entry had the profiles revolving with the geometry of the bench, however this poses problems with how we slot the bench profiles with the deck underneath. now, we are continuing the primary profiles of the roof and floor into the bench. for the cross-members we are looking at taking the sections in one direction for the vertical surfaces of the bench (horizontal sections) and another for the horizontal (seating) surfaces of the bench. we still need to work on the density and spacing of the profiles, but i think we now have the principle in place.

as the deadline approaches and the complexity of joints is exposed, we are basing many decisions on the infamous "KISS Principle".

however, as usual, all sections are completely planar, but never parrallel...

more from theverymany...

here's a few more images from Marc Fornes of theverymany, and his entry into the DRL10 pavilion competition. Marc is doing some fascinating generative geometries in RhinoScript. This entry explores the use of a Danzer Tile, which is an aperiodic tiling system. Check out his site for more. As seen in previous posts, Marc has generously been supplying us with coding support to help us automate the process of producing the documentation for our pavilion.

more joint options

work in progress...

the bench profiles are slowly beginning to take shape....here are some screen shots from today's progress...

moving towards mock-ups...

A few options on joint details. We have a range of 30° on the angle of intersection for each joint. We are proposing that there are then 6 details, each of which can acommodate a variation of 5°. Rieder is confident that their material is strong enough, and their fabrication process acurate enough, that we could do a concrete on concrete joint, with no gaskets or fixings. AKT is not confident about this approach. Not only do they want the gasket, but they also want a mechanical or adhesive fixing. Above is a range of options showing a range of 6 variations each on a gasket-less variable notch detail, and two rubber gasket details.

AKT is also proposing a series of details, which we will post later.

All of the details will be tested through AKT.

In parrallel, we are producing a 1:1 scale plywood mock-up to test fitting and assembly. Here's some images of the 3D model of this mock-up. Yosuke will be coordinating a team of the current DRL phase one students to fabricate and assemble this mock up.


Once again, special thanks to Marc Fornes of theverymany for his assistance with providing the Rhino scripting tools to notch and unroll these surfaces!!

the 58 unrolled profiles for the mock up, notched and ready to be cut...