Above: exterior render image of the digital fabrication block
The project departed from the concentric city plan of Mandalay, acting as an urban intervention that reflects the grid layout. The workshop and residential blocks are linked via an elevated walkway on the second level.
The wood and digital fabrication blocks are differentiated through the language of the columns – rectilinear for the wood production block and curved for the digital fabrication block – in order to reflect the processes taking place in the workshop. The roof uses Kielsteg as a long-span solution.
Simulations were done using Autodesk Simulation Mechanical 360 in order to check the feasibility of the shape of the columns as well as to find the optimum column grid.
Prototypes of the columns were also made in order to better understand the nature of timber as a construction material.
Inspired by the same structural principles as the Dougong Cloud project, Cloud V is the vertical variation of the DouGong (bracket and block) timber joinery system. Designed as a hybrid timber (beams, columns) and concrete (floor slabs, cores) high-rise service apartment block in Mandalay (Myanmar), considerations were given mainly to structural tectonics and integrating with on-site climatic conditions.
Like a DouGong column head, the main structural strategy is to support a large volume of space using a small footprint. Forces would be distributed downwards from a larger congregation of members to a smaller assembly of members connected perpendicularly to one another in layers. Through the massing phase, the 3 blocks are angled to reduce accumulated solar radiation from the South and East sun, while maximising daylight entry for each room. This creates a large scale cantilevered structure, similar to how a DouGong is assembled. To manifest the fore-mentioned idea, the overall structure was inspired by the construction of a Japanese Pagoda, which employs 2 main structural principles:
1) The central structural member is non load-bearing.
Vertical and Horizontal forces are transferred strictly from beam to column without depending on the central column, or in this case, the core. Thus, the central core is freed up from its load bearing functions. This is expressed by the skylight void around the core to emphasize the departure of the vertical core as a recognised strong load-bearing element. Thus, the core is used for vertical circulation and housing M&E services. Moreover, in terms of climatic considerations, daylighting is able to be achieved for the lower floors. Light is able to enter from the top and is channelled down to the first few storeys.
2) Staggered Column Grid
To avoid the concentration of forces on vertical members, a staggered column grid is generated to allow distribution of load-bearing forces. As such, no single column takes the accumulated load of more than one floor plate and this reduces the amount of material invested in each column. Furthermore, this arrangement of columns allows natural ventilation to occur as wind can be channeled through along the East-West axis between the rooms.
Visualisation of Structural Components
Due to the cantilevered structure, stress will be concentrated on areas where the column meets the beam. This is resolved by identifying and strengthening these areas with more timber members, done with Scan&Solve along with Grasshopper’s image sampling. Thus, the structural framework is built by adding members to the existing column-beam system, allowing weaker wood pieces to gain strength through the layering process.
A von Mises simulation is run on the whole building frame (slabs and columns only) and the B/W stress images of each floor plate is used. Next, beams and transfer beams are then added to reduce the displacement of the floor plate. However, this does not resolve the amount of stress on each column and at the intersections where column and beam meet. Thus, a DouGong column head is layered at the top of every column. As from the results, both stress and displacement values are greatly reduced from the columns and floor slabs. This form of structural assembly is then generated for the subsequent floors.
The construction assembly is via a stacking system where the composite concrete-CLT floor plates are cast in place, followed by the addition of the timber columns and then, the addition of the window frames and railings. Timber Beams are added and capped by modular ceiling panels. Next, the timber column head DouGong is fitted in place with the timber transfer beams at the top to hold the load for the next floor slab.
Sectional Perspective of Structural Assembly
Rendering of Ground Floor lobby and Bicycle Shop
Design studio project by James LAU
Timber has always been difficult to specify – only specifying the cross-section is relatively flexible. Due to natural occurrences of undesirable defects (shakes, wind cracks, upsets, etc.), there is a large amount of waste generated from sawn timber not being able to meet the required specifications. The longer the timber required, the higher the chances of waste being generated. This design exploration seeks to tackle this problem and re-introduce timber architecture aesthetics.
This is a digital fabrication factory that takes inspiration from the traditional 斗拱 “dougong” construction system commonly found in traditional Chinese or Japanese architecture, especially pagodas. This modern interpretation of the system gives use the ability to utilise small standard cross sections available in the local market of the site and create large spans of approximately 35m. By joining the smaller timber components lengthwise using traditional Japanese timber joinery such as the 継手 “tsugite”, usable yields from sawn timber has increased. In addition, the new lengths of timber performs almost as if it was one solid timber sawn from a log.
With the porosity of the roof structure that the dougong system provides, there is the opportunity to bring in daylighting and reduce electrical lighting reliance during the working hours of the factory.
The Dougong Cloud is a long span space where the columns can be freely placed according to the programme of the space and the required span. Through the initial placement of the columns, the span is that calculated from column to column and column to building edge boundary. With the acquired spanning requirements, layers of dougong will be added based on the span and structural performance.
The following is a cropped elevation as seen from the exhibition building.
There is a visible play of varying number of layers and position of columns. It was a conscious design decision to retain the columns instead of using the dougong system to replace the column so as not to waste material through excessive use. In this case, the vertical element of the column would suffice. It is almost as if the column head itself morphs into the roof.
Section Detail of Digital Fabrication Factory
Section Perspective of Digital Fabrication Factory
The following is a diagram of the exploration beginning with the traditional column, beam construction concept. It is then replaced with the standard timber sections and tested using Autodesk Simulation 360 to understand its performance.
Roof layers: 6
Dougong Layers: 8
Step size: 1500
Roof layers: 6
Dougong Layers: 6
Step size: 2000
Roof layers: 6
Dougong Layers: 5
Step size: 2500
There were three basic column configuration to test the structural properties of this system when being used for the following programme respectively: digital fabrication factory, traditional timber factory and exhibition hall.
Each building is then modelled in Rhino and Grasshopper. Subsequently, simulations are then executed with the same 3D model, followed by fabrication.
The lux levels in the building is simulated using the weather file for the site and the average was taken across the year in an overcast condition. Within the space, the average lowest lux levels was found to be approximately 1200 lux. This proves that during the operation hours when daylight is available, for the tasks in this space, daylighting itself is more than sufficient.
Autodesk Simulation Mechanical 360
Displacement [Isometric, Front, Right]
Most of the displacement occur as the beam spans away from the column, where the further it spans, the greater the displacement. However, the maximum displacement that occurred in the direction of the 35m span is within the threshold of 30mm.
von Mises [Isometric, Front, Right]
Most of the stresses occur in the column and in the furthest cantilevered parts of the structure. The maximum stress still falls well within the mechanical properties of teak.
Traditional Timber Factory
Displacement [Isometric, Front, Right]
Displacement [Isometric, Front, Right]
In the application of this technology, there is a chamber where the reaction occurs to provide the nourishments. The limiting supplies are the carbon dioxide, water, surface areas of reaction on the catalyst for reaction and the intensity of sunlight received. The first three are easily solved by:
Firstly, providing a scrub, which consistently extracts carbon dioxide from the surrounding air and resupplies it to the chamber.
Secondly, providing a constant flow of purified water pumped up from Irvin’s water collection intervention.
Thirdly, breaking the catalyst down into powdered form to maximize the amount of surface exposure.
This leaves the last factor, which studies of the natural photosynthesis has shown that rate of photosynthesis is directly proportionate to the intensity of sunlight received. The intensity of sunlight received is only partially controllable, where the supply is natural and the receiving surface depends on the geometry of the exposed surface.
In this exercise, I am attempting to prototype the clear surface to test out which characteristics of the geometry enables us to maximize the potential to receive sunlight.
I am to design digitally based on research on other technologies which require maximum sunlight intensity, namely the PV cells development, and then to send the models for computer numeric controlled milling to create a mould using MDF in preparation for vacuum forming using PVC. During some initial attempts, I have discovered that the vacuum forming is quite accurate in capturing even the texture of the mould, hence there was some manual preparation before we can actually send the mould for vacuum forming. The PVC geometry is then tested with a Lux Meter on top and below to get a relative difference in Sunlight absorbed by the surface. Comparing the amount of light diffracted and converted to heat within the material might not give us an accurate numerical result but it provides us with a base value that can be compared with other PVC geometries’ potentials.
The very initial design was a quick decision to develop a pod-like chamber to run water, carbon dioxide and metal catalyst through to maximize the other supplies besides sun intensity. Inspiration was taken from the natural process of photosynthesis where there are pod-like cells called the chlorophylls aligned side by side, housing the functions required to convert the supplies into glucose and oxygen. The design was generic and did not respond to it’s surroundings.
With this, I started off by looking more into the natural geometries of plants, since they grow in the manner to maximise the receiving of what they need. I chose 2 characteristics of geometries that I was to test on: Fractal Geometry and the Fibonacci Sequence (Golden Curves)
Taking into account that to maximise sunlight intensity received, one important factor is that maximum sunlight is obtained when surface is perpendicular to the sun rays, hence the exposure to the sun path, East to West is important.
Also the Sunlight intensity near mid-day is highest and hence this should be maximized as well. Hence the base geometry is a wave-like surface, with larger amounts of surfaces facing up to maximise mid-day sunlight.
2 more iterations were created to test the effects of fractal growth. In this case, the base geometry is re-panelled onto itself into a more complex geometry. This is repeated again for a 3rd time.
Fibonacci Sequence (Golden Curves):
These geometries are found in broccoli, sunflowers and other plants. The lines subdivides the surfaces into smaller parts to increase the overall surface exposure. The Golden Curve is generated using the Fibonacci sequence of numbers to dictate the size of squares aligned next to each other in a circular manner. A quarter circle is drawn using the points with the longest distance within two adjacent squares. In nature, this repeats itself fractally as well where the fibonacci sequence is used to decide the number of curves repeated radially in one direction and the next number for the curves in the opposite direction to provide the subdivisions for example the sunflower.
Fractal Geometry has a great impact on surface exposure, leading to higher sunlight received. From the experiment, layer 1 absorbed 50 Lux
Layer 2 absorbed 60 Lux
Layer 3 absorbed 80 Lux
This aligns with the tests that MIT did with accordion shaped PV Cells where they managed to increase the amount of energy generated by increasing surface exposure by mimicking the folds on an accordion instrument.
The golden curve is a natural occurring method for subdivision, based on how cells regenerate itself. It is a supporting element to fractal growth but is not entirely affected by the need to maximize sunlight.
From the experiment, the reading is 100 Lux
After the CNC milling is done, the mould needs to be processed such that the surface is smooth in order to replicate the transparent nature of the reaction chamber when placed for vacuum forming. I have picked up the industrial design method of using putty and primer to coat the top surface layer of the mould to create an extremely smooth surface.
This is done by applying an initial layer of putty on a loose particle free MDF surface to fill up the gaps in the material. The model is sanded down using a 400 grade sandpaper after the putty dries to remove any unevenness. A second layer of putty is applied across to coat the entire model. It is re-sanded down to create a base for priming. after sanding it down till relatively smooth surface, spray coat a layer of primer and leave to dry for about 30 minutes. Sand it down using the 400 grade sandpaper until surface is even and that there is a slight gloss. This step can be repeated 2 to 3 times to achieve a better consistency. Sand it further with a grade 1500 sandpaper until surface is glossy. The mould is ready for vacuum forming.
While attempting to vacuum form, there was an issue with a loose table template, which causes the air to leak and creates an issue that the PVC does not wrap around the surface of the mould properly. Check for leaks and in this cause, the table top template is not fitted properly. The PVC should cling onto the form nicely.
When heating the PVC, in my case a 1mm PVC, heat it for about 150 seconds and because of the ability for pvc to stretch for quite a lot, try not to use the air blast to prestretch the pvc. try to raise the mould about 10 cm below the PVC before starting the vacuum and raising the table. There was an attempt when i pre stretched the PVC and the mould was too far from it that when I started the vacuum, the pvc overstretched and by the time the table is raised to the top, the PVC folded and some of it went under the table top template. causing the PVC and the mould to be stuck on the machine.
Before setting the mould for vacuum forming, note that PVC then to cling tightly to the mould and when air cooled, it contracts further and have an even firmer grip onto the mould. MDF being of a compressed powder form, does not provide firm adhesion between itself and the putty/primer. In my attempts, the putty and primer became stuck onto the PVC.
I have done all my prototypes the same way but after the exercise, I realised that I should have coated the top of the primer with some lubricating oil, like WD40, so that the mould can be easily extracted from the PVC.
In the experiment to test the Sunlight, be sure to take readings on top and below the PVC quickly because the change in cloud cover might influence the readings. It would be better if there were 2 Lux Meter and that the readings are taken simultaneously.
The basis for my design was a deployable structure with the express purpose of aid relief. Therefore, the development of my design was focused specifically towards three key issues: it had to be compact and easily transportable when in storage, quickly deployed with minimum of action, and stable when fully open. These three drivers would direct the formation of material, joints and assembly concerns.
The inspiration for the structure was a portable folding stool, with a swivelling mechanism that resulted in a stable tripod base when deployed, but a compact bundle when closed. The spaces created beneath the legs, as well as the span created by the seat, was the suggestions of a possibly useful space that interested me in further exploration.
The final purpose of such a tripod was to provide an elevated rail on which food cargo can be suspended, for ease of movement and to facilitate distribution. Thus, the carrying capacity of the structure was a big issue that had to be addressed.
The aim of the first prototype was to explore if it was possible to have four legs, because my initial hypothesis was that the four point, square footprint would be more helpful in organising space, when multiple structures are used together. From there, I hoped to examine the space created underneath the legs, to develop the use of the structure further.
From the swivel motion of the stool, I designed a joint that I hoped would allow the same freedom of movement while being stronger when scaled up to the size of a full tent. The idea was that the collar would restrict movement to only the basic motion needed, to prevent wobbling and destabilization. Furthermore, it was supposed to take the reciprocal arrangement of the legs and direct them through the collar in a controlled manner, to ensure loads are properly distributed.
The first prototype showed that the motion and deployment was possible and functioned as intended. However the tubular legs of the first prototype were inefficient and bulky at full scale. Thus, I redesigned them with more depth and less width, with a rectangular profile which would hopefully be more efficient in carrying loads. Next, to further reduce weight I also cut out circular sections from the legs, which gave the structure some added functionality bonuses, since I believed I could develop those areas as attachment points for additional structures.
However, the next problem was that fitting a collar around the new legs would re-introduce the bulkiness problem, as the diameter required to fit around the angular centre portion would be proportionally much larger than the slim legs, especially when closed, which is disadvantageous for my intended usage.
Therefore, I decided to do away with the collar, and create the reciprocal mechanism directly into the form of the legs.
I introduced notches into the legs, which, when fully opened, would provide a wider surface for each leg to rest on the next.
This model was laser cut from 5mm acrylic. The test was to ensure that the notches functioned as intended, and that there were no collisions in the movement. At a 5mm thickness and a 1:20 scale, the model accurate represented the actual dimensions of the full scale structure.
From this model, I observed that:
- The reciprocal structure worked as intended. The legs properly leaned on one another, creating a complete load bearing structure.
- The central axis does not take any load, once the legs fully lock into place. This is assuming no loads are applied at the center, which is my intention. The roof structure was intended to continue from the upper portion of the legs. However, the axis would have to be very exact in limiting the legs to prevent wobbling while they are in the process of opening.
- The legs were not thick enough. The actual structure was intended to be aluminium, so in digital form, the slender legs seemed strong enough. But the physical model showed, especially next to scale human figures, that the legs would need to be thicker.
- The length of the legs could be shortened without compromising functionality.
- Most importantly, the notches would have to be wider. The model had a tendency to slip, and once the reciprocal pattern was broken, the structure would collapse. Thus, a larger contact area between each leg was necessary.
Lastly, the choice of material for the model was, in my opinion, not satisfactory in conveying the structural strength and stability of the intended usage. Thus, I decided to construct the third prototype out of aluminium.
One more major benefit of using aluminium would be that I could also accurately show the I-beam profile that I intend for the legs, to show the structural strength of such a form.
The aluminium was purchased from http://www.teckcheong.com/
Rectangular flat bar profile, 76.2mm width by 9.53mm height. The minimum length sold was 6100mm, which they cut into 1 meter portions for me.
The plan was to use the CNC router to hollow out the I-beam portion, and then cut out the outer shape, to save time. Also, I decided not to fully cut the circular portions out either, to save time as well.
All these time saving measure had to be taken because cutting aluminium is a slow process. The stepover for the 6mm drill bit was limited to a maximum of 50% of the width, 3mm, and the pass depth for each run was a tiny 0.5mm. For a 3mm cut, this was 6 passes, multiplied by two sides. To Cut entirely through the 9.5mm would be a significant increase
The machine used was the Versatil 2500.
Because my material was so small, it had to be attached to a wooden surface for the vacuum suction to properly hold in place.
There were two major concerns while cutting:
- That the material stay firmly still
- That the drill bit stay clean to cut properly
For both these concerns, overheating was the main culprit. Despite being air cooled, the friction from the drilling still creates a lot of heat, resulting in several negative outcomes.
Firstly, the metal expands when heated. Very slightly, but enough to shift it from its fastenings. Coupled with the movements of the drill bit, the material began to move when the drill bit cut along the narrow side of the metal. Furthermore, the vibrations loosened the screws, and the metal began to move significantly. This was addressed by adding more screws around the edge of the metal, wedging it in place.
The expansion also created a slight bulge in the middle, as the metal bar curved outward, varying the height of the cut. This was a minor issue, which did not drastically alter the final outcome.
A more major concern was that the heat also softens the aluminium, creating a build-up of melted aluminium along the cutting edge of the drill bit.
This is problematic because it reduces the cutting power, and also creates even more friction, heating up the metal further. If the cutting edge gets too clogged that it cannot cut, but the machine still tries to move the drill bit, DISASTER STRIKES.
The combination of all the abovementioned factors meant that the drill bit was cutting significantly deeper than 0.5mm each pass, and bit broke. In the above picture you can see the jagged, uneven cut resulting from excessive movement.
Thankfully, the bit broke high up, above the cutting edge, with enough space that it could still be reattached and used. Learning from the mistakes of the first run, the metal was more securely fastened, the cut speed was reduced, WD40 was used to lubricate the drill, and frequent pauses were made to ensure the drill bit was clean. Subsequent cuts were much cleaner, and without incident. Time taken: 4 hours per run, four runs total, for a combined time of about 16 hours cutting time.
Cutting out the edges took another hour per leg, as the metal saw cuts at an agonizingly slow pace. The edges were sanded down, and the entire leg was sandblasted to remove blemishes and smooth down the surface. The central axis was 3D printed, designed to be assembled with the metal legs in place.
The final model:
The full structure consists of a roof as well. The deployable motion of the roof was created in Grasshopper, but translating that movement into physical joints was a manual process.
The overall roof consists of four portions, which are basically a series of interlocking C-channels, on which each row of segments can slide along. The intention was to create greater coverage when fully extended, but a compact form when closed. Thus, the C-channels allow for a collapsible arrangement to maximise the difference between opened/closed positions.
In reality, the joint would likely have to consist of a rolling wheel mounted on a ball-and-socket joint, to allow for the freedom of movement required. In the model, this was achieved using simpler T-shaped joints that slotted into the C-channel. The 0.5mm gap required for 3D printing moving parts is responsible for this allowance. To save costs, the final, longest segments of the roof were not printed. Instead, sockets were prepared where I could attach on plastic styrene rods.
Full deployment sequence: